Book Matlab Document Stats
Uploaded by
Daxs WangBook Matlab Document Stats
Uploaded by
Daxs WangStatistics Toolbox™ 7
User’s Guide 4.5.7.9.10.11.12
How to Contact The MathWorks
www.mathworks.com Web
comp.soft-sys.matlab Newsgroup
www.mathworks.com/contact_TS.html Technical Support
suggest@mathworks.com Product enhancement suggestions
bugs@mathworks.com Bug reports
doc@mathworks.com Documentation error reports
service@mathworks.com Order status, license renewals, passcodes
info@mathworks.com Sales, pricing, and general information
508-647-7000 (Phone)
508-647-7001 (Fax)
The MathWorks, Inc.
3 Apple Hill Drive
Natick, MA 01760-2098
For contact information about worldwide offices, see the MathWorks Web site.
Statistics Toolbox™ User’s Guide
© COPYRIGHT 1993–2010 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used
or copied only under the terms of the license agreement. No part of this manual may be photocopied or
reproduced in any form without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation
by, for, or through the federal government of the United States. By accepting delivery of the Program
or Documentation, the government hereby agrees that this software or documentation qualifies as
commercial computer software or commercial computer software documentation as such terms are used
or defined in FAR 12.212, DFARS Part 227.72, and DFARS 252.227-7014. Accordingly, the terms and
conditions of this Agreement and only those rights specified in this Agreement, shall pertain to and govern
the use, modification, reproduction, release, performance, display, and disclosure of the Program and
Documentation by the federal government (or other entity acquiring for or through the federal government)
and shall supersede any conflicting contractual terms or conditions. If this License fails to meet the
government’s needs or is inconsistent in any respect with federal procurement law, the government agrees
to return the Program and Documentation, unused, to The MathWorks, Inc.
Trademarks
MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See
www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand
names may be trademarks or registered trademarks of their respective holders.
Patents
The MathWorks products are protected by one or more U.S. patents. Please see
www.mathworks.com/patents for more information.
Revision History
September 1993 First printing Version 1.0
March 1996 Second printing Version 2.0
January 1997 Third printing Version 2.11
November 2000 Fourth printing Revised for Version 3.0 (Release 12)
May 2001 Fifth printing Minor revisions
July 2002 Sixth printing Revised for Version 4.0 (Release 13)
February 2003 Online only Revised for Version 4.1 (Release 13.0.1)
June 2004 Seventh printing Revised for Version 5.0 (Release 14)
October 2004 Online only Revised for Version 5.0.1 (Release 14SP1)
March 2005 Online only Revised for Version 5.0.2 (Release 14SP2)
September 2005 Online only Revised for Version 5.1 (Release 14SP3)
March 2006 Online only Revised for Version 5.2 (Release 2006a)
September 2006 Online only Revised for Version 5.3 (Release 2006b)
March 2007 Eighth printing Revised for Version 6.0 (Release 2007a)
September 2007 Ninth printing Revised for Version 6.1 (Release 2007b)
March 2008 Online only Revised for Version 6.2 (Release 2008a)
October 2008 Online only Revised for Version 7.0 (Release 2008b)
March 2009 Online only Revised for Version 7.1 (Release 2009a)
September 2009 Online only Revised for Version 7.2 (Release 2009b)
March 2010 Online only Revised for Version 7.3 (Release 2010a)
Contents
Getting Started
1
Product Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2
Organizing Data
2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
MATLAB Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
Numerical Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
Heterogeneous Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-7
Statistical Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-9
Statistical Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-11
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-11
Categorical Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-13
Dataset Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-23
Grouped Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-34
Grouping Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-34
Functions for Grouped Data . . . . . . . . . . . . . . . . . . . . . . . . . 2-35
Using Grouping Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 2-36
Descriptive Statistics
3
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2
v
Measures of Central Tendency . . . . . . . . . . . . . . . . . . . . . . 3-3
Measures of Dispersion ............................ 3-5
Measures of Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7
Resampling Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-9
The Bootstrap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-9
The Jackknife . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-12
Parallel Computing Support for Resampling Methods . . . . 3-13
Data with Missing Values . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-16
Statistical Visualization
4
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2
Scatter Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-3
Box Plots ......................................... 4-6
Distribution Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-8
Normal Probability Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-8
Quantile-Quantile Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-10
Cumulative Distribution Plots . . . . . . . . . . . . . . . . . . . . . . . 4-12
Other Probability Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-14
Probability Distributions
5
Using Probability Distributions . . . . . . . . . . . . . . . . . . . . . 5-2
vi Contents
Supported Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3
Parametric Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-4
Nonparametric Distributions . . . . . . . . . . . . . . . . . . . . . . . . 5-8
Working with Distributions Through GUIs . . . . . . . . . . . 5-9
Exploring Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9
Modeling Your Data Using the Distribution Fitting GUI . . 5-11
Visually Exploring Random Number Generation . . . . . . . . 5-49
Statistics Toolbox Distribution Functions . . . . . . . . . . . 5-52
Probability Density Functions . . . . . . . . . . . . . . . . . . . . . . . 5-52
Cumulative Distribution Functions . . . . . . . . . . . . . . . . . . . 5-62
Inverse Cumulative Distribution Functions . . . . . . . . . . . . 5-66
Distribution Statistics Functions . . . . . . . . . . . . . . . . . . . . . 5-68
Distribution Fitting Functions . . . . . . . . . . . . . . . . . . . . . . . 5-70
Negative Log-Likelihood Functions . . . . . . . . . . . . . . . . . . . 5-77
Random Number Generators . . . . . . . . . . . . . . . . . . . . . . . . 5-80
Using Probability Distribution Objects . . . . . . . . . . . . . . 5-84
Using Distribution Objects . . . . . . . . . . . . . . . . . . . . . . . . . . 5-84
What are Objects? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-85
Creating Distribution Objects . . . . . . . . . . . . . . . . . . . . . . . 5-88
Object-Supported Distributions . . . . . . . . . . . . . . . . . . . . . . 5-89
Performing Calculations Using Distribution Objects . . . . . 5-90
Capturing Results Using Distribution Objects . . . . . . . . . . 5-97
Probability Distributions Used for Multivariate
Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-99
Gaussian Mixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-99
Copulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-107
Random Number Generation
6
Generating Random Data . . . . . . . . . . . . . . . . . . . . . . . . . . 6-2
Random Number Generation Functions . . . . . . . . . . . . . 6-3
vii
Common Generation Methods . . . . . . . . . . . . . . . . . . . . . . 6-5
Direct Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-5
Inversion Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-7
Acceptance-Rejection Methods . . . . . . . . . . . . . . . . . . . . . . . 6-9
Parallel Computing Support for Random Number
Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-13
What is Parallel Computing? . . . . . . . . . . . . . . . . . . . . . . . . 6-13
Reproducing Computations . . . . . . . . . . . . . . . . . . . . . . . . . 6-13
Assigning Random Number Generators . . . . . . . . . . . . . . . 6-14
Representing Sampling Distributions Using Markov
Chain Samplers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-15
Using the Metropolis-Hastings Algorithm . . . . . . . . . . . . . . 6-15
Using Slice Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-16
Generating Quasi-Random Numbers . . . . . . . . . . . . . . . . 6-17
Quasi-Random Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-17
Quasi-Random Point Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-18
Quasi-Random Streams . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-25
Generating Data Using Flexible Families of
Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-27
Pearson and Johnson Systems . . . . . . . . . . . . . . . . . . . . . . . 6-27
Generating Data Using the Pearson System . . . . . . . . . . . . 6-28
Generating Data Using the Johnson System . . . . . . . . . . . 6-30
Hypothesis Tests
7
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-2
Hypothesis Test Terminology . . . . . . . . . . . . . . . . . . . . . . . 7-3
Hypothesis Test Assumptions . . . . . . . . . . . . . . . . . . . . . . . 7-5
Example: Hypothesis Testing . . . . . . . . . . . . . . . . . . . . . . . 7-7
viii Contents
Available Hypothesis Tests . . . . . . . . . . . . . . . . . . . . . . . . . 7-13
Analysis of Variance
8
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2
ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-3
One-Way ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-3
Two-Way ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-9
N-Way ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-12
Other ANOVA Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-26
Analysis of Covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-27
Nonparametric Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-35
MANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-39
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-39
ANOVA with Multiple Responses . . . . . . . . . . . . . . . . . . . . 8-39
Regression Analysis
9
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-2
Linear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3
Linear Regression Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3
Multiple Linear Regression . . . . . . . . . . . . . . . . . . . . . . . . . 9-8
Robust Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-14
Stepwise Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-19
Ridge Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-29
Partial Least Squares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-32
Polynomial Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-37
Response Surface Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-45
Generalized Linear Models . . . . . . . . . . . . . . . . . . . . . . . . . . 9-52
Multivariate Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-57
ix
Nonlinear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-58
Nonlinear Regression Models . . . . . . . . . . . . . . . . . . . . . . . . 9-58
Parametric Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-59
Mixed-Effects Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-64
Regression Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-94
Multivariate Methods
10
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-2
MDS Multidimensional Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . 10-3
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-3
Classical Multidimensional Scaling . . . . . . . . . . . . . . . . . . . 10-3
Nonclassical Multidimensional Scaling . . . . . . . . . . . . . . . . 10-8
Nonmetric Multidimensional Scaling . . . . . . . . . . . . . . . . . 10-10
Procrustes Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-14
Comparing Landmark Data . . . . . . . . . . . . . . . . . . . . . . . . . 10-14
Data Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-14
Preprocessing Data for Accurate Results . . . . . . . . . . . . . . 10-15
Example: Comparing Handwritten Shapes . . . . . . . . . . . . . 10-16
Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-23
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-23
Sequential Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . 10-23
Feature Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-28
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-28
Nonnegative Matrix Factorization . . . . . . . . . . . . . . . . . . . . 10-28
Principal Component Analysis . . . . . . . . . . . . . . . . . . . . . . . 10-31
Factor Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-45
x Contents
Cluster Analysis
11
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-2
Hierarchical Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3
Algorithm Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3
Similarity Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-4
Linkages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-6
Dendrograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-8
Verifying the Cluster Tree . . . . . . . . . . . . . . . . . . . . . . . . . . 11-10
Creating Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-16
K-Means Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-21
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-21
Creating Clusters and Determining Separation . . . . . . . . . 11-22
Determining the Correct Number of Clusters . . . . . . . . . . . 11-23
Avoiding Local Minima . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-26
Gaussian Mixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . 11-28
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-28
Clustering with Gaussian Mixtures . . . . . . . . . . . . . . . . . . . 11-28
Classification
12
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-2
Discriminant Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-3
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-3
Example: Discriminant Analysis . . . . . . . . . . . . . . . . . . . . . 12-3
Naive Bayes Classification . . . . . . . . . . . . . . . . . . . . . . . . . 12-6
Supported Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-6
Classification Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-9
xi
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-9
Example: Classification Trees . . . . . . . . . . . . . . . . . . . . . . . 12-9
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-13
Classification Using Nearest Neighbors . . . . . . . . . . . . . . 12-14
Pairwise Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-14
k-Nearest Neighbor Search . . . . . . . . . . . . . . . . . . . . . . . . . 12-17
Regression and Classification by Bagging Decision
Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-30
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-30
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-31
Regression of Insurance Risk Rating for Car Imports . . . . 12-31
Classifying Radar Returns for Ionosphere Data . . . . . . . . . 12-40
Plotting a Performance Curve . . . . . . . . . . . . . . . . . . . . . . . 12-49
Performance Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-53
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-53
What are ROC Curves? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-53
Evaluating Classifier Performance Using perfcurve . . . . . 12-53
Markov Models
13
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2
Markov Chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-3
Hidden Markov Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-5
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-5
Analyzing Hidden Markov Models . . . . . . . . . . . . . . . . . . . . 13-7
xii Contents
Design of Experiments
14
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-2
Full Factorial Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-3
Multilevel Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-3
Two-Level Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-4
Fractional Factorial Designs . . . . . . . . . . . . . . . . . . . . . . . 14-5
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-5
Plackett-Burman Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-5
General Fractional Designs . . . . . . . . . . . . . . . . . . . . . . . . . 14-6
Response Surface Designs . . . . . . . . . . . . . . . . . . . . . . . . . . 14-9
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-9
Central Composite Designs . . . . . . . . . . . . . . . . . . . . . . . . . 14-9
Box-Behnken Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-13
D-Optimal Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-15
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-15
Generating D-Optimal Designs . . . . . . . . . . . . . . . . . . . . . . 14-16
Augmenting D-Optimal Designs . . . . . . . . . . . . . . . . . . . . . 14-19
Specifying Fixed Covariate Factors . . . . . . . . . . . . . . . . . . . 14-20
Specifying Categorical Factors . . . . . . . . . . . . . . . . . . . . . . . 14-21
Specifying Candidate Sets . . . . . . . . . . . . . . . . . . . . . . . . . . 14-21
Statistical Process Control
15
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-2
Control Charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-3
Capability Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-6
xiii
Function Reference
16
File I/O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-2
Data Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-3
Categorical Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-3
Dataset Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-6
Grouped Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-7
Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-8
Summaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-8
Measures of Central Tendency . . . . . . . . . . . . . . . . . . . . . . . 16-8
Measures of Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-8
Measures of Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-9
Statistics Resampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-9
Data with Missing Values . . . . . . . . . . . . . . . . . . . . . . . . . . 16-9
Data Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-10
Statistical Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-11
Distribution Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-11
Scatter Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-12
ANOVA Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-12
Regression Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-13
Multivariate Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-13
Cluster Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-13
Classification Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-14
DOE Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-14
SPC Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-14
Probability Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-15
Distribution Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-15
Distribution Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-16
Probability Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-17
Cumulative Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-19
Inverse Cumulative Distribution . . . . . . . . . . . . . . . . . . . . . 16-21
Distribution Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-23
Distribution Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-24
Negative Log-Likelihood . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-26
Random Number Generators . . . . . . . . . . . . . . . . . . . . . . . . 16-26
Quasi-Random Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-28
Piecewise Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-29
xiv Contents
Hypothesis Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-31
Analysis of Variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-32
ANOVA Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-32
ANOVA Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-32
Regression Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-33
Regression Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-33
Linear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-34
Nonlinear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-35
Regression Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-35
Ensemble Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-36
Multivariate Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-38
Multivariate Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-38
Multidimensional Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-38
Procrustes Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-38
Feature Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-39
Feature Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-39
Cluster Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-40
Cluster Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-40
Hierarchical Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-40
K-Means Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-41
Gaussian Mixture Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-41
Model Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-41
Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-42
Classification Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-42
Discriminant Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-42
Classification Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-42
Naive Bayes Classification . . . . . . . . . . . . . . . . . . . . . . . . . 16-43
Ensemble Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-44
Hidden Markov Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-46
Design of Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-47
DOE Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-47
Full Factorial Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-47
xv
Fractional Factorial Designs . . . . . . . . . . . . . . . . . . . . . . . . 16-48
Response Surface Designs . . . . . . . . . . . . . . . . . . . . . . . . . . 16-48
D-Optimal Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-48
Latin Hypercube Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-48
Quasi-Random Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-49
Statistical Process Control . . . . . . . . . . . . . . . . . . . . . . . . . 16-51
SPC Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-51
SPC Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-51
GUIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-52
Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-53
Class Reference
17
Data Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-2
Categorical Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-2
Dataset Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-2
Probability Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-3
Distribution Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-3
Quasi-Random Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-3
Piecewise Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-4
Regression Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-4
Regression Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-4
Ensemble Method Classes . . . . . . . . . . . . . . . . . . . . . . . . . . 17-4
Gaussian Mixture Models .......................... 17-4
Model Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-5
Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-5
Classification Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-5
Naive Bayes Classification . . . . . . . . . . . . . . . . . . . . . . . . . 17-5
xvi Contents
Ensemble Method Classes . . . . . . . . . . . . . . . . . . . . . . . . . . 17-5
Quasi-Random Design of Experiments . . . . . . . . . . . . . . . 17-6
Functions — Alphabetical List
18
Data Sets
A
Distribution Reference
B
Bernoulli Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-3
Definition of the Bernoulli Distribution . . . . . . . . . . . . . . . . B-3
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-3
Beta Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-4
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-4
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-4
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-5
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-6
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-6
Binomial Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-7
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-7
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-7
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-8
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-9
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-9
Birnbaum-Saunders Distribution . . . . . . . . . . . . . . . . . . . B-10
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-10
xvii
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-10
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-11
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-11
Chi-Square Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-12
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-12
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-12
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-13
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-13
Copulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-14
Custom Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-15
Exponential Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . B-16
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-16
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-16
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-16
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-17
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-18
Extreme Value Distribution . . . . . . . . . . . . . . . . . . . . . . . . B-19
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-19
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-19
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-21
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-22
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-24
F Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-25
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-25
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-25
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-26
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-26
Gamma Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-27
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-27
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-27
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-28
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-29
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-29
xviii Contents
Gaussian Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-30
Gaussian Mixture Distributions . . . . . . . . . . . . . . . . . . . . . B-31
Generalized Extreme Value Distribution . . . . . . . . . . . . B-32
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-32
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-32
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-33
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-34
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-36
Generalized Pareto Distribution . . . . . . . . . . . . . . . . . . . . B-37
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-37
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-37
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-38
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-39
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-40
Geometric Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-41
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-41
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-41
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-41
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-42
Hypergeometric Distribution . . . . . . . . . . . . . . . . . . . . . . . B-43
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-43
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-43
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-44
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-44
Inverse Gaussian Distribution . . . . . . . . . . . . . . . . . . . . . . B-45
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-45
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-45
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-45
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-45
Inverse Wishart Distribution . . . . . . . . . . . . . . . . . . . . . . . B-46
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-46
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-46
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-46
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-47
xix
Johnson System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-48
Logistic Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-49
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-49
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-49
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-49
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-49
Loglogistic Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-50
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-50
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-50
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-50
Lognormal Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-51
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-51
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-51
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-52
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-53
Multinomial Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . B-54
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-54
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-54
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-54
Multivariate Gaussian Distribution . . . . . . . . . . . . . . . . . B-57
Multivariate Normal Distribution . . . . . . . . . . . . . . . . . . . B-58
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-58
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-58
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-59
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-63
Multivariate t Distribution . . . . . . . . . . . . . . . . . . . . . . . . . B-64
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-64
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-64
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-65
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-69
Nakagami Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-70
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-70
xx Contents
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-70
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-70
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-71
Negative Binomial Distribution . . . . . . . . . . . . . . . . . . . . . B-72
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-72
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-72
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-73
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-75
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-75
Noncentral Chi-Square Distribution . . . . . . . . . . . . . . . . . B-76
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-76
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-76
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-77
Noncentral F Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . B-78
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-78
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-78
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-79
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-79
Noncentral t Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . B-80
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-80
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-80
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-81
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-81
Nonparametric Distributions . . . . . . . . . . . . . . . . . . . . . . . B-82
Normal Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-83
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-83
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-83
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-84
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-85
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-85
Pareto Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-86
Pearson System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-87
xxi
Piecewise Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-88
Poisson Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-89
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-89
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-89
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-90
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-90
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-90
Rayleigh Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-91
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-91
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-91
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-92
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-92
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-92
Rician Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-93
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-93
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-93
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-93
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-94
Student’s t Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-95
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-95
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-95
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-96
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-96
t Location-Scale Distribution . . . . . . . . . . . . . . . . . . . . . . . B-97
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-97
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-97
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-97
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-98
Uniform Distribution (Continuous) . . . . . . . . . . . . . . . . . . B-99
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-99
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-99
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-99
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-99
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-100
xxii Contents
Uniform Distribution (Discrete) . . . . . . . . . . . . . . . . . . . . . B-101
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-101
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-101
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-101
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-102
Weibull Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-103
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-103
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-103
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-104
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-104
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-105
Wishart Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-106
Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-106
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-106
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-107
See Also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-107
Bibliography
C
Index
xxiii
xxiv Contents
1
Getting Started
1 Getting Started
Product Overview
Statistics Toolbox™ software extends MATLAB® to support a wide range of
common statistical tasks. The toolbox contains two categories of tools:
• Building-block statistical functions for use in MATLAB programming
• Graphical user interfaces (GUIs) for interactive data analysis
Code for the building-block functions is open and extensible. Use the MATLAB
Editor to review, copy, and edit code for any function. Extend the toolbox by
copying code to new files or by writing files that call toolbox functions.
Toolbox GUIs allow you to perform statistical visualization and analysis
without writing code. You interact with the GUIs using sliders, input fields,
push buttons, etc. and the GUIs automatically call building-block functions.
1-2
2
Organizing Data
• “Introduction” on page 2-2
• “MATLAB Arrays” on page 2-4
• “Statistical Arrays” on page 2-11
• “Grouped Data” on page 2-34
2 Organizing Data
Introduction
MATLAB data is placed into “data containers” in the form of workspace
variables. All workspace variables organize data into some form of array. For
statistical purposes, arrays are viewed as tables of values.
MATLAB variables use different structures to organize data:
• 2-D numerical arrays (matrices) organize observations and measured
variables by rows and columns, respectively. (See “Other Data Structures”
in the MATLAB documentation.)
• Multidimensional arrays organize multidimensional observations or
experimental designs. (See “Multidimensional Arrays” in the MATLAB
documentation.)
• Cell and structure arrays organize heterogeneous data of different types,
sizes, units, etc. (See “Cell Arrays” and “Structures” in the MATLAB
documentation.)
Data types determine the kind of data variables contain. (See “Classes (Data
Types)” in the MATLAB documentation.)
These basic MATLAB container variables are reviewed, in a statistical
context, in the section on “MATLAB Arrays” on page 2-4.
These variables are not specifically designed for statistical data, however.
Statistical data generally involves observations of multiple variables, with
measurements of heterogeneous type and size. Data may be numerical (of
type single or double), categorical, or in the form of descriptive metadata.
Fitting statistical data into basic MATLAB variables, and accessing it
efficiently, can be cumbersome.
Statistics Toolbox software offers two additional types of container variables
specifically designed for statistical data:
• “Categorical Arrays” on page 2-13 accommodate data in the form of discrete
levels, together with its descriptive metadata.
2-2
Introduction
• “Dataset Arrays” on page 2-23 encapsulate heterogeneous data and
metadata, including categorical data, which is accessed and manipulated
using familiar methods analogous to those for numerical matrices.
These statistical container variables are discussed in the section on
“Statistical Arrays” on page 2-11.
2-3
2 Organizing Data
MATLAB Arrays
In this section...
“Numerical Data” on page 2-4
“Heterogeneous Data” on page 2-7
“Statistical Functions” on page 2-9
Numerical Data
MATLAB two-dimensional numerical arrays (matrices) containing statistical
data use rows to represent observations and columns to represent measured
variables. For example,
load fisheriris % Fisher's iris data (1936)
loads the variables meas and species into the MATLAB workspace. The meas
variable is a 150-by-4 numerical matrix, representing 150 observations of 4
different measured variables (by column: sepal length, sepal width, petal
length, and petal width, respectively).
The observations in meas are of three different species of iris (setosa,
versicolor, and virginica), which can be separated from one another using the
150-by-1 cell array of strings species:
2-4
MATLAB® Arrays
setosa_indices = strcmp('setosa',species);
setosa = meas(setosa_indices,:);
The resulting setosa variable is 50-by-4, representing 50 observations of the
4 measured variables for iris setosa.
To access and display the first five observations in the setosa data, use row,
column parenthesis indexing:
SetosaObs = setosa(1:5,:)
SetosaObs =
5.1000 3.5000 1.4000 0.2000
4.9000 3.0000 1.4000 0.2000
4.7000 3.2000 1.3000 0.2000
4.6000 3.1000 1.5000 0.2000
5.0000 3.6000 1.4000 0.2000
The data are organized into a table with implicit column headers “Sepal
Length,” “Sepal Width,” “Petal Length,” and “Petal Width.” Implicit row
headers are “Observation 1,” “Observation 2,” “Observation 3,” etc.
Similarly, 50 observations for iris versicolor and iris virginica can be extracted
from the meas container variable:
versicolor_indices = strcmp('versicolor',species);
versicolor = meas(versicolor_indices,:);
virginica_indices = strcmp('virginica',species);
virginica = meas(virginica_indices,:);
Because the data sets for the three species happen to be of the same size, they
can be reorganized into a single 50-by-4-by-3 multidimensional array:
iris = cat(3,setosa,versicolor,virginica);
The iris array is a three-layer table with the same implicit row and column
headers as the setosa, versicolor, and virginica arrays. The implicit layer
names, along the third dimension, are “Setosa,” “Versicolor,” and “Virginica.”
The utility of such a multidimensional organization depends on assigning
meaningful properties of the data to each dimension.
2-5
2 Organizing Data
To access and display data in a multidimensional array, use parenthesis
indexing, as for 2-D arrays. The following gives the first five observations
of sepal lengths in the setosa data:
SetosaSL = iris(1:5,1,1)
SetosaSL =
5.1000
4.9000
4.7000
4.6000
5.0000
Multidimensional arrays provide a natural way to organize numerical data
for which the observations, or experimental designs, have many dimensions.
If, for example, data with the structure of iris are collected by multiple
observers, in multiple locations, over multiple dates, the entirety of the data
can be organized into a single higher dimensional array with dimensions
for “Observer,” “Location,” and “Date.” Likewise, an experimental design
calling for m observations of n p-dimensional variables could be stored in
an m-by-n-by-p array.
Numerical arrays have limitations when organizing more general statistical
data. One limitation is the implicit nature of the metadata. Another is the
requirement that multidimensional data be of commensurate size across all
dimensions. If variables have different lengths, or the number of variables
differs by layer, then multidimensional arrays must be artificially padded
with NaNs to indicate “missing values.” These limitations are addressed by
dataset arrays (see “Dataset Arrays” on page 2-23), which are specifically
designed for statistical data.
2-6
MATLAB® Arrays
Heterogeneous Data
MATLAB data types include two container variables—cell arrays and
structure arrays—that allow you to combine metadata with variables of
different types and sizes.
The data in the variables setosa, versicolor, and virginica created in
“Numerical Data” on page 2-4 can be organized in a cell array, as follows:
iris1 = cell(51,5,3); % Container variable
obsnames = strcat({'Obs'},num2str((1:50)','%-d'));
iris1(2:end,1,:) = repmat(obsnames,[1 1 3]);
varnames = {'SepalLength','SepalWidth',...
'PetalLength','PetalWidth'};
iris1(1,2:end,:) = repmat(varnames,[1 1 3]);
iris1(2:end,2:end,1) = num2cell(setosa);
iris1(2:end,2:end,2) = num2cell(versicolor);
iris1(2:end,2:end,3) = num2cell(virginica);
iris1{1,1,1} = 'Setosa';
iris1{1,1,2} = 'Versicolor';
iris1{1,1,3} = 'Virginica';
To access and display the cells, use parenthesis indexing. The following
displays the first five observations in the setosa sepal data:
SetosaSLSW = iris1(1:6,1:3,1)
SetosaSLSW =
'Setosa' 'SepalLength' 'SepalWidth'
'Obs1' [ 5.1000] [ 3.5000]
'Obs2' [ 4.9000] [ 3]
'Obs3' [ 4.7000] [ 3.2000]
'Obs4' [ 4.6000] [ 3.1000]
'Obs5' [ 5] [ 3.6000]
Here, the row and column headers have been explicitly labeled with metadata.
To extract the data subset, use row, column curly brace indexing:
2-7
2 Organizing Data
subset = reshape([iris1{2:6,2:3,1}],5,2)
subset =
5.1000 3.5000
4.9000 3.0000
4.7000 3.2000
4.6000 3.1000
5.0000 3.6000
While cell arrays are useful for organizing heterogeneous data, they may
be cumbersome when it comes to manipulating and analyzing the data.
MATLAB and Statistics Toolbox statistical functions do not accept data in the
form of cell arrays. For processing, data must be extracted from the cell array
to a numerical container variable, as in the preceding example. The indexing
can become complicated for large, heterogeneous data sets. This limitation of
cell arrays is addressed by dataset arrays (see “Dataset Arrays” on page 2-23),
which are designed to store general statistical data and provide easy access.
The data in the preceding example can also be organized in a structure array,
as follows:
iris2.data = cat(3,setosa,versicolor,virginica);
iris2.varnames = {'SepalLength','SepalWidth',...
'PetalLength','PetalWidth'};
iris2.obsnames = strcat({'Obs'},num2str((1:50)','%-d'));
iris2.species = {'setosa','versicolor','virginica'};
The data subset is then returned using a combination of dot and parenthesis
indexing:
subset = iris2.data(1:5,1:2,1)
subset =
5.1000 3.5000
4.9000 3.0000
4.7000 3.2000
4.6000 3.1000
5.0000 3.6000
For statistical data, structure arrays have many of the same limitations as
cell arrays. Once again, dataset arrays (see “Dataset Arrays” on page 2-23),
designed specifically for general statistical data, address these limitations.
2-8
MATLAB® Arrays
Statistical Functions
One of the advantages of working in the MATLAB language is that functions
operate on entire arrays of data, not just on single scalar values. The
functions are said to be vectorized. Vectorization allows for both efficient
problem formulation, using array-based data, and efficient computation,
using vectorized statistical functions.
When MATLAB and Statistics Toolbox statistical functions operate on a
vector of numerical data (either a row vector or a column vector), they return
a single computed statistic:
% Fisher's setosa data:
load fisheriris
setosa_indices = strcmp('setosa',species);
setosa = meas(setosa_indices,:);
% Single variable from the data:
setosa_sepal_length = setosa(:,1);
% Standard deviation of the variable:
std(setosa_sepal_length)
ans =
0.3525
When statistical functions operate on a matrix of numerical data, they treat
the columns independently, as separate measured variables, and return a
vector of statistics—one for each variable:
std(setosa)
ans =
0.3525 0.3791 0.1737 0.1054
The four standard deviations are for measurements of sepal length, sepal
width, petal length, and petal width, respectively.
Compare this to
std(setosa(:))
ans =
1.8483
2-9
2 Organizing Data
which gives the standard deviation across the entire array (all measurements).
Compare the preceding statistical calculations to the more generic
mathematical operation
sin(setosa)
This operation returns a 50-by-4 array the same size as setosa. The sin
function is vectorized in a different way than the std function, computing one
scalar value for each element in the array.
MATLAB and Statistics Toolbox statistical functions, like std, must be
distinguished from general mathematical functions like sin. Both are
vectorized, and both are useful for working with array-based data, but
only statistical functions summarize data across observations (rows) while
preserving variables (columns). This property of statistical functions may be
explicit, as with std, or implicit, as with regress. To see how a particular
function handles array-based data, consult its reference page.
MATLAB statistical functions expect data input arguments to be in the form
of numerical arrays. If data is stored in a cell or structure array, it must
be extracted to a numerical array, via indexing, for processing. Statistics
Toolbox functions are more flexible. Many toolbox functions accept data input
arguments in the form of both numerical arrays and dataset arrays (see
“Dataset Arrays” on page 2-23), which are specifically designed for storing
general statistical data.
2-10
Statistical Arrays
Statistical Arrays
In this section...
“Introduction” on page 2-11
“Categorical Arrays” on page 2-13
“Dataset Arrays” on page 2-23
Introduction
As discussed in “MATLAB Arrays” on page 2-4, MATLAB data types include
arrays for numerical, logical, and character data, as well as cell and structure
arrays for heterogeneous collections of data.
Statistics Toolbox software offers two additional types of arrays specifically
designed for statistical data:
• “Categorical Arrays” on page 2-13
• “Dataset Arrays” on page 2-23
Categorical arrays store data with values in a discrete set of levels. Each level
is meant to capture a single, defining characteristic of an observation. If no
ordering is encoded in the levels, the data and the array are nominal. If an
ordering is encoded, the data and the array are ordinal.
Categorical arrays also store labels for the levels. Nominal labels typically
suggest the type of an observation, while ordinal labels suggest the position
or rank.
Dataset arrays collect heterogeneous statistical data and metadata, including
categorical data, into a single container variable. Like the numerical matrices
discussed in “Numerical Data” on page 2-4, dataset arrays can be viewed as
tables of values, with rows representing different observations and columns
representing different measured variables. Like the cell and structure
arrays discussed in “Heterogeneous Data” on page 2-7, dataset arrays can
accommodate variables of different types, sizes, units, etc.
2-11
2 Organizing Data
Dataset arrays combine the organizational advantages of these basic
MATLAB data types while addressing their shortcomings with respect to
storing complex statistical data.
Both categorical and dataset arrays have associated methods for assembling,
accessing, manipulating, and processing the collected data. Basic array
operations parallel those for numerical, cell, and structure arrays.
2-12
Statistical Arrays
Categorical Arrays
• “Categorical Data” on page 2-13
• “Categorical Arrays” on page 2-14
• “Using Categorical Arrays” on page 2-16
Categorical Data
Categorical data take on values from only a finite, discrete set of categories
or levels. Levels may be determined before the data are collected, based on
the application, or they may be determined by the distinct values in the data
when converting them to categorical form. Predetermined levels, such as a
set of states or numerical intervals, are independent of the data they contain.
Any number of values in the data may attain a given level, or no data at all.
Categorical data show which measured values share common levels, and
which do not.
Levels may have associated labels. Labels typically express a defining
characteristic of an observation, captured by its level.
If no ordering is encoded in the levels, the data are nominal. Nominal
labels typically indicate the type of an observation. Examples of nominal
labels are {false, true}, {male, female}, and {Afghanistan, ..., Zimbabwe}.
For nominal data, the numeric or lexicographic order of the labels is
irrelevant—Afghanistan is not considered to be less than, equal to, or greater
than Zimbabwe.
If an ordering is encoded in the levels—for example, if levels labeled “red”,
“green”, and “blue” represent wavelengths—the data are ordinal. Labels
for ordinal levels typically indicate the position or rank of an observation.
Examples of ordinal labels are {0, 1}, {mm, cm, m, km}, and {poor, satisfactory,
outstanding}. The ordering of the levels may or may not correspond to the
numeric or lexicographic order of the labels.
2-13
2 Organizing Data
Categorical Arrays
Categorical data can be represented using MATLAB integer arrays, but
this method has a number of drawbacks. First, it removes all of the useful
metadata that might be captured in labels for the levels. Labels must be
stored separately, in character arrays or cell arrays of strings. Secondly, this
method suggests that values stored in the integer array have their usual
numeric meaning, which, for categorical data, they may not. Finally, integer
types have a fixed set of levels (for example, -128:127 for all int8 arrays),
which cannot be changed.
Categorical arrays, available in Statistics Toolbox software, are specifically
designed for storing, manipulating, and processing categorical data and
metadata. Unlike integer arrays, each categorical array has its own set of
levels, which can be changed. Categorical arrays also accommodate labels for
levels in a natural way. Like numerical arrays, categorical arrays take on
different shapes and sizes, from scalars to N-D arrays.
Organizing data in a categorical array can be an end in itself. Often, however,
categorical arrays are used for further statistical processing. They can be
used to index into other variables, creating subsets of data based on the
category of observation, or they can be used with statistical functions that
accept categorical inputs. For examples, see “Grouped Data” on page 2-34.
Categorical arrays come in two types, depending on whether the collected
data is understood to be nominal or ordinal. Nominal arrays are constructed
with nominal; ordinal arrays are constructed with ordinal. For example,
load fisheriris
ndata = nominal(species,{'A','B','C'});
creates a nominal array with levels A, B, and C from the species data in
fisheriris.mat, while
odata = ordinal(ndata,{},{'C','A','B'});
encodes an ordering of the levels with C < A < B. See “Using Categorical
Arrays” on page 2-16, and the reference pages for nominal and ordinal, for
further examples.
Categorical arrays are implemented as objects of the categorical class.
The class is abstract, defining properties and methods common to both
2-14
Statistical Arrays
the nominal class and ordinal class. Use the corresponding constructors,
nominal or ordinal, to create categorical arrays. Methods of the classes are
used to display, summarize, convert, concatenate, and access the collected
data. Many of these methods are invoked using operations analogous to those
for numerical arrays, and do not need to be called directly (for example, []
invokes horzcat). Other methods, such as reorderlevels, must be called
directly.
2-15
2 Organizing Data
Using Categorical Arrays
This section provides an extended tutorial example demonstrating the use of
categorical arrays with methods of the nominal class and ordinal class.
• “Constructing Categorical Arrays” on page 2-16
• “Accessing Categorical Arrays” on page 2-18
• “Combining Categorical Arrays” on page 2-19
• “Computing with Categorical Arrays” on page 2-20
Constructing Categorical Arrays.
1 Load the 150-by-4 numerical array meas and the 150-by-1 cell array of
strings species:
load fisheriris % Fisher's iris data (1936)
The data are 150 observations of four measured variables (by column
number: sepal length, sepal width, petal length, and petal width,
respectively) over three species of iris (setosa, versicolor, and virginica).
2 Use nominal to create a nominal array from species:
n1 = nominal(species);
3 Open species and n1 side by side in the Variable Editor (see “Viewing and
Editing Workspace Variables with the Variable Editor” in the MATLAB
documentation). Note that the string information in species has been
converted to categorical form, leaving only information on which data share
the same values, indicated by the labels for the levels.
By default, levels are labeled with the distinct values in the data (in this
case, the strings in species). Give alternate labels with additional input
arguments to the nominal constructor:
n2 = nominal(species,{'species1','species2','species3'});
4 Open n2 in the Variable Editor, and compare it with species and n1. The
levels have been relabeled.
2-16
Statistical Arrays
5 Suppose that the data are considered to be ordinal. A characteristic of the
data that is not reflected in the labels is the diploid chromosome count,
which orders the levels corresponding to the three species as follows:
species1 < species3 < species2
Use ordinal to cast n2 as an ordinal array:
o1 = ordinal(n2,{},{'species1','species3','species2'});
The second input argument to ordinal is the same as for nominal—a list
of labels for the levels in the data. If it is unspecified, as above, the labels
are inherited from the data, in this case n2. The third input argument of
ordinal indicates the ordering of the levels, in ascending order.
6 When displayed side by side in the Variable Editor, o1 does not appear any
different than n2. This is because the data in o1 have not been sorted. It
is important to recognize the difference between the ordering of the levels
in an ordinal array and sorting the actual data according to that ordering.
Use sort to sort ordinal data in ascending order:
o2 = sort(o1);
When displayed in the Variable Editor, o2 shows the data sorted by diploid
chromosome count.
7 To find which elements moved up in the sort, use the < operator for ordinal
arrays:
moved_up = (o1 < o2);
The operation returns a logical array moved_up, indicating which elements
have moved up (the data for species3).
8 Use getlabels to display the labels for the levels in ascending order:
labels2 = getlabels(o2)
labels2 =
'species1' 'species3' 'species2'
9 The sort function reorders the display of the data, but not the order of the
levels. To reorder the levels, use reorderlevels:
2-17
2 Organizing Data
o3 = reorderlevels(o2,labels2([1 3 2]));
labels3 = getlabels(o3)
labels3 =
'species1' 'species2' 'species3'
o4 = sort(o3);
These operations return the levels in the data to their original ordering, by
species number, and then sort the data for display purposes.
Accessing Categorical Arrays. Categorical arrays are accessed using
parenthesis indexing, with syntax that parallels similar operations for
numerical arrays (see “Numerical Data” on page 2-4).
Parenthesis indexing on the right-hand side of an assignment is used to
extract the lowest 50 elements from the ordinal array o4:
low50 = o4(1:50);
Suppose you want to categorize the data in o4 with only two levels: low (the
data in low50) and high (the rest of the data). One way to do this is to use an
assignment with parenthesis indexing on the left-hand side:
o5 = o4; % Copy o4
o5(1:50) = 'low';
Warning: Categorical level 'low' being added.
o5(51:end) = 'high';
Warning: Categorical level 'high' being added.
Note the warnings: the assignments move data to new levels. The old levels,
though empty, remain:
getlabels(o5)
ans =
'species1' 'species2' 'species3' 'low' 'high'
The old levels are removed using droplevels:
o5 = droplevels(o5,{'species1','species2','species3'});
Another approach to creating two categories in o5 from the three categories in
o4 is to merge levels, using mergelevels:
2-18
Statistical Arrays
o5 = mergelevels(o4,{'species1'},'low');
o5 = mergelevels(o5,{'species2','species3'},'high');
getlabels(o5)
ans =
'low' 'high'
The merged levels are removed and replaced with the new levels.
Combining Categorical Arrays. Categorical arrays are concatenated using
square brackets. Again, the syntax parallels similar operations for numerical
arrays (see “Numerical Data” on page 2-4). There are, however, restrictions:
• Only categorical arrays of the same type can be combined. You cannot
concatenate a nominal array with an ordinal array.
• Only ordinal arrays with the same levels, in the same order, can be
combined.
• Nominal arrays with different levels can be combined to produce a nominal
array whose levels are the union of the levels in the component arrays.
First use ordinal to create ordinal arrays from the variables for sepal length
and sepal width in meas. Categorize the data as short or long depending on
whether they are below or above the median of the variable, respectively:
sl = meas(:,1); % Sepal length data
sw = meas(:,2); % Sepal width data
SL1 = ordinal(sl,{'short','long'},[],...
[min(sl),median(sl),max(sl)]);
SW1 = ordinal(sw,{'short','long'},[],...
[min(sw),median(sw),max(sw)]);
Because SL1 and SW1 are ordinal arrays with the same levels, in the same
order, they can be concatenated:
S1 = [SL1,SW1];
S1(1:10,:)
ans =
short long
short long
short long
2-19
2 Organizing Data
short long
short long
short long
short long
short long
short short
short long
The result is an ordinal array S1 with two columns.
If, on the other hand, the measurements are cast as nominal, different levels
can be used for the different variables, and the two nominal arrays can still
be combined:
SL2 = nominal(sl,{'short','long'},[],...
[min(sl),median(sl),max(sl)]);
SW2 = nominal(sw,{'skinny','wide'},[],...
[min(sw),median(sw),max(sw)]);
S2 = [SL2,SW2];
getlabels(S2)
ans =
'short' 'long' 'skinny' 'wide'
S2(1:10,:)
ans =
short wide
short wide
short wide
short wide
short wide
short wide
short wide
short wide
short skinny
short wide
Computing with Categorical Arrays. Categorical arrays are used to
index into other variables, creating subsets of data based on the category
of observation, and they are used with statistical functions that accept
categorical inputs, such as those described in “Grouped Data” on page 2-34.
2-20
Statistical Arrays
Use ismember to create logical variables based on the category of observation.
For example, the following creates a logical index the same size as species
that is true for observations of iris setosa and false elsewhere. Recall that
n1 = nominal(species):
SetosaObs = ismember(n1,'setosa');
Since the code above compares elements of n1 to a single value, the same
operation is carried out by the equality operator:
SetosaObs = (n1 == 'setosa');
The SetosaObs variable is used to index into meas to extract only the setosa
data:
SetosaData = meas(SetosaObs,:);
Categorical arrays are also used as grouping variables. The following plot
summarizes the sepal length data in meas by category:
boxplot(sl,n1)
2-21
2 Organizing Data
2-22
Statistical Arrays
Dataset Arrays
• “Statistical Data” on page 2-23
• “Dataset Arrays” on page 2-24
• “Using Dataset Arrays” on page 2-25
Statistical Data
MATLAB data containers (variables) are suitable for completely homogeneous
data (numeric, character, and logical arrays) and for completely heterogeneous
data (cell and structure arrays). Statistical data, however, are often a mixture
of homogeneous variables of heterogeneous types and sizes. Dataset arrays
are suitable containers for this kind of data.
Dataset arrays can be viewed as tables of values, with rows representing
different observations or cases and columns representing different measured
variables. In this sense, dataset arrays are analogous to the numerical
arrays for statistical data discussed in “Numerical Data” on page 2-4. Basic
methods for creating and manipulating dataset arrays parallel the syntax of
corresponding methods for numerical arrays.
While each column of a dataset array must be a variable of a single type,
each row may contain an observation consisting of measurements of different
types. In this sense, dataset arrays lie somewhere between variables that
enforce complete homogeneity on the data and those that enforce nothing.
Because of the potentially heterogeneous nature of the data, dataset arrays
have indexing methods with syntax that parallels corresponding methods for
cell and structure arrays (see “Heterogeneous Data” on page 2-7).
2-23
2 Organizing Data
Dataset Arrays
Dataset arrays are variables created with dataset. For example, the
following creates a dataset array from observations that are a combination of
categorical and numerical measurements:
load fisheriris
NumObs = size(meas,1);
NameObs = strcat({'Obs'},num2str((1:NumObs)','%-d'));
iris = dataset({nominal(species),'species'},...
{meas,'SL','SW','PL','PW'},...
'ObsNames',NameObs);
iris(1:5,:)
ans =
species SL SW PL PW
Obs1 setosa 5.1 3.5 1.4 0.2
Obs2 setosa 4.9 3 1.4 0.2
Obs3 setosa 4.7 3.2 1.3 0.2
Obs4 setosa 4.6 3.1 1.5 0.2
Obs5 setosa 5 3.6 1.4 0.2
When creating a dataset array, variable names and observation names can be
assigned together with the data. Other metadata associated with the array
can be assigned with set and accessed with get:
iris = set(iris,'Description','Fisher''s Iris Data');
get(iris)
Description: 'Fisher's Iris Data'
Units: {}
DimNames: {'Observations' 'Variables'}
UserData: []
ObsNames: {150x1 cell}
VarNames: {'species' 'SL' 'SW' 'PL' 'PW'}
Dataset arrays are implemented as objects of the dataset class. Methods of
the class are used to display, summarize, convert, concatenate, and access
the collected data. Many of these methods are invoked using operations
analogous to those for numerical arrays, and do not need to be called directly
(for example, [] invokes horzcat). Other methods, such as sortrows, must
be called directly.
2-24
Statistical Arrays
Using Dataset Arrays
This section provides an extended tutorial example demonstrating the use of
dataset arrays with methods of the dataset class.
• “Constructing Dataset Arrays” on page 2-25
• “Accessing Dataset Arrays” on page 2-27
• “Combining Dataset Arrays” on page 2-29
• “Removing Observations from Dataset Arrays” on page 2-31
• “Computing with Dataset Arrays” on page 2-31
Constructing Dataset Arrays. Load the 150-by-4 numerical array meas and
the 150-by-1 cell array of strings species:
load fisheriris % Fisher's iris data (1936)
The data are 150 observations of four measured variables (by column number:
sepal length, sepal width, petal length, and petal width, respectively) over
three species of iris (setosa, versicolor, and virginica).
Use dataset to create a dataset array iris from the data, assigning variable
names species, SL, SW, PL, and PW and observation names Obs1, Obs2, Obs3,
etc.:
NumObs = size(meas,1);
NameObs = strcat({'Obs'},num2str((1:NumObs)','%-d'));
iris = dataset({nominal(species),'species'},...
{meas,'SL','SW','PL','PW'},...
'ObsNames',NameObs);
iris(1:5,:)
ans =
species SL SW PL PW
Obs1 setosa 5.1 3.5 1.4 0.2
Obs2 setosa 4.9 3 1.4 0.2
Obs3 setosa 4.7 3.2 1.3 0.2
Obs4 setosa 4.6 3.1 1.5 0.2
Obs5 setosa 5 3.6 1.4 0.2
2-25
2 Organizing Data
The cell array of strings species is first converted to a categorical array of
type nominal before inclusion in the dataset array. For further information
on categorical arrays, see “Categorical Arrays” on page 2-13.
Use set to set properties of the array:
desc = 'Fisher''s iris data (1936)';
units = [{''} repmat({'cm'},1,4)];
info = 'http://en.wikipedia.org/wiki/R.A._Fisher';
iris = set(iris,'Description',desc,...
'Units',units,...
'UserData',info);
Use get to view properties of the array:
get(iris)
Description: 'Fisher's iris data (1936)'
Units: {'' 'cm' 'cm' 'cm' 'cm'}
DimNames: {'Observations' 'Variables'}
UserData: 'http://en.wikipedia.org/wiki/R.A._Fisher'
ObsNames: {150x1 cell}
VarNames: {'species' 'SL' 'SW' 'PL' 'PW'}
get(iris(1:5,:),'ObsNames')
ans =
'Obs1'
'Obs2'
'Obs3'
'Obs4'
'Obs5'
For a table of accessible properties of dataset arrays, with descriptions, see
the reference on the dataset class.
2-26
Statistical Arrays
Accessing Dataset Arrays. Dataset arrays support multiple types of
indexing. Like the numerical matrices described in “Numerical Data” on page
2-4, parenthesis () indexing is used to access data subsets. Like the cell
and structure arrays described in “Heterogeneous Data” on page 2-7, dot .
indexing is used to access data variables and curly brace {} indexing is used
to access data elements.
Use parenthesis indexing to assign a subset of the data in iris to a new
dataset array iris1:
iris1 = iris(1:5,2:3)
iris1 =
SL SW
Obs1 5.1 3.5
Obs2 4.9 3
Obs3 4.7 3.2
Obs4 4.6 3.1
Obs5 5 3.6
Similarly, use parenthesis indexing to assign new data to the first variable
in iris1:
iris1(:,1) = dataset([5.2;4.9;4.6;4.6;5])
iris1 =
SL SW
Obs1 5.2 3.5
Obs2 4.9 3
Obs3 4.6 3.2
Obs4 4.6 3.1
Obs5 5 3.6
Variable and observation names can also be used to access data:
SepalObs = iris1({'Obs1','Obs3','Obs5'},'SL')
SepalObs =
SL
Obs1 5.2
Obs3 4.6
Obs5 5
2-27
2 Organizing Data
Dot indexing is used to access variables in a dataset array, and can be
combined with other indexing methods. For example, apply zscore to the
data in SepalObs as follows:
ScaledSepalObs = zscore(iris1.SL([1 3 5]))
ScaledSepalObs =
0.8006
-1.1209
0.3203
The following code extracts the sepal lengths in iris1 corresponding to sepal
widths greater than 3:
BigSWLengths = iris1.SL(iris1.SW > 3)
BigSWLengths =
5.2000
4.6000
4.6000
5.0000
Dot indexing also allows entire variables to be deleted from a dataset array:
iris1.SL = []
iris1 =
SW
Obs 1 3.5
Obs 2 3
Obs 3 3.2
Obs 4 3.1
Obs 5 3.6
Dynamic variable naming works for dataset arrays just as it does for structure
arrays. For example, the units of the SW variable are changed in iris1 as
follows:
varname = 'SW';
iris1.(varname) = iris1.(varname)*10
iris1 =
SW
Obs1 35
Obs2 30
2-28
Statistical Arrays
Obs3 32
Obs4 31
Obs5 36
iris1 = set(iris1,'Units',{'mm'});
Curly brace indexing is used to access individual data elements. The following
are equivalent:
iris1{1,1}
ans =
35
iris1{'Obs1','SW'}
ans =
35
Combining Dataset Arrays. Combine two dataset arrays into a single
dataset array using square brackets:
SepalData = iris(:,{'SL','SW'});
PetalData = iris(:,{'PL','PW'});
newiris = [SepalData,PetalData];
size(newiris)
ans =
150 4
For horizontal concatenation, as in the preceding example, the number of
observations in the two dataset arrays must agree. Observations are matched
up by name (if given), regardless of their order in the two data sets.
The following concatenates variables within a dataset array and then deletes
the component variables:
newiris.SepalData = [newiris.SL,newiris.SW];
newiris.PetalData = [newiris.PL,newiris.PW];
newiris(:,{'SL','SW','PL','PW'}) = [];
size(newiris)
ans =
150 2
size(newiris.SepalData)
ans =
2-29
2 Organizing Data
150 2
newiris is now a 150-by-2 dataset array containing two 150-by-2 numerical
arrays as variables.
Vertical concatenation is also handled in a manner analogous to numerical
arrays:
newobs = dataset({[5.3 4.2; 5.0 4.1],'PetalData'},...
{[5.5 2; 4.8 2.1],'SepalData'});
newiris = [newiris;newobs];
size(newiris)
ans =
152 2
For vertical concatenation, as in the preceding example, the names of the
variables in the two dataset arrays must agree. Variables are matched up by
name, regardless of their order in the two data sets.
Expansion of variables is also accomplished using direct assignment to new
rows:
newiris(153,:) = dataset({[5.1 4.0],'PetalData'},...
{[5.1 4.2],'SepalData'});
A different type of concatenation is performed by join, which takes the data
in one dataset array and assigns it to the rows of another dataset array, based
on matching values in a common key variable. For example, the following
creates a dataset array with diploid chromosome counts for each species of iris:
snames = nominal({'setosa';'versicolor';'virginica'});
CC = dataset({snames,'species'},{[38;108;70],'cc'})
CC =
species cc
setosa 38
versicolor 108
virginica 70
This data is broadcast to the rows of iris using join:
iris2 = join(iris,CC);
2-30
Statistical Arrays
iris2([1 2 51 52 101 102],:)
ans =
species SL SW PL PW cc
Obs1 setosa 5.1 3.5 1.4 0.2 38
Obs2 setosa 4.9 3 1.4 0.2 38
Obs51 versicolor 7 3.2 4.7 1.4 108
Obs52 versicolor 6.4 3.2 4.5 1.5 108
Obs101 virginica 6.3 3.3 6 2.5 70
Obs102 virginica 5.8 2.7 5.1 1.9 70
Removing Observations from Dataset Arrays. Use one of the following
commands to remove observations or variables from a dataset (ds):
• Remove a variable by name:
ds.var = [];
• Remove the jth variable:
ds(:,j) = [];
• Remove the jth variable:
ds = ds(:,[1:(j-1) (j+1):end]);
• Remove the ith observation:
ds(i,:) = [];
• Remove the ith observation:
ds = ds([1:(i-1) (i+1):end],:);
• Remove the jth variable and ith observation:
ds = ds([1:(i-1) (i+1):end],[1:(j-1) (j+1):end]);
Computing with Dataset Arrays. Use summary to provide summary
statistics for the component variables of a dataset array:
summary(newiris)
Fisher's iris data (1936)
SepalData: [153x2 double]
2-31
2 Organizing Data
min 4.3000 2
1st Q 5.1000 2.8000
median 5.8000 3
3rd Q 6.4000 3.3250
max 7.9000 4.4000
PetalData: [153x2 double]
min 1 0.1000
1st Q 1.6000 0.3000
median 4.4000 1.3000
3rd Q 5.1000 1.8000
max 6.9000 4.2000
To apply other statistical functions, use dot indexing to access relevant
variables:
SepalMeans = mean(newiris.SepalData)
SepalMeans =
5.8294 3.0503
The same result is obtained with datasetfun, which applies functions to
dataset array variables:
means = datasetfun(@mean,newiris,'UniformOutput',false)
means =
[1x2 double] [1x2 double]
SepalMeans = means{1}
SepalMeans =
5.8294 3.0503
An alternative approach is to cast data in a dataset array as double and
apply statistical functions directly. Compare the following two methods
for computing the covariance of the length and width of the SepalData in
newiris:
covs = datasetfun(@cov,newiris,'UniformOutput',false)
covs =
[2x2 double] [2x2 double]
SepalCovs = covs{1}
SepalCovs =
0.6835 -0.0373
-0.0373 0.2054
2-32
Statistical Arrays
SepalCovs = cov(double(newiris(:,1)))
SepalCovs =
0.6835 -0.0373
-0.0373 0.2054
2-33
2 Organizing Data
Grouped Data
In this section...
“Grouping Variables” on page 2-34
“Functions for Grouped Data” on page 2-35
“Using Grouping Variables” on page 2-36
Grouping Variables
Grouping variables are utility variables used to indicate which elements
in a data set are to be considered together when computing statistics and
creating visualizations. They may be numeric vectors, string arrays, cell
arrays of strings, or categorical arrays. Logical vectors can be used to indicate
membership (or not) in a single group.
Grouping variables have the same length as the variables (columns) in a data
set. Observations (rows) i and j are considered to be in the same group if the
values of the corresponding grouping variable are identical at those indices.
Grouping variables with multiple columns are used to specify different groups
within multiple variables.
For example, the following loads the 150-by-4 numerical array meas and the
150-by-1 cell array of strings species into the workspace:
load fisheriris % Fisher's iris data (1936)
The data are 150 observations of four measured variables (by column number:
sepal length, sepal width, petal length, and petal width, respectively)
over three species of iris (setosa, versicolor, and virginica). To group the
observations by species, the following are all acceptable (and equivalent)
grouping variables:
group1 = species; % Cell array of strings
group2 = grp2idx(species); % Numeric vector
group3 = char(species); % Character array
group4 = nominal(species); % Categorical array
2-34
Grouped Data
These grouping variables can be supplied as input arguments to any of the
functions described in “Functions for Grouped Data” on page 2-35. Examples
are given in “Using Grouping Variables” on page 2-36.
Functions for Grouped Data
The following table lists Statistics Toolbox functions that accept a grouping
variable group as an input argument. The grouping variable may be in the
form of a vector, string array, cell array of strings, or categorical array, as
described in “Grouping Variables” on page 2-34.
For a full description of the syntax of any particular function, and examples
of its use, consult its reference page, linked from the table. “Using Grouping
Variables” on page 2-36 also includes examples.
Function Basic Syntax for Grouped Data
andrewsplot andrewsplot(X, ... ,'Group',group)
anova1 p = anova1(X,group)
anovan p = anovan(x,group)
aoctool aoctool(x,y,group)
boxplot boxplot(x,group)
classify class = classify(sample,training,group)
controlchart controlchart(x,group)
crosstab crosstab(group1,group2)
cvpartition c = cvpartition(group,'Kfold',k) or c =
cvpartition(group,'Holdout',p)
dummyvar D = dummyvar(group)
gagerr gagerr(x,group)
gplotmatrix gplotmatrix(x,y,group)
grp2idx [G,GN] = grp2idx(group)
grpstats means = grpstats(X,group)
gscatter gscatter(x,y,group)
interactionplot interactionplot(X,group)
2-35
2 Organizing Data
Function Basic Syntax for Grouped Data
kruskalwallis p = kruskalwallis(X,group)
maineffectsplot maineffectsplot(X,group)
manova1 d = manova1(X,group)
multivarichart multivarichart(x,group)
parallelcoords parallelcoords(X, ... ,'Group',group)
silhouette silhouette(X,group)
tabulate tabulate(group)
treefit T = treefit(X,y,'cost',S) or T =
treefit(X,y,'priorprob',S), where S.group
= group
vartestn vartestn(X,group)
Using Grouping Variables
This section provides an example demonstrating the use of grouping variables
and associated functions. Grouping variables are introduced in “Grouping
Variables” on page 2-34. A list of functions accepting grouping variables as
input arguments is given in “Functions for Grouped Data” on page 2-35.
Load the 150-by-4 numerical array meas and the 150-by-1 cell array of strings
species:
load fisheriris % Fisher's iris data (1936)
The data are 150 observations of four measured variables (by column number:
sepal length, sepal width, petal length, and petal width, respectively) over
three species of iris (setosa, versicolor, and virginica).
Create a categorical array (see “Categorical Arrays” on page 2-13) from
species to use as a grouping variable:
group = nominal(species);
2-36
Grouped Data
While species, as a cell array of strings, is itself a grouping variable, the
categorical array has the advantage that it can be easily manipulated with
methods of the categorical class.
Compute some basic statistics for the data (median and interquartile range),
by group, using the grpstats function:
[order,number,group_median,group_iqr] = ...
grpstats(meas,group,{'gname','numel',@median,@iqr})
order =
'setosa'
'versicolor'
'virginica'
number =
50 50 50 50
50 50 50 50
50 50 50 50
group_median =
5.0000 3.4000 1.5000 0.2000
5.9000 2.8000 4.3500 1.3000
6.5000 3.0000 5.5500 2.0000
group_iqr =
0.4000 0.5000 0.2000 0.1000
0.7000 0.5000 0.6000 0.3000
0.7000 0.4000 0.8000 0.5000
The statistics appear in 3-by-4 arrays, corresponding to the 3 groups
(categories) and 4 variables in the data. The order of the groups (not encoded
in the nominal array group) is indicated by the group names in order.
To improve the labeling of the data, create a dataset array (see “Dataset
Arrays” on page 2-23) from meas:
NumObs = size(meas,1);
NameObs = strcat({'Obs'},num2str((1:NumObs)','%-d'));
iris = dataset({group,'species'},...
{meas,'SL','SW','PL','PW'},...
'ObsNames',NameObs);
2-37
2 Organizing Data
When you call grpstats with a dataset array as an argument, you invoke the
grpstats method of the dataset class, rather than the grpstats function.
The method has a slightly different syntax than the function, but it returns
the same results, with better labeling:
stats = grpstats(iris,'species',{@median,@iqr})
stats =
species GroupCount
setosa setosa 50
versicolor versicolor 50
virginica virginica 50
median_SL iqr_SL
setosa 5 0.4
versicolor 5.9 0.7
virginica 6.5 0.7
median_SW iqr_SW
setosa 3.4 0.5
versicolor 2.8 0.5
virginica 3 0.4
median_PL iqr_PL
setosa 1.5 0.2
versicolor 4.35 0.6
virginica 5.55 0.8
median_PW iqr_PW
setosa 0.2 0.1
versicolor 1.3 0.3
virginica 2 0.5
Grouping variables are also used to create visualizations based on categories
of observations. The following scatter plot, created with the gscatter
function, shows the correlation between sepal length and sepal width in two
species of iris. Use ismember to subset the two species from group:
subset = ismember(group,{'setosa','versicolor'});
scattergroup = group(subset);
gscatter(iris.SL(subset),...
iris.SW(subset),...
2-38
Grouped Data
scattergroup)
xlabel('Sepal Length')
ylabel('Sepal Width')
2-39
2 Organizing Data
2-40
3
Descriptive Statistics
• “Introduction” on page 3-2
• “Measures of Central Tendency” on page 3-3
• “Measures of Dispersion” on page 3-5
• “Measures of Shape” on page 3-7
• “Resampling Statistics” on page 3-9
• “Data with Missing Values” on page 3-16
3 Descriptive Statistics
Introduction
You may need to summarize large, complex data sets—both numerically
and visually—to convey their essence to the data analyst and to allow for
further processing. This chapter focuses on numerical summaries; Chapter 4,
“Statistical Visualization” focuses on visual summaries.
3-2
Measures of Central Tendency
Measures of Central Tendency
Measures of central tendency locate a distribution of data along an
appropriate scale.
The following table lists the functions that calculate the measures of central
tendency.
Function Name Description
geomean Geometric mean
harmmean Harmonic mean
mean Arithmetic average
median 50th percentile
mode Most frequent value
trimmean Trimmed mean
The average is a simple and popular estimate of location. If the data sample
comes from a normal distribution, then the sample mean is also optimal
(MVUE of µ).
Unfortunately, outliers, data entry errors, or glitches exist in almost all
real data. The sample mean is sensitive to these problems. One bad data
value can move the average away from the center of the rest of the data by
an arbitrarily large distance.
The median and trimmed mean are two measures that are resistant (robust)
to outliers. The median is the 50th percentile of the sample, which will only
change slightly if you add a large perturbation to any value. The idea behind
the trimmed mean is to ignore a small percentage of the highest and lowest
values of a sample when determining the center of the sample.
The geometric mean and harmonic mean, like the average, are not robust
to outliers. They are useful when the sample is distributed lognormal or
heavily skewed.
3-3
3 Descriptive Statistics
The following example shows the behavior of the measures of location for a
sample with one outlier.
x = [ones(1,6) 100]
x =
1 1 1 1 1 1 100
locate = [geomean(x) harmmean(x) mean(x) median(x)...
trimmean(x,25)]
locate =
1.9307 1.1647 15.1429 1.0000 1.0000
You can see that the mean is far from any data value because of the influence
of the outlier. The median and trimmed mean ignore the outlying value and
describe the location of the rest of the data values.
3-4
Measures of Dispersion
Measures of Dispersion
The purpose of measures of dispersion is to find out how spread out the data
values are on the number line. Another term for these statistics is measures
of spread.
The table gives the function names and descriptions.
Function
Name Description
iqr Interquartile range
mad Mean absolute deviation
moment Central moment of all orders
range Range
std Standard deviation
var Variance
The range (the difference between the maximum and minimum values) is the
simplest measure of spread. But if there is an outlier in the data, it will be the
minimum or maximum value. Thus, the range is not robust to outliers.
The standard deviation and the variance are popular measures of spread that
are optimal for normally distributed samples. The sample variance is the
MVUE of the normal parameter σ2. The standard deviation is the square root
of the variance and has the desirable property of being in the same units as
the data. That is, if the data is in meters, the standard deviation is in meters
as well. The variance is in meters2, which is more difficult to interpret.
Neither the standard deviation nor the variance is robust to outliers. A data
value that is separate from the body of the data can increase the value of the
statistics by an arbitrarily large amount.
The mean absolute deviation (MAD) is also sensitive to outliers. But the
MAD does not move quite as much as the standard deviation or variance in
response to bad data.
3-5
3 Descriptive Statistics
The interquartile range (IQR) is the difference between the 75th and 25th
percentile of the data. Since only the middle 50% of the data affects this
measure, it is robust to outliers.
The following example shows the behavior of the measures of dispersion for a
sample with one outlier.
x = [ones(1,6) 100]
x =
1 1 1 1 1 1 100
stats = [iqr(x) mad(x) range(x) std(x)]
stats =
0 24.2449 99.0000 37.4185
3-6
Measures of Shape
Measures of Shape
Quantiles and percentiles provide information about the shape of data as
well as its location and spread.
The quantile of order p (0 ≤ p ≤ 1) is the smallest x value where the cumulative
distribution function equals or exceeds p. The function quantile computes
quantiles as follows:
1 n sorted data points are the 0.5/n, 1.5/n, ..., (n–0.5)/n quantiles.
2 Linear interpolation is used to compute intermediate quantiles.
3 The data min or max are assigned to quantiles outside the range.
4 Missing values are treated as NaN, and removed from the data.
Percentiles, computed by the prctile function, are quantiles for a certain
percentage of the data, specified for 0 ≤ p ≤ 100.
The following example shows the result of looking at every quartile (quantiles
with orders that are multiples of 0.25) of a sample containing a mixture of
two distributions.
x = [normrnd(4,1,1,100) normrnd(6,0.5,1,200)];
p = 100*(0:0.25:1);
y = prctile(x,p);
z = [p;y]
z =
0 25.0000 50.0000 75.0000 100.0000
1.8293 4.6728 5.6459 6.0766 7.1546
A box plot helps to visualize the statistics:
boxplot(x)
3-7
3 Descriptive Statistics
The long lower tail and plus signs show the lack of symmetry in the sample
values. For more information on box plots, see “Box Plots” on page 4-6.
The shape of a data distribution is also measured by the Statistics Toolbox
functions skewness, kurtosis, and, more generally, moment.
3-8
Resampling Statistics
Resampling Statistics
In this section...
“The Bootstrap” on page 3-9
“The Jackknife” on page 3-12
“Parallel Computing Support for Resampling Methods” on page 3-13
The Bootstrap
The bootstrap procedure involves choosing random samples with replacement
from a data set and analyzing each sample the same way. Sampling with
replacement means that each observation is selected separately at random
from the original dataset. So a particular data point from the original data
set could appear multiple times in a given bootstrap sample. The number of
elements in each bootstrap sample equals the number of elements in the
original data set. The range of sample estimates you obtain enables you to
establish the uncertainty of the quantity you are estimating.
This example from Efron and Tibshirani [33] compares Law School Admission
Test (LSAT) scores and subsequent law school grade point average (GPA) for
a sample of 15 law schools.
load lawdata
plot(lsat,gpa,'+')
lsline
3-9
3 Descriptive Statistics
The least-squares fit line indicates that higher LSAT scores go with higher
law school GPAs. But how certain is this conclusion? The plot provides some
intuition, but nothing quantitative.
You can calculate the correlation coefficient of the variables using the corr
function.
rhohat = corr(lsat,gpa)
rhohat =
0.7764
Now you have a number describing the positive connection between LSAT
and GPA; though it may seem large, you still do not know if it is statistically
significant.
3-10
Resampling Statistics
Using the bootstrp function you can resample the lsat and gpa vectors as
many times as you like and consider the variation in the resulting correlation
coefficients.
Here is an example.
rhos1000 = bootstrp(1000,'corr',lsat,gpa);
This command resamples the lsat and gpa vectors 1000 times and computes
the corr function on each sample. Here is a histogram of the result.
hist(rhos1000,30)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
Nearly all the estimates lie on the interval [0.4 1.0].
3-11
3 Descriptive Statistics
It is often desirable to construct a confidence interval for a parameter
estimate in statistical inferences. Using the bootci function, you can use
bootstrapping to obtain a confidence interval. The confidence interval for the
lsat and gpa data is computed as:
ci = bootci(5000,@corr,lsat,gpa)
ci =
0.3313
0.9427
Therefore, a 95% confidence interval for the correlation coefficient between
LSAT and GPA is [0.33 0.94]. This is strong quantitative evidence that LSAT
and subsequent GPA are positively correlated. Moreover, this evidence does
not require any strong assumptions about the probability distribution of the
correlation coefficient.
Although the bootci function computes the Bias Corrected and accelerated
(BCa) interval as the default type, it is also able to compute various other
types of bootstrap confidence intervals, such as the studentized bootstrap
confidence interval.
The Jackknife
Similar to the bootstrap is the jackknife, which uses resampling to estimate
the bias of a sample statistic. Sometimes it is also used to estimate standard
error of the sample statistic. The jackknife is implemented by the Statistics
Toolbox function jackknife.
The jackknife resamples systematically, rather than at random as the
bootstrap does. For a sample with n points, the jackknife computes sample
statistics on n separate samples of size n-1. Each sample is the original data
with a single observation omitted.
In the previous bootstrap example you measured the uncertainty in
estimating the correlation coefficient. You can use the jackknife to estimate
the bias, which is the tendency of the sample correlation to over-estimate or
under-estimate the true, unknown correlation. First compute the sample
correlation on the data:
3-12
Resampling Statistics
load lawdata
rhohat = corr(lsat,gpa)
rhohat =
0.7764
Next compute the correlations for jackknife samples, and compute their mean:
jackrho = jackknife(@corr,lsat,gpa);
meanrho = mean(jackrho)
meanrho =
0.7759
Now compute an estimate of the bias:
n = length(lsat);
biasrho = (n-1) * (meanrho-rhohat)
biasrho =
-0.0065
The sample correlation probably underestimates the true correlation by about
this amount.
Parallel Computing Support for Resampling Methods
Parallel computing is the technique of using multiple processors on a single
problem. The primary reason to use parallel computing is to shorten the
computation time.
Resampling methods all take as input a statistical function and a set of
supplied data, and evaluate the statistical function repeatedly, on multiple
samples drawn from the supplied data. Resampling methods are statistically
informative but they can be very time-consuming. But because the repeat
evaluations are independent of one another, you can reduce computation time
3-13
3 Descriptive Statistics
by performing those repeat evaluations in parallel. The following functions
support parallel computing:
• bootci
• bootstrp
• crossval
• jackknife
• TreeBagger
• TreeBagger.growTrees
These functions use parallel resampling under the following conditions:
• You have a license for Parallel Computing Toolbox™ software and the
software is installed.
• A group of processors has been prepared for parallel computation using the
matlabpool command of the Parallel Computing Toolbox.
• The option UseParallel is set to 'always'. The default value of this option
is 'never'. You specify this option using the 'Options' argument that all
of these resampling functions accept.
When these conditions hold, the functions resample in parallel. For more
information on the Parallel Computing Toolbox, see its User Guide..
Nested Parallel Functions
Resampling methods employ the Parallel Computing Toolbox function parfor
to perform parallel evaluations. parfor does not work in parallel when called
from within another parfor loop. Parallelization occurs only at the outermost
level if you combine parallel resampling methods with parallel functionality
in your statistical function or in the code that calls the resampling methods.
Suppose, for example, you want to apply the jackknife to your function
userfcn, which calls parfor, and you wish to call jackknife in a loop.
Suppose also that the conditions for parallel resampling of bootstrp, as given
in the section above, are satisfied. The following figure shows three cases:
1 The outermost loop is parfor. Only that loop runs in parallel.
3-14
Resampling Statistics
2 The outermost parfor loop is in jackknife. Only jackknife runs in
parallel.
3 The outermost parfor loop is in userfcn. userfcn can use parfor in
parallel.
When parfor Runs In Parallel
3-15
3 Descriptive Statistics
Data with Missing Values
Many data sets have one or more missing values. It is convenient to code
missing values as NaN (Not a Number) to preserve the structure of data sets
across multiple variables and observations.
For example:
X = magic(3);
X([1 5]) = [NaN NaN]
X =
NaN 1 6
3 NaN 7
4 9 2
Normal MATLAB arithmetic operations yield NaN values when operands
are NaN:
s1 = sum(X)
s1 =
NaN NaN 15
Removing the NaN values would destroy the matrix structure. Removing
the rows containing the NaN values would discard data. Statistics Toolbox
functions in the following table remove NaN values only for the purposes of
computation.
Function Description
nancov Covariance matrix, ignoring NaN values
nanmax Maximum, ignoring NaN values
nanmean Mean, ignoring NaN values
nanmedian Median, ignoring NaN values
nanmin Minimum, ignoring NaN values
nanstd Standard deviation, ignoring NaN values
nansum Sum, ignoring NaN values
nanvar Variance, ignoring NaN values
3-16
Data with Missing Values
For example:
s2 = nansum(X)
s2 =
7 10 15
Other Statistics Toolbox functions also ignore NaN values. These include iqr,
kurtosis, mad, prctile, range, skewness, and trimmean.
3-17
3 Descriptive Statistics
3-18
4
Statistical Visualization
• “Introduction” on page 4-2
• “Scatter Plots” on page 4-3
• “Box Plots” on page 4-6
• “Distribution Plots” on page 4-8
4 Statistical Visualization
Introduction
形象化 大量的
Statistics Toolbox data visualization functions add to the extensive graphics
能力
capabilities already in MATLAB.
分散 多變量
形象化
• Scatter plots are a basic visualization tool for multivariate data. They
辨識 relationships among variables.
are used to identify 變因 版本 of
Grouped versions
指出 group membership.
these plots use different plotting symbols to indicate
The gname function is used to label points on these plots with a text label
or an observation number.
• Box plots display a five-number summary of a set of data: the median,
the two ends of the interquartile range (the box), and two extreme values
(the whiskers) above and below the box. Because they show less detail
than histograms, box plots are most useful for side-by-side comparisons
of two distributions.
• Distribution plots help you identify an appropriate distribution family
for your data. They include normal and Weibull probability plots,
quantile-quantile plots, and empirical cumulative distribution plots.
Advanced Statistics Toolbox visualization functions are available for
specialized statistical analyses.
Note For information on creating visualizations of data by group, see
“Grouped Data” on page 2-34.
4-2
Scatter Plots
Scatter Plots 變數對的關係
A scatter plot is a simple plot of one variable against another. The MATLAB
functions plot and scatter produce scatter plots. The MATLAB function
plotmatrix can produce a matrix of such plots showing the relationship
between several pairs of variables.
Statistics Toolbox functions gscatter and gplotmatrix produce grouped
versions of these plots. These are useful for determining whether the values
of two variables or the relationship between those variables is the same in
each group.
Suppose you want to examine the weight and mileage of cars from three
different model years.
load carsmall
gscatter(Weight,MPG,Model_Year,'','xos')
4-3
4 Statistical Visualization
This shows that not only is there a strong relationship between the weight of
a car and its mileage, but also that newer cars tend to be lighter and have
better gas mileage than older cars.
參數
The default arguments for gscatter produce a scatter plot with the different
groups shown with the same symbol but different colors. The last two
arguments above request that all groups be shown in default colors and with
different symbols.
The carsmall data set contains other variables that describe different aspects
of cars. You can examine several of them in a single display by creating a
grouped plot matrix.
xvars = [Weight Displacement Horsepower];
yvars = [MPG Acceleration];
4-4
Scatter Plots
gplotmatrix(xvars,yvars,Model_Year,'','xos')
The upper right subplot displays MPG against Horsepower, and shows that
over the years the horsepower of the cars has decreased but the gas mileage
has improved.
The gplotmatrix function can also graph all pairs from a single list of
variables, along with histograms for each variable. See “MANOVA” on page
8-39.
4-5
4 Statistical Visualization
Box Plots
The graph below, created with the boxplot command, compares petal lengths
in samples from two species of iris.
load fisheriris
s1 = meas(51:100,3);
s2 = meas(101:150,3);
boxplot([s1 s2],'notch','on',...
'labels',{'versicolor','virginica'})
This plot has the following features:
• The tops and bottoms of each “box” are the 25th and 75th percentiles of the
samples, respectively. The distances between the tops and bottoms are the
interquartile ranges.
4-6
Box Plots
• The line in the middle of each box is the sample median. If the median is
not centered in the box, it shows sample skewness.
• The whiskers are lines extending above and below each box. Whiskers are
drawn from the ends of the interquartile ranges to the furthest observations
within the whisker length (the adjacent values).
• Observations beyond the whisker length are marked as outliers. By
default, an outlier is a value that is more than 1.5 times the interquartile
range away from the top or bottom of the box, but this value can be adjusted
with additional input arguments. Outliers are displayed with a red + sign.
• Notches display the variability of the median between samples. The width
of a notch is computed so that box plots whose notches do not overlap (as
above) have different medians at the 5% significance level. The significance
level is based on a normal distribution assumption, but comparisons of
medians are reasonably robust for other distributions. Comparing box-plot
medians is like a visual hypothesis test, analogous to the t test used for
means.
4-7
4 Statistical Visualization
Distribution Plots
In this section...
“Normal Probability Plots” on page 4-8
“Quantile-Quantile Plots” on page 4-10
“Cumulative Distribution Plots” on page 4-12
“Other Probability Plots” on page 4-14
Normal Probability Plots
Normal probability plots are used to assess whether data comes from a
normal distribution. Many statistical procedures make the assumption that
an underlying distribution is normal, so normal probability plots can provide
some assurance that the assumption is justified, or else provide a warning of
problems with the assumption. An analysis of normality typically combines
normal probability plots with hypothesis tests for normality, as described in
Chapter 7, “Hypothesis Tests”.
The following example shows a normal probability plot created with the
normplot function.
x = normrnd(10,1,25,1);
normplot(x)
4-8
Distribution Plots
The plus signs plot the empirical probability versus the data value for each
point in the data. A solid line connects the 25th and 75th percentiles in the
data, and a dashed line extends it to the ends of the data. The y-axis values
are probabilities from zero to one, but the scale is not linear. The distance
between tick marks on the y-axis matches the distance between the quantiles
of a normal distribution. The quantiles are close together near the median
(probability = 0.5) and stretch out symmetrically as you move away from
the median.
In a normal probability plot, if all the data points fall near the line, an
assumption of normality is reasonable. Otherwise, the points will curve away
from the line, and an assumption of normality is not justified.
For example:
x = exprnd(10,100,1);
4-9
4 Statistical Visualization
normplot(x)
The plot is strong evidence that the underlying distribution is not normal.
Quantile-Quantile Plots
Quantile-quantile plots are used to determine whether two samples come from
the same distribution family. They are scatter plots of quantiles computed
from each sample, with a line drawn between the first and third quartiles. If
the data falls near the line, it is reasonable to assume that the two samples
come from the same distribution. The method is robust with respect to
changes in the location and scale of either distribution.
To create a quantile-quantile plot, use the qqplot function.
4-10
Distribution Plots
The following example shows a quantile-quantile plot of two samples from
Poisson distributions.
x = poissrnd(10,50,1);
y = poissrnd(5,100,1);
qqplot(x,y);
Even though the parameters and sample sizes are different, the approximate
linear relationship suggests that the two samples may come from the same
distribution family. As with normal probability plots, hypothesis tests,
as described in Chapter 7, “Hypothesis Tests”, can provide additional
justification for such an assumption. For statistical procedures that depend
on the two samples coming from the same distribution, however, a linear
quantile-quantile plot is often sufficient.
4-11
4 Statistical Visualization
The following example shows what happens when the underlying distributions
are not the same.
x = normrnd(5,1,100,1);
y = wblrnd(2,0.5,100,1);
qqplot(x,y);
These samples clearly are not from the same distribution family.
Cumulative Distribution Plots
An empirical cumulative distribution function (cdf) plot shows the proportion
of data less than each x value, as a function of x. The scale on the y-axis is
linear; in particular, it is not scaled to any particular distribution. Empirical
cdf plots are used to compare data cdfs to cdfs for particular distributions.
4-12
Distribution Plots
To create an empirical cdf plot, use the cdfplot function (or ecdf and stairs).
The following example compares the empirical cdf for a sample from an
extreme value distribution with a plot of the cdf for the sampling distribution.
In practice, the sampling distribution would be unknown, and would be
chosen to match the empirical cdf.
y = evrnd(0,3,100,1);
cdfplot(y)
hold on
x = -20:0.1:10;
f = evcdf(x,0,3);
plot(x,f,'m')
legend('Empirical','Theoretical','Location','NW')
4-13
4 Statistical Visualization
Other Probability Plots
A probability plot, like the normal probability plot, is just an empirical cdf plot
scaled to a particular distribution. The y-axis values are probabilities from
zero to one, but the scale is not linear. The distance between tick marks is the
distance between quantiles of the distribution. In the plot, a line is drawn
between the first and third quartiles in the data. If the data falls near the
line, it is reasonable to choose the distribution as a model for the data.
To create probability plots for different distributions, use the probplot
function.
For example, the following plot assesses two samples, one from a Weibull
distribution and one from a Rayleigh distribution, to see if they may have
come from a Weibull population.
x1 = wblrnd(3,3,100,1);
x2 = raylrnd(3,100,1);
probplot('weibull',[x1 x2])
legend('Weibull Sample','Rayleigh Sample','Location','NW')
4-14
Distribution Plots
The plot gives justification for modeling the first sample with a Weibull
distribution; much less so for the second sample.
A distribution analysis typically combines probability plots with hypothesis
tests for a particular distribution, as described in Chapter 7, “Hypothesis
Tests”.
4-15
4 Statistical Visualization
4-16
5
Probability Distributions
• “Using Probability Distributions” on page 5-2
• “Supported Distributions” on page 5-3
• “Working with Distributions Through GUIs” on page 5-9
• “Statistics Toolbox Distribution Functions” on page 5-52
• “Using Probability Distribution Objects” on page 5-84
• “Probability Distributions Used for Multivariate Modeling” on page 5-99
5 Probability Distributions
Using Probability Distributions
Probability distributions are theoretical distributions based on
assumptions about a source population. They assign probability to the event
that a random variable takes on a specific, discrete value, or falls within a
specified range of continuous values. There are two main types of models:
• Parametric Models—Choose a model based on a parametric family of
probability distributions and then adjust the parameters to fit the data.
For information on supported parametric distributions, see “Parametric
Distributions” on page 5-4.
• Nonparametric Models—When data or statistics do not follow
any standard probability distribution, nonparametric models may be
appropriate. For information on supported nonparametric distributions,
see “Nonparametric Distributions” on page 5-8.
The Statistics Toolbox provides several ways of working with both parametric
and nonparametric probability distributions:
• Graphic User Interfaces (GUIs)—Interact with the distributions to
visualize distributions, fit a distribution to your data, or generate random
data using a specific distribution. For more information, see “Working with
Distributions Through GUIs” on page 5-9.
• Command Line Functions—Use command-line functions to further
explore the distributions, fit relevant models to your data, or generate
random data. For more information on using functions, see “Statistics
Toolbox Distribution Functions” on page 5-52.
• Distribution Objects—Use objects to explore and fit your data to a
distribution, save the results to a single entity, and generate random
data from the resulting parameters. For more information, see “Using
Probability Distribution Objects” on page 5-84.
5-2
Supported Distributions
Supported Distributions
In this section...
“Parametric Distributions” on page 5-4
“Nonparametric Distributions” on page 5-8
Probability distributions supported by the Statistics Toolbox are
cross-referenced with their supporting functions and GUIs in the following
tables. The tables use the following abbreviations for distribution functions:
• pdf — Probability density functions
• cdf — Cumulative distribution functions
• inv — Inverse cumulative distribution functions
• stat — Distribution statistics functions
• fit — Distribution fitting functions
• like — Negative log-likelihood functions
• rnd — Random number generators
For more detailed explanations of each supported distribution, see Appendix
B, “Distribution Reference”.
5-3
5 Probability Distributions
Parametric Distributions
Continuous Distributions (Data)
Name pdf cdf inv stat fit like rnd
Beta betapdf, betacdf, betainv, betastat betafit, betalike betarnd,
pdf cdf icdf mle random,
randtool
Birnbaum- dfittool
Saunders
Exponential exppdf, expcdf, expinv, expstat expfit, explike exprnd,
pdf cdf icdf mle, random,
dfittool randtool
Extreme evpdf, evcdf, evinv, evstat evfit, mle, evlike evrnd,
value pdf cdf icdf dfittool random,
randtool
Gamma gampdf, gamcdf, gaminv, gamstat gamfit, gamlike gamrnd,
pdf cdf icdf mle, randg,
dfittool random,
randtool
Generalized gevpdf, gevcdf, gevinv, gevstat gevfit, gevlike gevrnd,
extreme pdf cdf icdf mle, random,
value dfittool randtool
Generalized gppdf, gpcdf, gpinv, gpstat gpfit, mle, gplike gprnd,
Pareto pdf cdf icdf dfittool random,
randtool
Inverse dfittool
Gaussian
Johnson johnsrnd johnsrnd
system
Logistic dfittool
Loglogistic dfittool
5-4
Supported Distributions
Name pdf cdf inv stat fit like rnd
Lognormal lognpdf, logncdf, logninv, lognstat lognfit, lognlike lognrnd,
pdf cdf icdf mle, random,
dfittool randtool
Nakagami dfittool
Normal normpdf, normcdf, norminv, normstat normfit, normlike normrnd,
(Gaussian) pdf cdf icdf mle, randn,
dfittool random,
randtool
Pearson pearsrnd pearsrnd
system
Piecewise pdf cdf icdf paretotails random
Rayleigh raylpdf, raylcdf, raylinv, raylstat raylfit, raylrnd,
pdf cdf icdf mle, random,
dfittool randtool
Rician dfittool
Uniform unifpdf, unifcdf, unifinv, unifstat unifit, mle unifrnd,
(continuous) pdf cdf icdf rand,
random
Weibull wblpdf, wblcdf, wblinv, wblstat wblfit, wbllike wblrnd,
pdf cdf icdf mle, random
dfittool
5-5
5 Probability Distributions
Continuous Distributions (Statistics)
Name pdf cdf inv stat fit like rnd
Chi-square chi2pdf, chi2cdf, chi2inv, chi2stat chi2rnd,
pdf cdf icdf random,
randtool
F fpdf, pdf fcdf, cdf finv, fstat frnd,
icdf random,
randtool
Noncentral ncx2pdf, ncx2cdf, ncx2inv, ncx2stat ncx2rnd,
chi-square pdf cdf icdf random,
randtool
Noncentral ncfpdf, ncfcdf, ncfinv, ncfstat ncfrnd,
F pdf cdf icdf random,
randtool
Noncentral nctpdf, nctcdf, nctinv, nctstat nctrnd,
t pdf cdf icdf random,
randtool
Student’s t tpdf, pdf tcdf, cdf tinv, tstat trnd,
icdf random,
randtool
t location- dfittool
scale
5-6
Supported Distributions
Discrete Distributions
Name pdf cdf inv stat fit like rnd
Binomial binopdf, binocdf, binoinv, binostat binofit, binornd,
pdf cdf icdf mle, random,
dfittool randtool
Bernoulli mle
Geometric geopdf, geocdf, geoinv, geostat mle geornd,
pdf cdf icdf random,
randtool
hygepdf,
Hypergeometric hygecdf, hygeinv, hygestat hygernd,
pdf cdf icdf random
Multinomial mnpdf mnrnd
Negative nbinpdf, nbincdf, nbininv, nbinstat nbinfit, nbinrnd,
binomial pdf cdf icdf mle, random,
dfittool randtool
Poisson poisspdf, poisscdf, poissinv, poisstat poissfit, poissrnd,
pdf cdf icdf mle, random,
dfittool randtool
Uniform unidpdf, unidcdf, unidinv, unidstat mle unidrnd,
(discrete) pdf cdf icdf random,
randtool
5-7
5 Probability Distributions
Multivariate Distributions
Name pdf cdf inv stat fit like rnd
Gaussian copulapdf copulacdf copulastat copulafit copularnd
copula
Gaussian pdf cdf fit random
mixture
t copula copulapdf copulacdf copulastat copulafit copularnd
Clayton copulapdf copulacdf copulastat copulafit copularnd
copula
Frank copulapdf copulacdf copulastat copulafit copularnd
copula
Gumbel copulapdf copulacdf copulastat copulafit copularnd
copula
Inverse iwishrnd
Wishart
Multivariate mvnpdf mvncdf mvnrnd
normal
Multivariate mvtpdf mvtcdf mvtrnd
t
Wishart wishrnd
Nonparametric Distributions
Name pdf cdf inv stat fit like rnd
ksdensity,
Nonparametric ksdensity ksdensity ksdensity
dfittool
5-8
Working with Distributions Through GUIs
Working with Distributions Through GUIs
In this section...
“Exploring Distributions” on page 5-9
“Modeling Your Data Using the Distribution Fitting GUI” on page 5-11
“Visually Exploring Random Number Generation” on page 5-49
This section describes Statistics Toolbox GUIs that provide convenient,
interactive access to the distribution functions described in “Statistics Toolbox
Distribution Functions” on page 5-52.
Exploring Distributions
To interactively see the influence of parameter changes on the shapes of the
pdfs and cdfs of supported Statistics Toolbox distributions, use the Probability
Distribution Function Tool.
Run the tool by typing disttool at the command line.
5-9
5 Probability Distributions
Choose Function type
distribution (cdf or pdf)
Function
plot
Function
value Draggable
reference
lines
Parameter Parameter Parameter Additional
bounds value control parameters
Start by selecting a distribution. Then choose the function type: probability
density function (pdf) or cumulative distribution function (cdf).
5-10
Working with Distributions Through GUIs
Once the plot displays, you can
• Calculate a new function value by
- Typing a new x value in the text box on the x-axis
- Dragging the vertical reference line.
- Clicking in the figure where you want the line to be.
The new function value displays in the text box to the left of the plot.
• For cdf plots, find critical values corresponding to a specific probability by
typing the desired probability in the text box on the y-axis or by dragging
the horizontal reference line.
• Use the controls at the bottom of the window to set parameter values for
the distribution and to change their upper and lower bounds.
Modeling Your Data Using the Distribution Fitting GUI
The Distribution Fitting Tool is a GUI for fitting univariate distributions to
data. This section describes how to use the Distribution Fitting Tool and
covers the following topics:
• “Starting the Distribution Fitting Tool” on page 5-12
• “Creating and Managing Data Sets” on page 5-14
• “Creating a New Fit” on page 5-19
• “Displaying Results” on page 5-25
• “Managing Fits” on page 5-26
• “Evaluating Fits” on page 5-28
• “Excluding Data” on page 5-32
• “Saving and Loading Sessions” on page 5-38
• “Example: Fitting a Distribution” on page 5-39
• “Generating a File to Fit and Plot Distributions” on page 5-46
• “Using Custom Distributions” on page 5-48
• “Additional Distributions Available in the Distribution Fitting Tool” on
page 5-49
5-11
5 Probability Distributions
Starting the Distribution Fitting Tool
To open the Distribution Fitting Tool, enter the command
dfittool
The following figure shows the main window of the Distribution Fitting Tool.
Select display Select distribution (probability plot only)
Task buttons
Import data
from workspace
Create a new fit
Manage multiple fits Evaluate distribution Exclude data
at selected points from fit
Adjusting the Plot. Buttons at the top of the tool allow you to adjust the
plot displayed in the main window:
5-12
Working with Distributions Through GUIs
• — Toggle the legend on (default) or off.
• — Toggle grid lines on or off (default).
• — Restore default axes limits.
Displaying the Data. The Display Type field specifies the type of plot
displayed in the main window. Each type corresponds to a probability
function, for example, a probability density function. The following display
types are available:
• Density (PDF) — Displays a probability density function (PDF) plot for
the fitted distribution.
• Cumulative probability (CDF) — Displays a cumulative probability
plot of the data.
• Quantile (inverse CDF) — Displays a quantile (inverse CDF) plot.
• Probability plot — Displays a probability plot.
• Survivor function — Displays a survivor function plot of the data.
• Cumulative hazard — Displays a cumulative hazard plot of the data.
Inputting and Fitting Data. The task buttons enable you to perform the
tasks necessary to fit distributions to data. Each button opens a new window
in which you perform the task. The buttons include
• Data — Import and manage data sets. See “Creating and Managing Data
Sets” on page 5-14.
• New Fit — Create new fits. See “Creating a New Fit” on page 5-19.
• Manage Fits — Manage existing fits. See “Managing Fits” on page 5-26.
• Evaluate — Evaluate fits at any points you choose. See “Evaluating Fits”
on page 5-28.
• Exclude — Create rules specifying which values to exclude when fitting a
distribution. See “Excluding Data” on page 5-32.
The display pane displays plots of the data sets and fits you create. Whenever
you make changes in one of the task windows, the results in the display pane
update.
5-13
5 Probability Distributions
Saving and Customizing Distributions. The Distribution Fitting Tool
menus contain items that enable you to do the following:
• Save and load sessions. See “Saving and Loading Sessions” on page 5-38.
• Generate a file with which you can fit distributions to data and plot the
results independently of the Distribution Fitting Tool. See “Generating a
File to Fit and Plot Distributions” on page 5-46.
• Define and import custom distributions. See “Using Custom Distributions”
on page 5-48.
Creating and Managing Data Sets
This section describes how create and manage data sets.
To begin, click the Data button in the main window of the Distribution Fitting
Tool to open the Data window shown in the following figure.
5-14
Working with Distributions Through GUIs
Importing Data. The Import workspace vectors pane enables you to
create a data set by importing a vector from the MATLAB workspace. The
following sections describe the fields of the Import workspace vectors
pane and give appropriate values for vectors imported from the MATLAB
workspace:
• Data—The drop-down list in the Data field contains the names of all
matrices and vectors, other than 1-by-1 matrices (scalars) in the MATLAB
workspace. Select the array containing the data you want to fit. The actual
data you import must be a vector. If you select a matrix in the Data field,
the first column of the matrix is imported by default. To select a different
column or row of the matrix, click Select Column or Row. This displays
5-15
5 Probability Distributions
the matrix in the Variable Editor, where you can select a row or column
by highlighting it with the mouse.
Alternatively, you can enter any valid MATLAB expression in the Data
field.
When you select a vector in the Data field, a histogram of the data is
displayed in the Data preview pane.
• Censoring—If some of the points in the data set are censored, enter
a Boolean vector, of the same size as the data vector, specifying the
censored entries of the data. A 1 in the censoring vector specifies that the
corresponding entry of the data vector is censored, while a 0 specifies that
the entry is not censored. If you enter a matrix, you can select a column or
row by clicking Select Column or Row. If you do not want to censor any
data, leave the Censoring field blank.
• Frequency—Enter a vector of positive integers of the same size as the
data vector to specify the frequency of the corresponding entries of the data
vector. For example, a value of 7 in the 15th entry of frequency vector
specifies that there are 7 data points corresponding to the value in the 15th
entry of the data vector. If all entries of the data vector have frequency 1,
leave the Frequency field blank.
• Data set name—Enter a name for the data set you import from the
workspace, such as My data.
After you have entered the information in the preceding fields, click Create
Data Set to create the data set My data.
Managing Data Sets. The Manage data sets pane enables you to view
and manage the data sets you create. When you create a data set, its name
appears in the Data sets list. The following figure shows the Manage data
sets pane after creating the data set My data.
5-16
Working with Distributions Through GUIs
For each data set in the Data sets list, you can
• Select the Plot check box to display a plot of the data in the main
Distribution Fitting Tool window. When you create a new data set, Plot is
selected by default. Clearing the Plot check box removes the data from the
plot in the main window. You can specify the type of plot displayed in the
Display Type field in the main window.
• If Plot is selected, you can also select Bounds to display confidence
interval bounds for the plot in the main window. These bounds are
pointwise confidence bounds around the empirical estimates of these
functions. The bounds are only displayed when you set Display Type in
the main window to one of the following:
- Cumulative probability (CDF)
- Survivor function
- Cumulative hazard
The Distribution Fitting Tool cannot display confidence bounds on density
(PDF), quantile (inverse CDF), or probability plots. Clearing the Bounds
check box removes the confidence bounds from the plot in the main window.
When you select a data set from the list, the following buttons are enabled:
• View — Displays the data in a table in a new window.
• Set Bin Rules — Defines the histogram bins used in a density (PDF) plot.
• Rename — Renames the data set.
5-17
5 Probability Distributions
• Delete — Deletes the data set.
Setting Bin Rules. To set bin rules for the histogram of a data set, click Set
Bin Rules. This opens the dialog box shown in the following figure.
You can select from the following rules:
• Freedman-Diaconis rule — Algorithm that chooses bin widths and
locations automatically, based on the sample size and the spread of the
data. This rule, which is the default, is suitable for many kinds of data.
• Scott rule — Algorithm intended for data that are approximately normal.
The algorithm chooses bin widths and locations automatically.
• Number of bins — Enter the number of bins. All bins have equal widths.
• Bins centered on integers — Specifies bins centered on integers.
5-18
Working with Distributions Through GUIs
• Bin width — Enter the width of each bin. If you select this option, you can
make the following choices:
- Automatic bin placement — Places the edges of the bins at integer
multiples of the Bin width.
- Bin boundary at — Enter a scalar to specify the boundaries of the
bins. The boundary of each bin is equal to this scalar plus an integer
multiple of the Bin width.
The Set Bin Width Rules dialog box also provides the following options:
• Apply to all existing data sets — When selected, the rule is applied to
all data sets. Otherwise, the rule is only applied to the data set currently
selected in the Data window.
• Save as default — When selected, the current rule is applied to any
new data sets that you create. You can also set default bin width rules
by selecting Set Default Bin Rules from the Tools menu in the main
window.
Creating a New Fit
This section describes how to create a new fit. To begin, click the New Fit
button at the top of the main window to open a New Fit window. If you
created the data set My data, it appears in the Data field:
5-19
5 Probability Distributions
5-20
Working with Distributions Through GUIs
Field Description
Name
Fit Name Enter a name for the fit in the Fit Name field.
Data The Data field contains a drop-down list of the data sets you
have created. Select the data set to which you want to fit a
distribution.
Distribution Select the type of distribution you want to fit from the
Distribution drop-down list. See “Available Distributions”
on page 5-22 for a list of distributions supported by the
Distribution Fitting Tool.
Only the distributions that apply to the values of the selected
data set are displayed in the Distribution field. For example,
positive distributions are not displayed when the data include
values that are zero or negative.
You can specify either a parametric or a nonparametric
distribution. When you select a parametric distribution from
the drop-down list, a description of its parameters is displayed
in the pane below the Exclusion rule field. The Distribution
Fitting Tool estimates these parameters to fit the distribution
to the data set. When you select Nonparametric fit, options
for the fit appear in the pane, as described in “Further Options
for Nonparametric Fits” on page 5-23.
Exclusion You can specify a rule to exclude some the data in the
Rule Exclusion rule field. You can create an exclusion rule by
clicking Exclude in the main window of the Distribution
Fitting Tool. For more information, see “Excluding Data” on
page 5-32.
Apply the New Fit. Click Apply to fit the distribution. For a parametric
fit, the Results pane displays the values of the estimated parameters. For a
nonparametric fit, the Results pane displays information about the fit.
When you click Apply, the main window of Distribution Fitting Tool displays
a plot of the distribution, along with the corresponding data.
5-21
5 Probability Distributions
Note When you click Apply, the title of the window changes to Edit Fit. You
can now make changes to the fit you just created and click Apply again to
save them. After closing the Edit Fit window, you can reopen it from the Fit
Manager window at any time to edit the fit.
Available Distributions. This section lists the distributions available in
the Distribution Fitting Tool.
Most, but not all, of the distributions available in the Distribution Fitting
Tool are supported elsewhere in Statistics Toolbox software (see “Supported
Distributions” on page 5-3), and have dedicated distribution fitting functions.
These functions are used to compute the majority of the fits in the Distribution
Fitting Tool, and are referenced in the list below.
Other fits are computed using functions internal to the Distribution Fitting
Tool. Distributions that do not have corresponding Statistics Toolbox
fitting functions are described in “Additional Distributions Available in the
Distribution Fitting Tool” on page 5-49.
Not all of the distributions listed below are available for all data sets. The
Distribution Fitting Tool determines the extent of the data (nonnegative, unit
interval, etc.) and displays appropriate distributions in the Distribution
drop-down list. Distribution data ranges are given parenthetically in the
list below.
• Beta (unit interval values) distribution, fit using the function betafit.
• Binomial (nonnegative values) distribution, fit using the function binopdf.
• Birnbaum-Saunders (positive values) distribution.
• Exponential (nonnegative values) distribution, fit using the function
expfit.
• Extreme value (all values) distribution, fit using the function evfit.
• Gamma (positive values) distribution, fit using the function gamfit.
• Generalized extreme value (all values) distribution, fit using the function
gevfit.
5-22
Working with Distributions Through GUIs
• Generalized Pareto (all values) distribution, fit using the function gpfit.
• Inverse Gaussian (positive values) distribution.
• Logistic (all values) distribution.
• Loglogistic (positive values) distribution.
• Lognormal (positive values) distribution, fit using the function lognfit.
• Nakagami (positive values) distribution.
• Negative binomial (nonnegative values) distribution, fit using the function
nbinpdf.
• Nonparametric (all values) distribution, fit using the function ksdensity.
See “Further Options for Nonparametric Fits” on page 5-23 for a description
of available options.
• Normal (all values) distribution, fit using the function normfit.
• Poisson (nonnegative integer values) distribution, fit using the function
poisspdf.
• Rayleigh (positive values) distribution using the function raylfit.
• Rician (positive values) distribution.
• t location-scale (all values) distribution.
• Weibull (positive values) distribution using the function wblfit.
Further Options for Nonparametric Fits. When you select
Non-parametric in the Distribution field, a set of options appears in the
pane below Exclusion rule, as shown in the following figure.
5-23
5 Probability Distributions
The options for nonparametric distributions are
• Kernel — Type of kernel function to use. The options are
- Normal
- Box
- Triangle
- Epanechnikov
• Bandwidth — The bandwidth of the kernel smoothing window. Select
auto for a default value that is optimal for estimating normal densities.
This value is displayed in the Fit results pane after you click Apply.
Select specify and enter a smaller value to reveal features such as multiple
modes or a larger value to make the fit smoother.
• Domain — The allowed x-values for the density. The options are
- unbounded — The density extends over the whole real line.
- positive — The density is restricted to positive values.
- specify — Enter lower and upper bounds for the domain of the density.
When you select positive or specify, the nonparametric fit has zero
probability outside the specified domain.
5-24
Working with Distributions Through GUIs
Displaying Results
This section explains the different ways to display results in the main window
of the Distribution Fitting Tool. The main window displays plots of
• The data sets for which you select Plot in the Data window.
• The fits for which you select Plot in the Fit Manager window.
• Confidence bounds for
- Data sets for which you select Bounds in the Data window.
- Fits for which you select Bounds in the Fit Manager.
The following fields are available.
Display Type. The Display Type field in the main window specifies the type
of plot displayed. Each type corresponds to a probability function, for example,
a probability density function. The following display types are available:
• Density (PDF) — Displays a probability density function (PDF) plot
for the fitted distribution. The main window displays data sets using a
probability histogram, in which the height of each rectangle is the fraction
of data points that lie in the bin divided by the width of the bin. This makes
the sum of the areas of the rectangles equal to 1.
• Cumulative probability (CDF) — Displays a cumulative probability
plot of the data. The main window displays data sets using a cumulative
probability step function. The height of each step is the cumulative sum of
the heights of the rectangles in the probability histogram.
• Quantile (inverse CDF) — Displays a quantile (inverse CDF) plot.
• Probability plot — Displays a probability plot of the data. You can
specify the type of distribution used to construct the probability plot in the
Distribution field, which is only available when you select Probability
plot. The choices for the distribution are
- Exponential
- Extreme value
- Logistic
- Log-Logistic
5-25
5 Probability Distributions
- Lognormal
- Normal
- Rayleigh
- Weibull
In addition to these choices, you can create a probability plot against a
parametric fit that you create in the New Fit panel. These fits are added
at the bottom of the Distribution drop-down list when you create them.
• Survivor function — Displays a survivor function plot of the data.
• Cumulative hazard — Displays a cumulative hazard plot of the data.
Note Some of these distributions are not available if the plotted data
includes 0 or negative values.
Confidence Bounds. You can display confidence bounds for data sets and
fits, provided that you set Display Type to Cumulative probability (CDF),
Survivor function, Cumulative hazard, or Quantile for fits only.
• To display bounds for a data set, select Bounds next to the data set in the
Data sets pane of the Data window.
• To display bounds for a fit, select Bounds next to the fit in the Fit
Manager window. Confidence bounds are not available for all fit types.
To set the confidence level for the bounds, select Confidence Level from the
View menu in the main window and choose from the options.
Managing Fits
This section describes how to manage fits that you have created. To begin,
click the Manage Fits button in the main window of the Distribution Fitting
Tool. This opens the Fit Manager window as shown in the following figure.
5-26
Working with Distributions Through GUIs
The Table of fits displays a list of the fits you create, with the following
options:
• Plot—Select Plot to display a plot of the fit in the main window of the
Distribution Fitting Tool. When you create a new fit, Plot is selected by
default. Clearing the Plot check box removes the fit from the plot in the
main window.
• Bounds—If Plot is selected, you can also select Bounds to display
confidence bounds in the plot. The bounds are displayed when you set
Display Type in the main window to one of the following:
- Cumulative probability (CDF)
- Quantile (inverse CDF)
- Survivor function
- Cumulative hazard
5-27
5 Probability Distributions
The Distribution Fitting Tool cannot display confidence bounds on density
(PDF) or probability plots. In addition, bounds are not supported for
nonparametric fits and some parametric fits.
Clearing the Bounds check box removes the confidence intervals from
the plot in the main window.
When you select a fit in the Table of fits, the following buttons are enabled
below the table:
- New Fit — Opens a New Fit window.
- Copy — Creates a copy of the selected fit.
- Edit — Opens an Edit Fit window, where you can edit the fit.
Note You can only edit the currently selected fit in the Edit Fit window.
To edit a different fit, select it in the Table of fits and click Edit to open
another Edit Fit window.
- Delete — Deletes the selected fit.
Evaluating Fits
The Evaluate window enables you to evaluate any fit at whatever points you
choose. To open the window, click the Evaluate button in the main window
of the Distribution Fitting Tool. The following figure shows the Evaluate
window.
5-28
Working with Distributions Through GUIs
The Evaluate window contains the following items:
• Fit pane — Displays the names of existing fits. Select one or more fits
that you want to evaluate. Using your platform specific functionality, you
can select multiple fits.
• Function — Select the type of probability function you want to evaluate
for the fit. The available functions are
- Density (PDF) — Computes a probability density function.
5-29
5 Probability Distributions
- Cumulative probability (CDF) — Computes a cumulative probability
function.
- Quantile (inverse CDF) — Computes a quantile (inverse CDF)
function.
- Survivor function — Computes a survivor function.
- Cumulative hazard — Computes a cumulative hazard function.
- Hazard rate — Computes the hazard rate.
• At x = — Enter a vector of points or the name of a workspace variable
containing a vector of points at which you want to evaluate the distribution
function. If you change Function to Quantile (inverse CDF), the field
name changes to At p = and you enter a vector of probability values.
• Compute confidence bounds — Select this box to compute confidence
bounds for the selected fits. The check box is only enabled if you set
Function to one of the following:
- Cumulative probability (CDF)
- Quantile (inverse CDF)
- Survivor function
- Cumulative hazard
The Distribution Fitting Tool cannot compute confidence bounds for
nonparametric fits and for some parametric fits. In these cases, the tool
returns NaN for the bounds.
• Level — Set the level for the confidence bounds.
• Plot function — Select this box to display a plot of the distribution
function, evaluated at the points you enter in the At x = field, in a new
window.
Note The settings for Compute confidence bounds, Level, and Plot
function do not affect the plots that are displayed in the main window of
the Distribution Fitting Tool. The settings only apply to plots you create by
clicking Plot function in the Evaluate window.
5-30
Working with Distributions Through GUIs
Click Apply to apply these settings to the selected fit. The following figure
shows the results of evaluating the cumulative density function for the fit My
fit, created in “Example: Fitting a Distribution” on page 5-39, at the points
in the vector -3:0.5:3.
The window displays the following values in the columns of the table to the
right of the Fit pane:
• X — The entries of the vector you enter in At x = field
• Y — The corresponding values of the CDF at the entries of X
5-31
5 Probability Distributions
• LB — The lower bounds for the confidence interval, if you select Compute
confidence bounds
• UB — The upper bounds for the confidence interval, if you select Compute
confidence bounds
To save the data displayed in the Evaluate window, click Export to
Workspace. This saves the values in the table to a matrix in the MATLAB
workspace.
Excluding Data
To exclude values from fit, click the Exclude button in the main window of
the Distribution Fitting Tool. This opens the Exclude window, in which you
can create rules for excluding specified values. You can use these rules to
exclude data when you create a new fit in the New Fit window. The following
figure shows the Exclude window.
To create an exclusion rule:
5-32
Working with Distributions Through GUIs
1 Exclusion Rule Name—Enter a name for the exclusion rule in the
Exclusion rule name field.
2 Exclude Sections—In the Exclude sections pane, you can specify
bounds for the excluded data:
• In the Lower limit: exclude Y drop-down list, select <= or < from the
drop-down list and enter a scalar in the field to the right. This excludes
values that are either less than or equal to or less than that scalar,
respectively.
• In the Upper limit: exclude Y drop-down list, select >= or > from the
drop-down list and enter a scalar in the field to the right to exclude
values that are either greater than or equal to or greater than the scalar,
respectively.
OR
Exclude Graphically—The Exclude Graphically button enables you
to define the exclusion rule by displaying a plot of the values in a data
set and selecting the bounds for the excluded data with the mouse. For
example, if you created the data set My data, described in “Creating
and Managing Data Sets” on page 5-14, select it from the drop-down list
next to Exclude graphically and then click the Exclude graphically
button. This displays the values in My data in a new window as shown in
the following figure.
5-33
5 Probability Distributions
To set a lower limit for the boundary of the excluded region, click Add
Lower Limit. This displays a vertical line on the left side of the plot
window. Move the line with the mouse to the point you where you want
the lower limit, as shown in the following figure.
5-34
Working with Distributions Through GUIs
Moving the vertical line changes the value displayed in the Lower limit:
exclude data field in the Exclude window, as shown in the following figure.
5-35
5 Probability Distributions
The value displayed corresponds to the x-coordinate of the vertical line.
Similarly, you can set the upper limit for the boundary of the excluded
region by clicking Add Upper Limit and moving the vertical line that
appears at the right side of the plot window. After setting the lower and
upper limits, click Close and return to the Exclude window.
3 Create Exclusion Rule—Once you have set the lower and upper limits
for the boundary of the excluded data, click Create Exclusion Rule
to create the new rule. The name of the new rule now appears in the
Existing exclusion rules pane.
When you select an exclusion rule in the Existing exclusion rules pane,
the following buttons are enabled:
• Copy — Creates a copy of the rule, which you can then modify. To save
the modified rule under a different name, click Create Exclusion Rule.
• View — Opens a new window in which you can see which data points
are excluded by the rule. The following figure shows a typical example.
5-36
Working with Distributions Through GUIs
The shaded areas in the plot graphically display which data points are
excluded. The table to the right lists all data points. The shaded rows
indicate excluded points:
• Rename — Renames the rule
• Delete — Deletes the rule
Once you define an exclusion rule, you can use it when you fit a distribution
to your data. The rule does not exclude points from the display of the data
set.
5-37
5 Probability Distributions
Saving and Loading Sessions
This section explains how to save your work in the current Distribution
Fitting Tool session and then load it in a subsequent session, so that you can
continue working where you left off.
Saving a Session. To save the current session, select Save Session from
the File menu in the main window. This opens a dialog box that prompts you
to enter a filename, such as my_session.dfit, for the session. Clicking Save
saves the following items created in the current session:
• Data sets
• Fits
• Exclusion rules
• Plot settings
• Bin width rules
Loading a Session. To load a previously saved session, select Load Session
from the File menu in the main window and enter the name of a previously
saved session. Clicking Open restores the information from the saved session
to the current session of the Distribution Fitting Tool.
5-38
Working with Distributions Through GUIs
Example: Fitting a Distribution
This section presents an example that illustrates how to use the Distribution
Fitting Tool. The example involves the following steps:
• “Step 1: Generate Random Data” on page 5-39
• “Step 2: Import Data” on page 5-39
• “Step 3: Create a New Fit” on page 5-42
Step 1: Generate Random Data. To try the example, first generate some
random data to which you will fit a distribution. The following command
generates a vector data, of length 100, whose entries are random numbers
from a normal distribution with mean.36 and standard deviation 1.4.
data = normrnd(.36, 1.4, 100, 1);
Step 2: Import Data. To import the vector data into the Distribution
Fitting Tool, click the Data button in main window. This opens the window
shown in the following figure.
5-39
5 Probability Distributions
Select data Enter name for data set
The Data field displays all numeric arrays in the MATLAB workspace. Select
data from the drop-down list, as shown in the following figure.
5-40
Working with Distributions Through GUIs
This displays a histogram of the data in the Data preview pane.
In the Data set name field, type a name for the data set, such as My data,
and click Create Data Set to create the data set. The main window of the
Distribution Fitting Tool now displays a larger version of the histogram in the
Data preview pane, as shown in the following figure.
5-41
5 Probability Distributions
Note Because the example uses random data, you might see a slightly
different histogram if you try this example for yourself.
Step 3: Create a New Fit. To fit a distribution to the data, click New Fit
in the main window of the Distribution Fitting Tool. This opens the window
shown in the following figure.
5-42
Working with Distributions Through GUIs
Select data set name Specify distribution type
To fit a normal distribution, the default entry of the Distribution field, to
My data:
1 Enter a name for the fit, such as My fit, in the Fit name field.
2 Select My data from the drop-down list in the Data field.
5-43
5 Probability Distributions
3 Click Apply.
The Results pane displays the mean and standard deviation of the normal
distribution that best fits My data, as shown in the following figure.
The main window of the Distribution Fitting Tool displays a plot of the
normal distribution with this mean and standard deviation, as shown in the
following figure.
5-44
Working with Distributions Through GUIs
5-45
5 Probability Distributions
Generating a File to Fit and Plot Distributions
The Generate M-file option in the File menu enables you to create a file that
• Fits the distributions used in the current session to any data vector in the
MATLAB workspace.
• Plots the data and the fits.
After you end the current session, you can use the file to create plots in a
standard MATLAB figure window, without having to reopen the Distribution
Fitting Tool.
As an example, assuming you created the fit described in “Creating a New
Fit” on page 5-19, do the following steps:
1 Select Generate M-file from the File menu.
2 Save the file as normal_fit.m in a folder on the MATLAB path.
You can then apply the function normal_fit to any vector of data in the
MATLAB workspace. For example, the following commands
new_data = normrnd(4.1, 12.5, 100, 1);
normal_fit(new_data)
legend('New Data', 'My fit')
fit a normal distribution to a data set and generate a plot of the data and
the fit.
5-46
Working with Distributions Through GUIs
Note By default, the file labels the data in the legend using the same name as
the data set in the Distribution Fitting Tool. You can change the label using
the legend command, as illustrated by the preceding example.
5-47
5 Probability Distributions
Using Custom Distributions
This section explains how to use custom distributions with the Distribution
Fitting Tool.
Defining Custom Distributions. To define a custom distribution, select
Define Custom Distribution from the File menu. This opens a file
template in the MATLAB editor. You then edit this file so that it computes
the distribution you want.
The template includes example code that computes the Laplace distribution,
beginning at the lines
% -
% Remove the following return statement to define the
% Laplace distributon
% -
return
To use this example, simply delete the command return and save the
file. If you save the template in a folder on the MATLAB path, under its
default name dfittooldists.m, the Distribution Fitting Tool reads it in
automatically when you start the tool. You can also save the template under a
different name, such as laplace.m, and then import the custom distribution
as described in the following section.
Importing Custom Distributions. To import a custom distribution, select
Import Custom Distributions from the File menu. This opens a dialog box
in which you can select the file that defines the distribution. For example, if
you created the file laplace.m, as described in the preceding section, you can
enter laplace.m and select Open in the dialog box. The Distribution field of
the New Fit window now contains the option Laplace.
5-48
Working with Distributions Through GUIs
Additional Distributions Available in the Distribution Fitting Tool
The following distributions are available in the Distribution Fitting Tool,
but do not have dedicated distribution functions as described in “Statistics
Toolbox Distribution Functions” on page 5-52. The distributions can be used
with the functions pdf, cdf, icdf, and mle in a limited capacity. See the
reference pages for these functions for details on the limitations.
• “Birnbaum-Saunders Distribution” on page B-10
• “Inverse Gaussian Distribution” on page B-45
• “Loglogistic Distribution” on page B-50
• “Logistic Distribution” on page B-49
• “Nakagami Distribution” on page B-70
• “Rician Distribution” on page B-93
• “t Location-Scale Distribution” on page B-97
For a complete list of the distributions available for use with the Distribution
Fitting Tool, see “Supported Distributions” on page 5-3. Distributions listing
dfittool in the fit column of the tables in that section can be used with
the Distribution Fitting Tool.
Visually Exploring Random Number Generation
The Random Number Generation Tool is a graphical user interface that
generates random samples from specified probability distributions and
displays the samples as histograms. Use the tool to explore the effects of
changing parameters and sample size on the distributions.
Run the tool by typing randtool at the command line.
5-49
5 Probability Distributions
Choose distribution Sample size
Histogram
Parameter
bounds
Parameter
value
Parameter
control Additional Sample again Export to
parameters from the same workspace
distribution
Start by selecting a distribution, then enter the desired sample size.
5-50
Working with Distributions Through GUIs
You can also
• Use the controls at the bottom of the window to set parameter values for
the distribution and to change their upper and lower bounds.
• Draw another sample from the same distribution, with the same size and
parameters.
• Export the current sample to your workspace. A dialog box enables you
to provide a name for the sample.
5-51
5 Probability Distributions
Statistics Toolbox Distribution Functions
In this section...
“Probability Density Functions” on page 5-52
“Cumulative Distribution Functions” on page 5-62
“Inverse Cumulative Distribution Functions” on page 5-66
“Distribution Statistics Functions” on page 5-68
“Distribution Fitting Functions” on page 5-70
“Negative Log-Likelihood Functions” on page 5-77
“Random Number Generators” on page 5-80
For each distribution supported by Statistics Toolbox software, a selection of
the distribution functions described in this section is available for statistical
programming. This section gives a general overview of the use of each type
of function, independent of the particular distribution. For specific functions
available for specific distributions, see “Supported Distributions” on page 5-3.
Probability Density Functions
• “Estimating PDFs with Parameters” on page 5-52
• “Estimating PDFs without Parameters” on page 5-55
Estimating PDFs with Parameters
Probability density functions (pdfs) for supported Statistics Toolbox
distributions all end with pdf, as in binopdf or exppdf. For more information
on specific function names for specific distributions see “Supported
Distributions” on page 5-3.
Each function represents a parametric family of distributions. Input
arguments are arrays of outcomes followed by a list of parameter values
specifying a particular member of the distribution family.
For discrete distributions, the pdf assigns a probability to each outcome. In
this context, the pdf is often called a probability mass function (pmf).
5-52
Statistics Toolbox™ Distribution Functions
For example, the discrete binomial pdf
⎛ n⎞
f (k) = ⎜ ⎟ pk (1 − p)n− k
⎝ k⎠
assigns probability to the event of k successes in n trials of a Bernoulli process
(such as coin flipping) with probability p of success at each trial. Each of the
integers k = 0, 1, 2, ..., n is assigned a positive probability, with the sum of the
probabilities equal to 1. Compute the probabilities with the binopdf function:
p = 0.2; % Probability of success for each trial
n = 10; % Number of trials
k = 0:n; % Outcomes
m = binopdf(k,n,p); % Probability mass vector
bar(k,m) % Visualize the probability distribution
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
grid on
5-53
5 Probability Distributions
For continuous distributions, the pdf assigns a probability density to each
outcome. The probability of any single outcome is zero. The pdf must be
integrated over a set of outcomes to compute the probability that an outcome
falls within that set. The integral over the entire set of outcomes is 1.
For example, the continuous exponential pdf
f (t) = λe−λt
is used to model the probability that a process with constant failure rate λ will
have a failure within time t . Each time t > 0 is assigned a positive probability
density. Densities are computed with the exppdf function:
lambda = 2; % Failure rate
t = 0:0.01:3; % Outcomes
f = exppdf(t,1/lambda); % Probability density vector
plot(t,f) % Visualize the probability distribution
grid on
5-54
Statistics Toolbox™ Distribution Functions
Probabilities for continuous pdfs can be computed with the quad function.
In the example above, the probability of failure in the time interval [0,1] is
computed as follows:
f_lambda = @(t)exppdf(t,1/lambda); % Pdf with fixed lambda
P = quad(f_lambda,0,1) % Integrate from 0 to 1
P =
0.8647
Alternatively, the cumulative distribution function (cdf) for the exponential
function, expcdf, can be used:
P = expcdf(1,1/lambda) % Cumulative probability from 0 to 1
P =
0.8647
Estimating PDFs without Parameters
A distribution of data can be described graphically with a histogram:
cars = load('carsmall','MPG','Origin');
MPG = cars.MPG;
hist(MPG)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
5-55
5 Probability Distributions
You can also describe a data distribution by estimating its density.
The ksdensity function does this using a kernel smoothing method. A
nonparametric density estimate of the previous data, using the default kernel
and bandwidth, is given by:
[f,x] = ksdensity(MPG);
plot(x,f);
title('Density estimate for MPG')
5-56
Statistics Toolbox™ Distribution Functions
Controlling Probability Density Curve Smoothness. The choice of
kernel bandwidth controls the smoothness of the probability density curve.
The following graph shows the density estimate for the same mileage data
using different bandwidths. The default bandwidth is in blue and looks
like the preceding graph. Estimates for smaller and larger bandwidths are
in red and green.
The first call to ksdensity returns the default bandwidth, u, of the kernel
smoothing function. Subsequent calls modify this bandwidth.
[f,x,u] = ksdensity(MPG);
plot(x,f)
title('Density estimate for MPG')
hold on
5-57
5 Probability Distributions
[f,x] = ksdensity(MPG,'width',u/3);
plot(x,f,'r');
[f,x] = ksdensity(MPG,'width',u*3);
plot(x,f,'g');
legend('default width','1/3 default','3*default')
hold off
The default bandwidth seems to be doing a good job—reasonably smooth,
but not so smooth as to obscure features of the data. This bandwidth is
the one that is theoretically optimal for estimating densities for the normal
distribution.
5-58
Statistics Toolbox™ Distribution Functions
The green curve shows a density with the kernel bandwidth set too high.
This curve smooths out the data so much that the end result looks just like
the kernel function. The red curve has a smaller bandwidth and is rougher
looking than the blue curve. It may be too rough, but it does provide an
indication that there might be two major peaks rather than the single peak
of the blue curve. A reasonable choice of width might lead to a curve that is
intermediate between the red and blue curves.
Specifying Kernel Smoothing Functions. You can also specify a kernel
function by supplying either the function name or a function handle. The four
preselected functions, 'normal', 'epanechnikov', 'box', and 'triangle',
are all scaled to have standard deviation equal to 1, so they perform a
comparable degree of smoothing.
Using default bandwidths, you can now plot the same mileage data, using
each of the available kernel functions.
hname = {'normal' 'epanechnikov' 'box' 'triangle'};
colors = {'r' 'b' 'g' 'm'};
for j=1:4
[f,x] = ksdensity(MPG,'kernel',hname{j});
plot(x,f,colors{j});
hold on;
end
legend(hname{:});
hold off
5-59
5 Probability Distributions
The density estimates are roughly comparable, but the box kernel produces a
density that is rougher than the others.
Comparing Density Estimates. While it is difficult to overlay two
histograms to compare them, you can easily overlay smooth density estimates.
For example, the following graph shows the MPG distributions for cars from
different countries of origin:
Origin = cellstr(cars.Origin);
I = strcmp('USA',Origin);
J = strcmp('Japan',Origin);
K = ~(I|J);
MPG_USA = MPG(I);
MPG_Japan = MPG(J);
MPG_Europe = MPG(K);
5-60
Statistics Toolbox™ Distribution Functions
[fI,xI] = ksdensity(MPG_USA);
plot(xI,fI,'b')
hold on
[fJ,xJ] = ksdensity(MPG_Japan);
plot(xJ,fJ,'r')
[fK,xK] = ksdensity(MPG_Europe);
plot(xK,fK,'g')
legend('USA','Japan','Europe')
hold off
For piecewise probability density estimation, using kernel smoothing in the
center of the distribution and Pareto distributions in the tails, see “Fitting
Piecewise Distributions” on page 5-72.
5-61
5 Probability Distributions
Cumulative Distribution Functions
• “Estimating Parametric CDFs” on page 5-62
• “Estimating Empirical CDFs” on page 5-63
Estimating Parametric CDFs
Cumulative distribution functions (cdfs) for supported Statistics Toolbox
distributions all end with cdf, as in binocdf or expcdf. Specific function
names for specific distributions can be found in “Supported Distributions”
on page 5-3.
Each function represents a parametric family of distributions. Input
arguments are arrays of outcomes followed by a list of parameter values
specifying a particular member of the distribution family.
For discrete distributions, the cdf F is related to the pdf f by
F ( x) = ∑ f ( y)
y≤ x
For continuous distributions, the cdf F is related to the pdf f by
x
F ( x) = ∫ f ( y) dy
−∞
Cdfs are used to compute probabilities of events. In particular, if F is a cdf
and x and y are outcomes, then
• P(y ≤ x) = F(x)
• P(y > x) = 1 – F(x)
• P(x1 < y ≤ x2) = F(x2) – F(x1)
For example, the t-statistic
5-62
Statistics Toolbox™ Distribution Functions
x−
t=
s/ n
follows a Student’s t distribution with n – 1 degrees of freedom when computed
from repeated random samples from a normal population with mean μ. Here
x is the sample mean, s is the sample standard deviation, and n is the sample
size. The probability of observing a t-statistic greater than or equal to the
value computed from a sample can be found with the tcdf function:
mu = 1; % Population mean
sigma = 2; % Population standard deviation
n = 100; % Sample size
x = normrnd(mu,sigma,n,1); % Random sample from population
xbar = mean(x); % Sample mean
s = std(x); % Sample standard deviation
t = (xbar-mu)/(s/sqrt(n)) % t-statistic
t =
0.2489
p = 1-tcdf(t,n-1) % Probability of larger t-statistic
p =
0.4020
This probability is the same as the p value returned by a t-test of the null
hypothesis that the sample comes from a normal population with mean μ:
[h,ptest] = ttest(x,mu,0.05,'right')
h =
0
ptest =
0.4020
Estimating Empirical CDFs
The ksdensity function produces an empirical version of a probability
density function (pdf). That is, instead of selecting a density with a particular
parametric form and estimating the parameters, it produces a nonparametric
density estimate that adapts itself to the data.
Similarly, it is possible to produce an empirical version of the cumulative
distribution function (cdf). The ecdf function computes this empirical cdf. It
5-63
5 Probability Distributions
returns the values of a function F such that F(x) represents the proportion of
observations in a sample less than or equal to x.
The idea behind the empirical cdf is simple. It is a function that assigns
probability 1/n to each of n observations in a sample. Its graph has a
stair-step appearance. If a sample comes from a distribution in a parametric
family (such as a normal distribution), its empirical cdf is likely to resemble
the parametric distribution. If not, its empirical distribution still gives an
estimate of the cdf for the distribution that generated the data.
The following example generates 20 observations from a normal distribution
with mean 10 and standard deviation 2. You can use ecdf to calculate the
empirical cdf and stairs to plot it. Then you overlay the normal distribution
curve on the empirical function.
x = normrnd(10,2,20,1);
[f,xf] = ecdf(x);
stairs(xf,f)
hold on
xx=linspace(5,15,100);
yy = normcdf(xx,10,2);
plot(xx,yy,'r:')
hold off
legend('Empirical cdf','Normal cdf',2)
5-64
Statistics Toolbox™ Distribution Functions
The empirical cdf is especially useful in survival analysis applications. In
such applications the data may be censored, that is, not observed exactly.
Some individuals may fail during a study, and you can observe their failure
time exactly. Other individuals may drop out of the study, or may not fail
until after the study is complete. The ecdf function has arguments for dealing
with censored data. In addition, you can use the coxphfit function with
individuals that have predictors that are not the same.
For piecewise probability density estimation, using the empirical cdf in the
center of the distribution and Pareto distributions in the tails, see “Fitting
Piecewise Distributions” on page 5-72.
5-65
5 Probability Distributions
Inverse Cumulative Distribution Functions
Inverse cumulative distribution functions for supported Statistics Toolbox
distributions all end with inv, as in binoinv or expinv. Specific function
names for specific distributions can be found in “Supported Distributions”
on page 5-3.
Each function represents a parametric family of distributions. Input
arguments are arrays of cumulative probabilities from 0 to 1 followed by a list
of parameter values specifying a particular member of the distribution family.
For continuous distributions, the inverse cdf returns the unique outcome
whose cdf value is the input cumulative probability.
For example, the expinv function can be used to compute inverses of
exponential cumulative probabilities:
x = 0.5:0.2:1.5 % Outcomes
x =
0.5000 0.7000 0.9000 1.1000 1.3000 1.5000
p = expcdf(x,1) % Cumulative probabilities
p =
0.3935 0.5034 0.5934 0.6671 0.7275 0.7769
expinv(p,1) % Return original outcomes
ans =
0.5000 0.7000 0.9000 1.1000 1.3000 1.5000
For discrete distributions, there may be no outcome whose cdf value is the
input cumulative probability. In these cases, the inverse cdf returns the first
outcome whose cdf value equals or exceeds the input cumulative probability.
For example, the binoinv function can be used to compute inverses of
binomial cumulative probabilities:
x = 0:5 % Some possible outcomes
p = binocdf(x,10,0.2) % Their cumulative probabilities
p =
0.1074 0.3758 0.6778 0.8791 0.9672 0.9936
q = [.1 .2 .3 .4] % New trial probabilities
5-66
Statistics Toolbox™ Distribution Functions
q =
0.1000 0.2000 0.3000 0.4000
binoinv(q,10,0.2) % Their corresponding outcomes
ans =
0 1 1 2
The inverse cdf is useful in hypothesis testing, where critical outcomes of a
test statistic are computed from cumulative significance probabilities. For
example, norminv can be used to compute a 95% confidence interval under
the assumption of normal variability:
p = [0.025 0.975]; % Interval containing 95% of [0,1]
x = norminv(p,0,1) % Assume standard normal variability
x =
-1.9600 1.9600 % 95% confidence interval
n = 20; % Sample size
y = normrnd(8,1,n,1); % Random sample (assume mean is unknown)
ybar = mean(y);
ci = ybar + (1/sqrt(n))*x % Confidence interval for mean
ci =
7.6779 8.5544
5-67
5 Probability Distributions
Distribution Statistics Functions
Distribution statistics functions for supported Statistics Toolbox distributions
all end with stat, as in binostat or expstat. Specific function names for
specific distributions can be found in “Supported Distributions” on page 5-3.
Each function represents a parametric family of distributions. Input
arguments are lists of parameter values specifying a particular member
of the distribution family. Functions return the mean and variance of the
distribution, as a function of the parameters.
For example, the wblstat function can be used to visualize the mean of the
Weibull distribution as a function of its two distribution parameters:
a = 0.5:0.1:3;
b = 0.5:0.1:3;
[A,B] = meshgrid(a,b);
M = wblstat(A,B);
surfc(A,B,M)
5-68
Statistics Toolbox™ Distribution Functions
5-69
5 Probability Distributions
Distribution Fitting Functions
• “Fitting Regular Distributions” on page 5-70
• “Fitting Piecewise Distributions” on page 5-72
Fitting Regular Distributions
Distribution fitting functions for supported Statistics Toolbox distributions all
end with fit, as in binofit or expfit. Specific function names for specific
distributions can be found in “Supported Distributions” on page 5-3.
Each function represents a parametric family of distributions. Input
arguments are arrays of data, presumed to be samples from some member
of the selected distribution family. Functions return maximum likelihood
estimates (MLEs) of distribution parameters, that is, parameters for the
distribution family member with the maximum likelihood of producing the
data as a random sample.
The Statistics Toolbox function mle is a convenient front end to the individual
distribution fitting functions, and more. The function computes MLEs for
distributions beyond those for which Statistics Toolbox software provides
specific pdf functions.
For some pdfs, MLEs can be given in closed form and computed directly.
For other pdfs, a search for the maximum likelihood must be employed. The
search can be controlled with an options input argument, created using
the statset function. For efficient searches, it is important to choose a
reasonable distribution model and set appropriate convergence tolerances.
MLEs can be heavily biased, especially for small samples. As sample size
increases, however, MLEs become unbiased minimum variance estimators
with approximate normal distributions. This is used to compute confidence
bounds for the estimates.
For example, consider the following distribution of means from repeated
random samples of an exponential distribution:
mu = 1; % Population parameter
n = 1e3; % Sample size
ns = 1e4; % Number of samples
5-70
Statistics Toolbox™ Distribution Functions
samples = exprnd(mu,n,ns); % Population samples
means = mean(samples); % Sample means
The Central Limit Theorem says that the means will be approximately
normally distributed, regardless of the distribution of the data in the samples.
The normfit function can be used to find the normal distribution that best
fits the means:
[muhat,sigmahat,muci,sigmaci] = normfit(means)
muhat =
1.0003
sigmahat =
0.0319
muci =
0.9997
1.0010
sigmaci =
0.0314
0.0323
The function returns MLEs for the mean and standard deviation and their
95% confidence intervals.
To visualize the distribution of sample means together with the fitted normal
distribution, you must scale the fitted pdf, with area = 1, to the area of the
histogram being used to display the means:
numbins = 50;
hist(means,numbins)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
hold on
[bincounts,binpositions] = hist(means,numbins);
binwidth = binpositions(2) - binpositions(1);
histarea = binwidth*sum(bincounts);
x = binpositions(1):0.001:binpositions(end);
y = normpdf(x,muhat,sigmahat);
plot(x,histarea*y,'r','LineWidth',2)
5-71
5 Probability Distributions
Fitting Piecewise Distributions
The parametric methods discussed in “Fitting Regular Distributions” on
page 5-70 fit data samples with smooth distributions that have a relatively
low-dimensional set of parameters controlling their shape. These methods
work well in many cases, but there is no guarantee that a given sample will be
described accurately by any of the supported Statistics Toolbox distributions.
The empirical distributions computed by ecdf and discussed in “Estimating
Empirical CDFs” on page 5-63 assign equal probability to each observation in
a sample, providing an exact match of the sample distribution. However, the
distributions are not smooth, especially in the tails where data may be sparse.
The paretotails function fits a distribution by piecing together the empirical
distribution in the center of the sample with smooth generalized Pareto
distributions (GPDs) in the tails. The output is an object of the paretotails
class, with associated methods to evaluate the cdf, inverse cdf, and other
functions of the fitted distribution.
5-72
Statistics Toolbox™ Distribution Functions
As an example, consider the following data, with about 20% outliers:
left_tail = -exprnd(1,10,1);
right_tail = exprnd(5,10,1);
center = randn(80,1);
data = [left_tail;center;right_tail];
Neither a normal distribution nor a t distribution fits the tails very well:
probplot(data);
p = fitdist(data,'tlocationscale');
h = probplot(gca,p);
set(h,'color','r','linestyle','-')
title('{\bf Probability Plot}')
legend('Normal','Data','t','Location','NW')
5-73
5 Probability Distributions
On the other hand, the empirical distribution provides a perfect fit, but the
outliers make the tails very discrete:
ecdf(data)
5-74
Statistics Toolbox™ Distribution Functions
Random samples generated from this distribution by inversion might include,
for example, values around 4.33 and 9.25, but nothing in-between.
The paretotails function provides a single, well-fit model for the entire
sample. The following uses generalized Pareto distributions (GPDs) for the
lower and upper 10% of the data:
pfit = paretotails(data,0.1,0.9)
pfit =
Piecewise distribution with 3 segments
-Inf < x < -1.30726 (0 < p < 0.1)
lower tail, GPD(-1.10167,1.12395)
-1.30726 < x < 1.27213 (0.1 < p < 0.9)
interpolated empirical cdf
1.27213 < x < Inf (0.9 < p < 1)
5-75
5 Probability Distributions
upper tail, GPD(1.03844,0.726038)
x = -4:0.01:10;
plot(x,cdf(pfit,x))
Access information about the fit using the methods of the paretotails class.
Options allow for nonparametric estimation of the center of the cdf.
5-76
Statistics Toolbox™ Distribution Functions
Negative Log-Likelihood Functions
Negative log-likelihood functions for supported Statistics Toolbox
distributions all end with like, as in explike. Specific function names for
specific distributions can be found in “Supported Distributions” on page 5-3.
Each function represents a parametric family of distributions. Input
arguments are lists of parameter values specifying a particular member of
the distribution family followed by an array of data. Functions return the
negative log-likelihood of the parameters, given the data.
Negative log-likelihood functions are used as objective functions in
search algorithms such as the one implemented by the MATLAB function
fminsearch. Additional search algorithms are implemented by Optimization
Toolbox™ functions and Global Optimization Toolbox functions.
When used to compute maximum likelihood estimates (MLEs), negative
log-likelihood functions allow you to choose a search algorithm and exercise
low-level control over algorithm execution. By contrast, the functions
discussed in “Distribution Fitting Functions” on page 5-70 use preset
algorithms with options limited to those set by the statset function.
Likelihoods are conditional probability densities. A parametric family of
distributions is specified by its pdf f(x,a), where x and a represent the
variables and parameters, respectively. When a is fixed, the pdf is used
to compute the density at x, f(x|a). When x is fixed, the pdf is used to
compute the likelihood of the parameters a, f(a|x). The joint likelihood of the
parameters over an independent random sample X is
L(a) = ∏ f (a| x)
x∈ X
Given X, MLEs maximize L(a) over all possible a.
In numerical algorithms, the log-likelihood function, log(L(a)), is
(equivalently) optimized. The logarithm transforms the product of potentially
small likelihoods into a sum of logs, which is easier to distinguish from 0
in computation. For convenience, Statistics Toolbox negative log-likelihood
functions return the negative of this sum, since the optimization algorithms to
which the values are passed typically search for minima rather than maxima.
5-77
5 Probability Distributions
For example, use gamrnd to generate a random sample from a specific gamma
distribution:
a = [1,2];
X = gamrnd(a(1),a(2),1e3,1);
Given X, the gamlike function can be used to visualize the likelihood surface
in the neighborhood of a:
mesh = 50;
delta = 0.5;
a1 = linspace(a(1)-delta,a(1)+delta,mesh);
a2 = linspace(a(2)-delta,a(2)+delta,mesh);
logL = zeros(mesh); % Preallocate memory
for i = 1:mesh
for j = 1:mesh
logL(i,j) = gamlike([a1(i),a2(j)],X);
end
end
[A1,A2] = meshgrid(a1,a2);
surfc(A1,A2,logL)
5-78
Statistics Toolbox™ Distribution Functions
The MATLAB function fminsearch is used to search for the minimum of
the likelihood surface:
LL = @(u)gamlike([u(1),u(2)],X); % Likelihood given X
MLES = fminsearch(LL,[1,2])
MLES =
1.0231 1.9729
These can be compared to the MLEs returned by the gamfit function, which
uses a combination search and solve algorithm:
ahat = gamfit(X)
ahat =
1.0231 1.9728
The MLEs can be added to the surface plot (rotated to show the minimum):
hold on
plot3(MLES(1),MLES(2),LL(MLES),...
'ro','MarkerSize',5,...
'MarkerFaceColor','r')
5-79
5 Probability Distributions
Random Number Generators
The Statistics Toolbox supports the generation of random numbers from
various distributions. Each RNG represents a parametric family of
distributions. RNGs return random numbers from the specified distribution
in an array of the specified dimensions. Specific RNG names for specific
distributions are in “Supported Distributions” on page 5-3.
Other random number generation functions which do not support specific
distributions include:
• cvpartition
• hmmgenerate
• lhsdesign
• lhsnorm
• mhsample
• random
• randsample
• slicesample
RNGs in Statistics Toolbox software depend on MATLAB’s default random
number stream via the rand and randn functions, each RNG uses one of
the techniques discussed in “Common Generation Methods” on page 6-5 to
generate random numbers from a given distribution.
By controlling the default random number stream and its state, you can
control how the RNGs in Statistics Toolbox software generate random values.
For example, to reproduce the same sequence of values from an RNG, you
can save and restore the default stream’s state, or reset the default stream.
For details on managing the default random number stream, see “Managing
the Default Stream”.
MATLAB initializes the default random number stream to the same state
each time it starts up. Thus, RNGs in Statistics Toolbox software will
generate the same sequence of values for each MATLAB session unless you
modify that state at startup. One simple way to do that is to add commands
to startup.m such as
5-80
Statistics Toolbox™ Distribution Functions
stream = RandStream('mt19937ar','seed',sum(100*clock));
RandStream.setDefaultStream(stream);
that initialize MATLAB’s default random number stream to a different state
for each session.
5-81
5 Probability Distributions
Dependencies of the Random Number Generators
The following table lists the dependencies of Statistics Toolbox RNGs on the
MATLAB base RNGs rand and/or randn.
RNG MATLAB Base RNG
betarnd rand, randn
binornd rand
chi2rnd rand, randn
evrnd rand
exprnd rand
frnd rand, randn
gamrnd rand, randn
geornd rand
gevrnd rand
gprnd rand
hygernd rand
iwishrnd rand, randn
johnsrnd randn
lhsdesign rand
lhsnorm rand
lognrnd randn
mhsample rand or randn, depending on
the RNG given for the proposal
distribution
mvnrnd randn
mvtrnd rand, randn
nbinrnd rand, randn
5-82
Statistics Toolbox™ Distribution Functions
RNG MATLAB Base RNG
ncfrnd rand, randn
nctrnd rand, randn
ncx2rnd randn
normrnd randn
pearsrnd rand or randn, depending on the
distribution type
poissrnd rand, randn
random rand or randn, depending on the
specified distribution
randsample rand
raylrnd randn
slicesample rand
trnd rand, randn
unidrnd rand
unifrnd rand
wblrnd rand
wishrnd rand, randn
5-83
5 Probability Distributions
Using Probability Distribution Objects
In this section...
“Using Distribution Objects” on page 5-84
“What are Objects?” on page 5-85
“Creating Distribution Objects” on page 5-88
“Object-Supported Distributions” on page 5-89
“Performing Calculations Using Distribution Objects” on page 5-90
“Capturing Results Using Distribution Objects” on page 5-97
Using Distribution Objects
For many distributions supported by Statistics Toolbox software, objects are
available for statistical analysis. This section gives a general overview of the
uses of distribution objects, including sample work flows. For information
on objects available for specific distributions, see “Object-Supported
Distributions” on page 5-89.
Probability distribution objects allow you to easily fit, access, and store
distribution information for a given data set. The following operations are
easier to perform using distribution objects:
• Grouping a single dataset in a number of different ways using group
names, and then fit a distribution to each group. For an example of how
to fit distributions to grouped data, see “Example: Fitting Distributions to
Grouped Data Within a Single Dataset” on page 5-91.
• Fitting different distributions to the same set of data. For an example of
how objects make fitting multiple distribution types easier, see “Example:
Fitting Multiple Distribution Types to a Single Dataset” on page 5-95.
• Sharing fitted distributions across workspaces. For an example of sharing
information using probability distribution objects, see “Example: Saving
and Sharing Distribution Fit Data” on page 5-97.
5-84
Using Probability Distribution Objects
Deciding to Use Distribution Objects
If you know the type of distribution you would like to use, objects provide a
less complex interface than functions and a more efficient functionality than
the dfittool GUI.
If you are a novice statistician who would like to explore how various
distributions look without having to manipulate data, see “Working with
Distributions Through GUIs” on page 5-9.
If you have no data to fit, but want to calculate a pdf, cdf, etc for various
parameters, see “Statistics Toolbox Distribution Functions” on page 5-52.
What are Objects?
Objects are, in short, a convenient way of storing data. They allow you to set
rules for the types of data to store, while maintaining some flexibility for the
actual values of the data. For example, in statistics groups of distributions
have some general things in common:
• All distributions have a name (ex, Normal).
• Parametric distributions have parameters.
• Nonparametric distributions have kernel-smoothing functions.
Objects store all this information within properties. Classes of related
objects (for example, all univariate parametric distributions) have the same
properties with values and types relevant to a specified distribution. In
addition to storing information within objects, you can perform certain actions
(called methods) on objects.
Subclasses (for example, ProbDistParametric is a subclass of ProbDist)
contain the same properties and methods as the original class, in addition to
other properties relevant to that subclass. This concept is called inheritance.
Inheritance means that subclasses of a class have all of its properties and
methods. For example, parametric distributions, which are a subset (subclass)
of probability distributions, have input data and a distribution name. The
following diagram illustrates this point:
5-85
5 Probability Distributions
The left side of this diagram shows the inheritance line from all probability
distributions down to univariate parametric probability distributions. The
right side shows the lineage down to univariate kernel distributions. Here is
how to interpret univariate parametric distribution lineage:
• ProbDist is a class of objects that includes all probability distributions. All
probability distribution objects have at least these properties:
- DistName — the name of the distribution (for example Normal or
Weibull)
- InputData — the data fit to the distribution
In addition, you can perform the following actions on these objects, using
the following methods:
- cdf — Return the cumulative distribution function for a specified
distribution.
5-86
Using Probability Distribution Objects
- pdf — Return the probability density function for a specified distribution.
- random — Generate random numbers based on a specified distribution.
• ProbDistParametric is a class of objects that includes all parametric
probability distributions. All parametric probability distribution objects
have the properties and methods of a ProbDist object, in addition to at
least the following properties:
- NLogL — Negative log likelihood for input data
- NumParams — Number of parameters for that distribution
- ParamCov — Covariance matrix of parameter estimates
- ParamDescription — Descriptions of parameters
- ParamNames — Names of parameters
- Params — Values of parameters
No additional unique methods apply to ProbDistParametric objects.
• ProbDistUnivParam is a class of objects that includes only univariate
parametric probability distributions. In addition to the properties and
methods of ProbDist and ProbDistParametric objects, these objects also
have at least the following methods:
- icdf — Return the inverse cumulative distribution function for a
specified distribution based on a given set of data.
- iqr — Return the interquartile range for a specified distribution based
on a given set of data.
- mean — Return the mean for a specified distribution based on a given
set of data.
- median — Return the median for a specified distribution based on a
given set of data.
- paramci — Return the parameter confidence intervals for a specified
distribution based on a given set of data.
- std — Return the standard deviation for a specified distribution based
on a given set of data.
- var — Return the variance for a specified distribution based on a given
set of data.
No additional unique properties apply to ProbDistUnivParam objects.
5-87
5 Probability Distributions
The univariate nonparametric lineage reads in a similar manner, with
different properties and methods. For more information on nonparametric
objects and their methods and properties, see ProbDistKernel and
ProbDistUnivKernel.
For more detailed information on object-oriented programming in MATLAB,
see Object-Oriented Programming.
Creating Distribution Objects
There are two ways to create distribution objects:
• Use the fitdist function. See “Creating Distribution Objects Using
fitdist” on page 5-88.
• Use the object constructor. See “Creating Distribution Objects Using
Constructors” on page 5-88.
Creating Distribution Objects Using fitdist
Using the fitdist function is the simplest way of creating distribution
objects. Like the *fit functions, fitdist fits your data to a specified
distribution and returns relevant distribution information. fitdist creates
an object relevant to the type of distribution you specify: if you specify a
parametric distribution, it returns a ProbDistUnivParam object. For examples
of how to use fitdist to fit your data, see “Performing Calculations Using
Distribution Objects” on page 5-90.
Creating Distribution Objects Using Constructors
If you know the distribution you would like to use and would like to create a
univariate parametric distribution with known parameters, you can use the
ProbDistUnivParam constructor. For example, create a normal distribution
with mean 100 and standard deviation 10:
pd = ProbDistUnivParam('normal',[100 10])
For nonparametric distributions, you must have a dataset. Using
fitdist is a simpler way to fit nonparametric data, but you can use
the ProbDistUnivKernel constructor as well. For example, create a
nonparametric distribution of the MPG data from carsmall.mat:
5-88
Using Probability Distribution Objects
load carsmall
pd = ProbDistUnivKernel(MPG)
Object-Supported Distributions
Object-oriented programming in the Statistics Toolbox supports the following
distributions.
Parametric Distributions
Use the following distribution to create ProbDistUnivParam objects using
fitdist. For more information on the cumulative distribution function (cdf)
and probability density function (pdf) methods, as well as other available
methods, see the ProbDistUnivParam class reference page.
Supported Distribution Input to fitdist
“Beta Distribution” on page B-4 'beta'
“Binomial Distribution” on page B-7 'binomial'
“Birnbaum-Saunders Distribution” 'birnbaumsaunders'
on page B-10
“Exponential Distribution” on page 'exponential'
B-16
“Extreme Value Distribution” on 'extreme value' or 'ev'
page B-19
“Gamma Distribution” on page B-27 'gamma'
“Generalized Extreme Value 'generalized extreme value' or
Distribution” on page B-32 'gev'
“Generalized Pareto Distribution” on 'generalized pareto' or 'gp'
page B-37
“Inverse Gaussian Distribution” on 'inversegaussian'
page B-45
“Logistic Distribution” on page B-49 'logistic'
“Loglogistic Distribution” on page 'loglogistic'
B-50
5-89
5 Probability Distributions
Supported Distribution Input to fitdist
“Lognormal Distribution” on page 'lognormal'
B-51
“Nakagami Distribution” on page 'nakagami'
B-70
“Negative Binomial Distribution” on 'negative binomial' or 'nbin'
page B-72
“Normal Distribution” on page B-83 'normal'
“Poisson Distribution” on page B-89 'poisson'
“Rayleigh Distribution” on page B-91 'rayleigh'
“Rician Distribution” on page B-93 'rician'
“t Location-Scale Distribution” on 'tlocationscale'
page B-97
“Weibull Distribution” on page B-103 'weibull' or 'wbl'
Nonparametric Distributions
Use the following distributions to create ProbDistUnivKernel objects.
For more information on the cumulative distribution function (cdf) and
probability density function (pdf) methods, as well as other available
methods, see the ProbDistUnivKernel class reference page.
Supported Distribution Input to fitdist
“Nonparametric Distributions” on 'kernel'
page B-82
Performing Calculations Using Distribution Objects
Distribution objects make it easier for you to perform calculations on complex
datasets. The following sample workflows show some of the functionality
of these objects.
• “Example: Fitting a Single Distribution to a Single Dataset” on page 5-91
5-90
Using Probability Distribution Objects
• “Example: Fitting Distributions to Grouped Data Within a Single Dataset”
on page 5-91
• “Example: Fitting Multiple Distribution Types to a Single Dataset” on
page 5-95
Example: Fitting a Single Distribution to a Single Dataset
Fit a single Normal distribution to a dataset using fitdist:
load carsmall
NormDist = fitdist(MPG,'normal')
NormDist =
normal distribution
mu = 23.7181
sigma = 8.03573
The output MATLAB returns is a ProbDistUnivParam object with a DistName
property of 'normal distribution'. The ParamNames property contains the
strings mu and sigma, while the Params property contains the parameter
values.
Example: Fitting Distributions to Grouped Data Within a Single
Dataset
Often, datasets are collections of data you can group in different ways. Using
fitdist and the data from carsmall.mat, group the MPG data by country of
origin, then fit a Weibull distribution each group:
load carsmall
[WeiByOrig, Country] = fitdist(MPG,'weibull','by',Origin)
Warning: Error while fitting group 'Italy':
Not enough data in X to fit this distribution.
> In fitdist at 171
WeiByOrig =
5-91
5 Probability Distributions
Columns 1 through 4
[1x1 ProbDistUnivParam] [1x1 ProbDistUnivParam] ...
[1x1 ProbDistUnivParam] [1x1 ProbDistUnivParam]
Columns 5 through 6
[1x1 ProbDistUnivParam] []
Country =
'USA'
'France'
'Japan'
'Germany'
'Sweden'
'Italy'
A warning appears informing you that, since the data only represents one
Italian car, fitdist cannot fit a Weibull distribution to that group. Each
one of the five other groups now has a distribution object associated with it,
represented in the cell array wd. Each object contains properties that hold
information about the data, the distribution, and the parameters. For more
information on what properties exist and what information they contain, see
ProbDistUnivParam or ProbDistUnivKernel.
Now access two of the objects and their properties:
% Get USA fit
distusa = WeiByOrig{1};
% Use the InputData property of ProbDistUnivParam objects to see
% the actual data used to fit the distribution:
dusa = distusa.InputData.data;
% Get Japan fit and data
distjapan = WeiByOrig{3};
djapan = distjapan.InputData.data;
5-92
Using Probability Distribution Objects
Now you can easily compare PDFs using the pdf method of the
ProbDistUnivParam class:
time = linspace(0,45);
pdfjapan = pdf(distjapan,time);
pdfusa = pdf(distusa,time);
hold on
plot(time,[pdfjapan;pdfusa])
l = legend('Japan','USA')
set(l,'Location','Best')
xlabel('MPG')
ylabel('Probability Density')
5-93
5 Probability Distributions
You could then further group the data and compare, for example, MPG by
year for American cars:
load carsmall
[WeiByYearOrig, Names] = fitdist(MPG,'weibull','by',...
{Origin Model_Year});
USA70 = WeiByYearOrig{1};
USA76 = WeiByYearOrig{2};
USA82 = WeiByYearOrig{3};
time = linspace(0,45);
pdf70 = pdf(USA70,time);
pdf76 = pdf(USA76,time);
pdf82 = pdf(USA82,time);
line(t,[pdf70;pdf76;pdf82])
l = legend('1970','1976','1982')
set(l,'Location','Best')
title('USA Car MPG by Year')
xlabel('MPG')
ylabel('Probability Density')
5-94
Using Probability Distribution Objects
Example: Fitting Multiple Distribution Types to a Single Dataset
Distribution objects make it easy to fit multiple distributions to the same
dataset, while minimizing workspace clutter. For example, use fitdist to
group the MPG data by country of origin, then fit Weibull, Normal, Logistic,
and nonparametric distributions for each group:
load carsmall;
[WeiByOrig, Country] = fitdist(MPG,'weibull','by',Origin);
[NormByOrig, Country] = fitdist(MPG,'normal','by',Origin);
[LogByOrig, Country] = fitdist(MPG,'logistic','by',Origin);
[KerByOrig, Country] = fitdist(MPG,'kernel','by',Origin);
5-95
5 Probability Distributions
Extract the fits for American cars and compare the fits visually against a
histogram of the original data:
WeiUSA = WeiByOrig{1};
NormUSA = NormByOrig{1};
LogUSA = LogByOrig{1};
KerUSA = KerByOrig{1};
% Since all three distributions use the same set of data,
% you can extract the data from any of them:
data = WeiUSA.InputData.data;
% Create a histogram of the data:
[n,y] = hist(data,10);
b = bar(y,n,'hist');
set(b,'FaceColor',[1,0.8,0])
% Scale the density by the histogram area, for easier display:
area = sum(n) * (y(2)-y(1));
time = linspace(0,45);
pdfWei = pdf(WeiUSA,time);
pdfNorm = pdf(NormUSA,time);
pdfLog = pdf(LogUSA,time);
pdfKer = pdf(KerUSA,time);
allpdf = [pdfWei;pdfNorm;pdfLog;pdfKer];
line(t,area * allpdf)
l = legend('Data','Weibull','Normal','Logistic','Kernel')
set(l,'Location','Best')
title('USA Car')
xlabel('MPG')
5-96
Using Probability Distribution Objects
You can see that only the nonparametric kernel distribution, KerUSA, comes
close to revealing the two modes in the data.
Capturing Results Using Distribution Objects
Distribution objects allow you to share both your dataset and your analysis
results simply by saving the information to a .mat file.
Example: Saving and Sharing Distribution Fit Data
Using the premise from the previous set of examples, group the MPG data
in carsmall.mat by country of origin and fit four different distributions to
each of the six sets of data:
5-97
5 Probability Distributions
load carsmall;
[WeiByOrig, Country] = fitdist(MPG,'weibull','by',Origin);
[NormByOrig, Country] = fitdist(MPG,'normal','by',Origin);
[LogByOrig, Country] = fitdist(MPG,'logistic','by',Origin);
[KerByOrig, Country] = fitdist(MPG,'kernel','by',Origin);
Combine all four fits and the country labels into a single cell array, including
“headers” to indicate which distributions correspond to which objects. Then,
save the array to a .mat file:
AllFits = cell(['Country' Country'; 'Weibull' WeiByOrig;...
'Normal' NormByOrig; 'Logistic' LogByOrig; 'Kernel',...
KerByOrig]);
save('CarSmallFits.mat','AllFits');
To show that the data is both safely saved and easily restored, clear your
workspace of relevant variables. This command clears only those variables
associated with this example:
clear('Weight','Acceleration','AllFits','Country',...
'Cylinders','Displacement','Horsepower','KerByOrig',...
'LogByOrig','MPG','Model','Model_Year','NormByOrig',...
'Origin','WeiByOrig')
Now, load the data:
load CarSmallFits
AllFits
You can now access the distributions objects as in the previous examples.
5-98
Probability Distributions Used for Multivariate Modeling
Probability Distributions Used for Multivariate Modeling
In this section...
“Gaussian Mixture Models” on page 5-99
“Copulas” on page 5-107
Gaussian Mixture Models
• “Creating Gaussian Mixture Models” on page 5-99
• “Simulating Gaussian Mixtures” on page 5-105
Gaussian mixture models are formed by combining multivariate normal
density components. For information on individual multivariate normal
densities, see “Multivariate Normal Distribution” on page B-58 and related
distribution functions listed under “Multivariate Distributions” on page 5-8.
In Statistics Toolbox software, use the gmdistribution class to fit data
using an expectation maximization (EM) algorithm, which assigns posterior
probabilities to each component density with respect to each observation. The
fitting method uses an iterative algorithm that converges to a local optimum.
Clustering using Gaussian mixture models is sometimes considered a soft
clustering method. The posterior probabilities for each point indicate that
each data point has some probability of belonging to each cluster.
For more information on clustering with Gaussian mixture models, see
“Gaussian Mixture Models” on page 11-28. This section describes their
creation.
Creating Gaussian Mixture Models
• “Specifying a Model” on page 5-100
• “Fitting a Model to Data” on page 5-102
5-99
5 Probability Distributions
Specifying a Model. Use the gmdistribution constructor to create
Gaussian mixture models with specified means, covariances, and mixture
proportions. The following creates an object of the gmdistribution class
defining a two-component mixture of bivariate Gaussian distributions:
MU = [1 2;-3 -5]; % Means
SIGMA = cat(3,[2 0;0 .5],[1 0;0 1]); % Covariances
p = ones(1,2)/2; % Mixing proportions
obj = gmdistribution(MU,SIGMA,p);
Display properties of the object with the MATLAB function fieldnames:
properties = fieldnames(obj)
properties =
'NDimensions'
'DistName'
'NComponents'
'PComponents'
'mu'
'Sigma'
'NlogL'
'AIC'
'BIC'
'Converged'
'Iters'
'SharedCov'
'CovType'
'RegV'
The gmdistribution reference page describes these properties. To access the
value of a property, use dot indexing:
dimension = obj.NDimensions
dimension =
2
name = obj.DistName
name =
gaussian mixture distribution
5-100
Probability Distributions Used for Multivariate Modeling
Use the methods pdf and cdf to compute values and visualize the object:
ezsurf(@(x,y)pdf(obj,[x y]),[-10 10],[-10 10])
ezsurf(@(x,y)cdf(obj,[x y]),[-10 10],[-10 10])
5-101
5 Probability Distributions
Fitting a Model to Data. You can also create Gaussian mixture models
by fitting a parametric model with a specified number of components to
data. The fit method of the gmdistribution class uses the syntax obj =
gmdistribution.fit(X,k), where X is a data matrix and k is the specified
number of components. Choosing a suitable number of components k is
essential for creating a useful model of the data—too few components fails to
model the data accurately; too many components leads to an over-fit model
with singular covariance matrices.
The following example illustrates this approach.
First, create some data from a mixture of two bivariate Gaussian distributions
using the mvnrnd function:
MU1 = [1 2];
SIGMA1 = [2 0; 0 .5];
5-102
Probability Distributions Used for Multivariate Modeling
MU2 = [-3 -5];
SIGMA2 = [1 0; 0 1];
X = [mvnrnd(MU1,SIGMA1,1000);
mvnrnd(MU2,SIGMA2,1000)];
scatter(X(:,1),X(:,2),10,'.')
Next, fit a two-component Gaussian mixture model:
options = statset('Display','final');
obj = gmdistribution.fit(X,2,'Options',options);
hold on
h = ezcontour(@(x,y)pdf(obj,[x y]),[-8 6],[-8 6]);
hold off
5-103
5 Probability Distributions
Among the properties of the fit are the parameter estimates:
ComponentMeans = obj.mu
ComponentMeans =
0.9391 2.0322
-2.9823 -4.9737
ComponentCovariances = obj.Sigma
ComponentCovariances(:,:,1) =
1.7786 -0.0528
-0.0528 0.5312
ComponentCovariances(:,:,2) =
1.0491 -0.0150
5-104
Probability Distributions Used for Multivariate Modeling
-0.0150 0.9816
MixtureProportions = obj.PComponents
MixtureProportions =
0.5000 0.5000
The two-component model minimizes the Akaike information:
AIC = zeros(1,4);
obj = cell(1,4);
for k = 1:4
obj{k} = gmdistribution.fit(X,k);
AIC(k)= obj{k}.AIC;
end
[minAIC,numComponents] = min(AIC);
numComponents
numComponents =
2
model = obj{2}
model =
Gaussian mixture distribution
with 2 components in 2 dimensions
Component 1:
Mixing proportion: 0.500000
Mean: 0.9391 2.0322
Component 2:
Mixing proportion: 0.500000
Mean: -2.9823 -4.9737
Both the Akaike and Bayes information are negative log-likelihoods for the
data with penalty terms for the number of estimated parameters. You can use
them to determine an appropriate number of components for a model when
the number of components is unspecified.
Simulating Gaussian Mixtures
Use the method random of the gmdistribution class to generate random data
from a Gaussian mixture model created with gmdistribution or fit.
5-105
5 Probability Distributions
For example, the following specifies a gmdistribution object consisting of a
two-component mixture of bivariate Gaussian distributions:
MU = [1 2;-3 -5];
SIGMA = cat(3,[2 0;0 .5],[1 0;0 1]);
p = ones(1,2)/2;
obj = gmdistribution(MU,SIGMA,p);
ezcontour(@(x,y)pdf(obj,[x y]),[-10 10],[-10 10])
hold on
Use random (gmdistribution) to generate 1000 random values:
Y = random(obj,1000);
scatter(Y(:,1),Y(:,2),10,'.')
5-106
Probability Distributions Used for Multivariate Modeling
Copulas
• “Determining Dependence Between Simulation Inputs” on page 5-108
• “Constructing Dependent Bivariate Distributions” on page 5-112
• “Using Rank Correlation Coefficients” on page 5-116
• “Using Bivariate Copulas” on page 5-119
• “Higher Dimension Copulas” on page 5-126
• “Archimedean Copulas” on page 5-128
• “Simulating Dependent Multivariate Data Using Copulas” on page 5-130
• “Example: Fitting Copulas to Data” on page 5-135
5-107
5 Probability Distributions
Copulas are functions that describe dependencies among variables, and
provide a way to create distributions that model correlated multivariate data.
Using a copula, you can construct a multivariate distribution by specifying
marginal univariate distributions, and then choose a copula to provide a
correlation structure between variables. Bivariate distributions, as well as
distributions in higher dimensions, are possible.
Determining Dependence Between Simulation Inputs
One of the design decisions for a Monte Carlo simulation is a choice of
probability distributions for the random inputs. Selecting a distribution
for each individual variable is often straightforward, but deciding what
dependencies should exist between the inputs may not be. Ideally, input
data to a simulation should reflect what you know about dependence among
the real quantities you are modeling. However, there may be little or no
information on which to base any dependence in the simulation. In such cases,
it is useful to experiment with different possibilities in order to determine
the model’s sensitivity.
It can be difficult to generate random inputs with dependence when they have
distributions that are not from a standard multivariate distribution. Further,
some of the standard multivariate distributions can model only limited types
of dependence. It is always possible to make the inputs independent, and
while that is a simple choice, it is not always sensible and can lead to the
wrong conclusions.
For example, a Monte-Carlo simulation of financial risk could have two
random inputs that represent different sources of insurance losses. You could
model these inputs as lognormal random variables. A reasonable question
to ask is how dependence between these two inputs affects the results of the
simulation. Indeed, you might know from real data that the same random
conditions affect both sources; ignoring that in the simulation could lead to
the wrong conclusions.
Example: Generate and Exponentiate Normal Random Variables.
The lognrnd function simulates independent lognormal random variables. In
the following example, the mvnrnd function generates n pairs of independent
normal random variables, and then exponentiates them. Notice that the
covariance matrix used here is diagonal:
5-108
Probability Distributions Used for Multivariate Modeling
n = 1000;
sigma = .5;
SigmaInd = sigma.^2 .* [1 0; 0 1]
SigmaInd =
0.25 0
0 0.25
ZInd = mvnrnd([0 0],SigmaInd,n);
XInd = exp(ZInd);
plot(XInd(:,1),XInd(:,2),'.')
axis([0 5 0 5])
axis equal
xlabel('X1')
ylabel('X2')
5-109
5 Probability Distributions
Dependent bivariate lognormal random variables are also easy to generate
using a covariance matrix with nonzero off-diagonal terms:
rho = .7;
SigmaDep = sigma.^2 .* [1 rho; rho 1]
SigmaDep =
0.25 0.175
0.175 0.25
ZDep = mvnrnd([0 0],SigmaDep,n);
XDep = exp(ZDep);
A second scatter plot demonstrates the difference between these two bivariate
distributions:
plot(XDep(:,1),XDep(:,2),'.')
axis([0 5 0 5])
axis equal
xlabel('X1')
ylabel('X2')
5-110
Probability Distributions Used for Multivariate Modeling
It is clear that there is a tendency in the second data set for large values of
X1 to be associated with large values of X2, and similarly for small values.
The correlation parameter, ρ, of the underlying bivariate normal determines
this dependence. The conclusions drawn from the simulation could well
depend on whether you generate X1 and X2 with dependence. The bivariate
lognormal distribution is a simple solution in this case; it easily generalizes
to higher dimensions in cases where the marginal distributions are different
lognormals.
Other multivariate distributions also exist. For example, the multivariate
t and the Dirichlet distributions simulate dependent t and beta random
variables, respectively. But the list of simple multivariate distributions is not
long, and they only apply in cases where the marginals are all in the same
family (or even the exact same distributions). This can be a serious limitation
in many situations.
5-111
5 Probability Distributions
Constructing Dependent Bivariate Distributions
Although the construction discussed in the previous section creates a
bivariate lognormal that is simple, it serves to illustrate a method that is
more generally applicable.
1 Generate pairs of values from a bivariate normal distribution. There is
statistical dependence between these two variables, and each has a normal
marginal distribution.
2 Apply a transformation (the exponential function) separately to each
variable, changing the marginal distributions into lognormals. The
transformed variables still have a statistical dependence.
If a suitable transformation can be found, this method can be generalized to
create dependent bivariate random vectors with other marginal distributions.
In fact, a general method of constructing such a transformation does exist,
although it is not as simple as exponentiation alone.
By definition, applying the normal cumulative distribution function (cdf),
denoted here by Φ, to a standard normal random variable results in a random
variable that is uniform on the interval [0,1]. To see this, if Z has a standard
normal distribution, then the cdf of U = Φ(Z) is
Pr{U ≤ u} = Pr{Φ( Z) ≤ u} = Pr( Z ≤ Φ −1 (u)} = u
and that is the cdf of a Unif(0,1) random variable. Histograms of some
simulated normal and transformed values demonstrate that fact:
n = 1000;
z = normrnd(0,1,n,1);
hist(z,-3.75:.5:3.75)
xlim([-4 4])
title('1000 Simulated N(0,1) Random Values')
xlabel('Z')
ylabel('Frequency')
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
5-112
Probability Distributions Used for Multivariate Modeling
u = normcdf(z);
hist(u,.05:.1:.95)
title('1000 Simulated N(0,1) Values Transformed to Unif(0,1)')
xlabel('U')
ylabel('Frequency')
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
5-113
5 Probability Distributions
Borrowing from the theory of univariate random number generation, applying
the inverse cdf of any distribution, F, to a Unif(0,1) random variable results in
a random variable whose distribution is exactly F (see “Inversion Methods”
on page 6-7). The proof is essentially the opposite of the preceding proof for
the forward case. Another histogram illustrates the transformation to a
gamma distribution:
x = gaminv(u,2,1);
hist(x,.25:.5:9.75)
title('1000 Simulated N(0,1) Values Transformed to Gamma(2,1)')
xlabel('X')
ylabel('Frequency')
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
5-114
Probability Distributions Used for Multivariate Modeling
You can apply this two-step transformation to each variable of a standard
bivariate normal, creating dependent random variables with arbitrary
marginal distributions. Because the transformation works on each component
separately, the two resulting random variables need not even have the same
marginal distributions. The transformation is defined as:
⎛ ⎡1 ⎤⎞
Z = [ Z1 , Z2 ] ∼ N ⎜ [0, 0] , ⎢ ⎟
⎝ ⎣ 1 ⎥⎦ ⎠
U = ⎡⎣Φ ( Z1 ) , Φ ( Z2 ) ⎤⎦
X = ⎡⎣G1 (U1 ) , G2 (U2 ) ⎤⎦
5-115
5 Probability Distributions
where G1 and G2 are inverse cdfs of two possibly different distributions. For
example, the following generates random vectors from a bivariate distribution
with t5 and Gamma(2,1) marginals:
n = 1000; rho = .7;
Z = mvnrnd([0 0],[1 rho; rho 1],n);
U = normcdf(Z);
X = [gaminv(U(:,1),2,1) tinv(U(:,2),5)];
scatterhist(X(:,1),X(:,2))
This plot has histograms alongside a scatter plot to show both the marginal
distributions, and the dependence.
Using Rank Correlation Coefficients
The correlation parameter, ρ, of the underlying bivariate normal determines
the dependence between X1 and X2 in this construction. However, the linear
correlation of X1 and X2 is not ρ. For example, in the original lognormal case,
a closed form for that correlation is:
5-116
Probability Distributions Used for Multivariate Modeling
e − 1
2
cor ( X 1, X 2) =
e − 1
2
which is strictly less than ρ, unless ρ is exactly 1. In more general cases such
as the Gamma/t construction, the linear correlation between X1 and X2 is
difficult or impossible to express in terms of ρ, but simulations show that the
same effect happens.
That is because the linear correlation coefficient expresses the linear
dependence between random variables, and when nonlinear transformations
are applied to those random variables, linear correlation is not preserved.
Instead, a rank correlation coefficient, such as Kendall’s τ or Spearman’s ρ,
is more appropriate.
Roughly speaking, these rank correlations measure the degree to which
large or small values of one random variable associate with large or small
values of another. However, unlike the linear correlation coefficient, they
measure the association only in terms of ranks. As a consequence, the rank
correlation is preserved under any monotonic transformation. In particular,
the transformation method just described preserves the rank correlation.
Therefore, knowing the rank correlation of the bivariate normal Z exactly
determines the rank correlation of the final transformed random variables,
X. While the linear correlation coefficient, ρ, is still needed to parameterize
the underlying bivariate normal, Kendall’s τ or Spearman’s ρ are more useful
in describing the dependence between random variables, because they are
invariant to the choice of marginal distribution.
For the bivariate normal, there is a simple one-to-one mapping between
Kendall’s τ or Spearman’s ρ, and the linear correlation coefficient ρ:
2 ⎛ ⎞
= arcsin ( )
or = sin ⎜ ⎟
⎝ 2⎠
6 ⎛⎞ ⎛ ⎞
s = arcsin ⎜ ⎟ or = 2sin ⎜ s ⎟
⎝2⎠ ⎝ 6⎠
The following plot shows the relationship:
rho = -1:.01:1;
5-117
5 Probability Distributions
tau = 2.*asin(rho)./pi;
rho_s = 6.*asin(rho./2)./pi;
plot(rho,tau,'b-','LineWidth',2)
hold on
plot(rho,rho_s,'g-','LineWidth',2)
plot([-1 1],[-1 1],'k:','LineWidth',2)
axis([-1 1 -1 1])
xlabel('rho')
ylabel('Rank correlation coefficient')
legend('Kendall''s {\it\tau}', ...
'Spearman''s {\it\rho_s}', ...
'location','NW')
5-118
Probability Distributions Used for Multivariate Modeling
Thus, it is easy to create the desired rank correlation between X1 and X2,
regardless of their marginal distributions, by choosing the correct ρ parameter
value for the linear correlation between Z1 and Z2.
For the multivariate normal distribution, Spearman’s rank correlation is
almost identical to the linear correlation. However, this is not true once you
transform to the final random variables.
Using Bivariate Copulas
The first step of the construction described in the previous section defines
what is known as a bivariate Gaussian copula. A copula is a multivariate
probability distribution, where each random variable has a uniform marginal
distribution on the unit interval [0,1]. These variables may be completely
independent, deterministically related (e.g., U2 = U1), or anything in between.
Because of the possibility for dependence among variables, you can use a
copula to construct a new multivariate distribution for dependent variables.
By transforming each of the variables in the copula separately using the
inversion method, possibly using different cdfs, the resulting distribution can
have arbitrary marginal distributions. Such multivariate distributions are
often useful in simulations, when you know that the different random inputs
are not independent of each other.
Statistics Toolbox functions compute:
• Probability density functions (copulapdf) and the cumulative distribution
functions (copulacdf) for Gaussian copulas
• Rank correlations from linear correlations (copulastat) and vice versa
(copulaparam)
• Random vectors (copularnd)
• Parameters for copulas fit to data (copulafit)
For example, use the copularnd function to create scatter plots of random
values from a bivariate Gaussian copula for various levels of ρ, to illustrate the
range of different dependence structures. The family of bivariate Gaussian
copulas is parameterized by the linear correlation matrix:
5-119
5 Probability Distributions
⎛1 ⎞
Ρ=⎜
⎝ 1 ⎟⎠
U1 and U2 approach linear dependence as ρ approaches ±1, and approach
complete independence as ρ approaches zero:
n = 500;
U = copularnd('Gaussian',[1 .8; .8 1],n);
subplot(2,2,1)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = 0.8')
xlabel('U1')
ylabel('U2')
U = copularnd('Gaussian',[1 .1; .1 1],n);
subplot(2,2,2)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = 0.1')
xlabel('U1')
ylabel('U2')
U = copularnd('Gaussian',[1 -.1; -.1 1],n);
subplot(2,2,3)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = -0.1')
xlabel('U1')
ylabel('U2')
U = copularnd('Gaussian',[1 -.8; -.8 1],n);
subplot(2,2,4)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = -0.8')
xlabel('U1')
ylabel('U2')
5-120
Probability Distributions Used for Multivariate Modeling
The dependence between U1 and U2 is completely separate from the marginal
distributions of X1 = G(U1) and X2 = G(U2). X1 and X2 can be given any
marginal distributions, and still have the same rank correlation. This is
one of the main appeals of copulas—they allow this separate specification
of dependence and marginal distribution. You can also compute the pdf
(copulapdf) and the cdf (copulacdf) for a copula. For example, these plots
show the pdf and cdf for ρ = .8:
u1 = linspace(1e-3,1-1e-3,50);
u2 = linspace(1e-3,1-1e-3,50);
[U1,U2] = meshgrid(u1,u2);
Rho = [1 .8; .8 1];
f = copulapdf('t',[U1(:) U2(:)],Rho,5);
f = reshape(f,size(U1));
surf(u1,u2,log(f),'FaceColor','interp','EdgeColor','none')
5-121
5 Probability Distributions
view([-15,20])
xlabel('U1')
ylabel('U2')
zlabel('Probability Density')
u1 = linspace(1e-3,1-1e-3,50);
u2 = linspace(1e-3,1-1e-3,50);
[U1,U2] = meshgrid(u1,u2);
F = copulacdf('t',[U1(:) U2(:)],Rho,5);
F = reshape(F,size(U1));
surf(u1,u2,F,'FaceColor','interp','EdgeColor','none')
view([-15,20])
xlabel('U1')
ylabel('U2')
zlabel('Cumulative Probability')
5-122
Probability Distributions Used for Multivariate Modeling
A different family of copulas can be constructed by starting from a bivariate t
distribution and transforming using the corresponding t cdf. The bivariate t
distribution is parameterized with P, the linear correlation matrix, and ν, the
degrees of freedom. Thus, for example, you can speak of a t1 or a t5 copula,
based on the multivariate t with one and five degrees of freedom, respectively.
Just as for Gaussian copulas, Statistics Toolbox functions for t copulas
compute:
• Probability density functions (copulapdf) and the cumulative distribution
functions (copulacdf) for Gaussian copulas
• Rank correlations from linear correlations (copulastat) and vice versa
(copulaparam)
• Random vectors (copularnd)
• Parameters for copulas fit to data (copulafit)
5-123
5 Probability Distributions
For example, use the copularnd function to create scatter plots of random
values from a bivariate t1 copula for various levels of ρ, to illustrate the range
of different dependence structures:
n = 500;
nu = 1;
U = copularnd('t',[1 .8; .8 1],nu,n);
subplot(2,2,1)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = 0.8')
xlabel('U1')
ylabel('U2')
U = copularnd('t',[1 .1; .1 1],nu,n);
subplot(2,2,2)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = 0.1')
xlabel('U1')
ylabel('U2')
U = copularnd('t',[1 -.1; -.1 1],nu,n);
subplot(2,2,3)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = -0.1')
xlabel('U1')
ylabel('U2')
U = copularnd('t',[1 -.8; -.8 1],nu, n);
subplot(2,2,4)
plot(U(:,1),U(:,2),'.')
title('{\it\rho} = -0.8')
xlabel('U1')
ylabel('U2')
5-124
Probability Distributions Used for Multivariate Modeling
A t copula has uniform marginal distributions for U1 and U2, just as a
Gaussian copula does. The rank correlation τ or ρs between components in a t
copula is also the same function of ρ as for a Gaussian. However, as these plots
demonstrate, a t1 copula differs quite a bit from a Gaussian copula, even when
their components have the same rank correlation. The difference is in their
dependence structure. Not surprisingly, as the degrees of freedom parameter
ν is made larger, a tν copula approaches the corresponding Gaussian copula.
As with a Gaussian copula, any marginal distributions can be imposed over
a t copula. For example, using a t copula with 1 degree of freedom, you can
again generate random vectors from a bivariate distribution with Gamma(2,1)
and t5 marginals using copularnd:
n = 1000;
rho = .7;
nu = 1;
5-125
5 Probability Distributions
U = copularnd('t',[1 rho; rho 1],nu,n);
X = [gaminv(U(:,1),2,1) tinv(U(:,2),5)];
scatterhist(X(:,1),X(:,2))
Compared to the bivariate Gamma/t distribution constructed earlier, which
was based on a Gaussian copula, the distribution constructed here, based on a
t1 copula, has the same marginal distributions and the same rank correlation
between variables but a very different dependence structure. This illustrates
the fact that multivariate distributions are not uniquely defined by their
marginal distributions, or by their correlations. The choice of a particular
copula in an application may be based on actual observed data, or different
copulas may be used as a way of determining the sensitivity of simulation
results to the input distribution.
Higher Dimension Copulas
The Gaussian and t copulas are known as elliptical copulas. It is easy to
generalize elliptical copulas to a higher number of dimensions. For example,
simulate data from a trivariate distribution with Gamma(2,1), Beta(2,2), and
t5 marginals using a Gaussian copula and copularnd, as follows:
5-126
Probability Distributions Used for Multivariate Modeling
n = 1000;
Rho = [1 .4 .2; .4 1 -.8; .2 -.8 1];
U = copularnd('Gaussian',Rho,n);
X = [gaminv(U(:,1),2,1) betainv(U(:,2),2,2) tinv(U(:,3),5)];
subplot(1,1,1)
plot3(X(:,1),X(:,2),X(:,3),'.')
grid on
view([-55, 15])
xlabel('X1')
ylabel('X2')
zlabel('X3')
Notice that the relationship between the linear correlation parameter ρ and,
for example, Kendall’s τ, holds for each entry in the correlation matrix P
used here. You can verify that the sample rank correlations of the data are
approximately equal to the theoretical values:
tauTheoretical = 2.*asin(Rho)./pi
tauTheoretical =
5-127
5 Probability Distributions
1 0.26198 0.12819
0.26198 1 -0.59033
0.12819 -0.59033 1
tauSample = corr(X,'type','Kendall')
tauSample =
1 0.27254 0.12701
0.27254 1 -0.58182
0.12701 -0.58182 1
Archimedean Copulas
Statistics Toolbox functions are available for three bivariate Archimedean
copula families:
• Clayton copulas
• Frank copulas
• Gumbel copulas
These are one-parameter families that are defined directly in terms of their
cdfs, rather than being defined constructively using a standard multivariate
distribution.
To compare these three Archimedean copulas to the Gaussian and t bivariate
copulas, first use the copulastat function to find the rank correlation for
a Gaussian or t copula with linear correlation parameter of 0.8, and then
use the copulaparam function to find the Clayton copula parameter that
corresponds to that rank correlation:
tau = copulastat('Gaussian',.8 ,'type','kendall')
tau =
0.59033
alpha = copulaparam('Clayton',tau,'type','kendall')
alpha =
2.882
Finally, plot a random sample from the Clayton copula with copularnd.
Repeat the same procedure for the Frank and Gumbel copulas:
5-128
Probability Distributions Used for Multivariate Modeling
n = 500;
U = copularnd('Clayton',alpha,n);
subplot(3,1,1)
plot(U(:,1),U(:,2),'.');
title(['Clayton Copula, {\it\alpha} = ',sprintf('%0.2f',alpha)])
xlabel('U1')
ylabel('U2')
alpha = copulaparam('Frank',tau,'type','kendall');
U = copularnd('Frank',alpha,n);
subplot(3,1,2)
plot(U(:,1),U(:,2),'.')
title(['Frank Copula, {\it\alpha} = ',sprintf('%0.2f',alpha)])
xlabel('U1')
ylabel('U2')
alpha = copulaparam('Gumbel',tau,'type','kendall');
U = copularnd('Gumbel',alpha,n);
subplot(3,1,3)
plot(U(:,1),U(:,2),'.')
title(['Gumbel Copula, {\it\alpha} = ',sprintf('%0.2f',alpha)])
xlabel('U1')
ylabel('U2')
5-129
5 Probability Distributions
Simulating Dependent Multivariate Data Using Copulas
To simulate dependent multivariate data using a copula, you must specify
each of the following:
5-130
Probability Distributions Used for Multivariate Modeling
• The copula family (and any shape parameters)
• The rank correlations among variables
• Marginal distributions for each variable
Suppose you have return data for two stocks and want to run a Monte Carlo
simulation with inputs that follow the same distributions as the data:
load stockreturns
nobs = size(stocks,1);
subplot(2,1,1)
hist(stocks(:,1),10)
xlim([-3.5 3.5])
xlabel('X1')
ylabel('Frequency')
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
subplot(2,1,2)
hist(stocks(:,2),10)
xlim([-3.5 3.5])
xlabel('X2')
ylabel('Frequency')
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
5-131
5 Probability Distributions
You could fit a parametric model separately to each dataset, and use those
estimates as the marginal distributions. However, a parametric model may
not be sufficiently flexible. Instead, you can use a nonparametric model
to transform to the marginal distributions. All that is needed is a way to
compute the inverse cdf for the nonparametric model.
The simplest nonparametric model is the empirical cdf, as computed by the
ecdf function. For a discrete marginal distribution, this is appropriate.
However, for a continuous distribution, use a model that is smoother than
the step function computed by ecdf. One way to do that is to estimate
the empirical cdf and interpolate between the midpoints of the steps with
a piecewise linear function. Another way is to use kernel smoothing with
ksdensity. For example, compare the empirical cdf to a kernel smoothed cdf
estimate for the first variable:
5-132
Probability Distributions Used for Multivariate Modeling
[Fi,xi] = ecdf(stocks(:,1));
stairs(xi,Fi,'b','LineWidth',2)
hold on
Fi_sm = ksdensity(stocks(:,1),xi,'function','cdf','width',.15);
plot(xi,Fi_sm,'r-','LineWidth',1.5)
xlabel('X1')
ylabel('Cumulative Probability')
legend('Empirical','Smoothed','Location','NW')
grid on
For the simulation, experiment with different copulas and correlations.
Here, you will use a bivariate t copula with a fairly small degrees of freedom
5-133
5 Probability Distributions
parameter. For the correlation parameter, you can compute the rank
correlation of the data, and then find the corresponding linear correlation
parameter for the t copula using copulaparam:
nu = 5;
tau = corr(stocks(:,1),stocks(:,2),'type','kendall')
tau =
0.51798
rho = copulaparam('t', tau, nu, 'type','kendall')
rho =
0.72679
Next, use copularnd to generate random values from the t copula and
transform using the nonparametric inverse cdfs. The ksdensity function
allows you to make a kernel estimate of distribution and evaluate the inverse
cdf at the copula points all in one step:
n = 1000;
U = copularnd('t',[1 rho; rho 1],nu,n);
X1 = ksdensity(stocks(:,1),U(:,1),...
'function','icdf','width',.15);
X2 = ksdensity(stocks(:,2),U(:,2),...
'function','icdf','width',.15);
Alternatively, when you have a large amount of data or need to simulate more
than one set of values, it may be more efficient to compute the inverse cdf
over a grid of values in the interval (0,1) and use interpolation to evaluate it
at the copula points:
p = linspace(0.00001,0.99999,1000);
G1 = ksdensity(stocks(:,1),p,'function','icdf','width',0.15);
X1 = interp1(p,G1,U(:,1),'spline');
G2 = ksdensity(stocks(:,2),p,'function','icdf','width',0.15);
X2 = interp1(p,G2,U(:,2),'spline');
scatterhist(X1,X2)
5-134
Probability Distributions Used for Multivariate Modeling
The marginal histograms of the simulated data are a smoothed version of the
histograms for the original data. The amount of smoothing is controlled by
the bandwidth input to ksdensity.
Example: Fitting Copulas to Data
The copulafit function is used to calibrate copulas with data. To generate
data Xsim with a distribution “just like” (in terms of marginal distributions
and correlations) the distribution of data in the matrix X:
1 Fit marginal distributions to the columns of X.
2 Use appropriate cdf functions to transform X to U, so that U has values
between 0 and 1.
3 Use copulafit to fit a copula to U.
4 Generate new data Usim from the copula.
5-135
5 Probability Distributions
5 Use appropriate inverse cdf functions to transform Usim to Xsim.
The following example illustrates the procedure.
Load and plot simulated stock return data:
load stockreturns
x = stocks(:,1);
y = stocks(:,2);
scatterhist(x,y)
Transform the data to the copula scale (unit square) using a kernel estimator
of the cumulative distribution function:
5-136
Probability Distributions Used for Multivariate Modeling
u = ksdensity(x,x,'function','cdf');
v = ksdensity(y,y,'function','cdf');
scatterhist(u,v)
xlabel('u')
ylabel('v')
Fit a t copula:
[Rho,nu] = copulafit('t',[u v],'Method','ApproximateML')
Rho =
1.0000 0.7220
0.7220 1.0000
nu =
3.2017e+006
5-137
5 Probability Distributions
Generate a random sample from the t copula:
r = copularnd('t',Rho,nu,1000);
u1 = r(:,1);
v1 = r(:,2);
scatterhist(u1,v1)
xlabel('u')
ylabel('v')
set(get(gca,'children'),'marker','.')
Transform the random sample back to the original scale of the data:
x1 = ksdensity(u,u1,'function','icdf');
5-138
Probability Distributions Used for Multivariate Modeling
y1 = ksdensity(v,v1,'function','icdf');
scatterhist(x1,y1)
set(get(gca,'children'),'marker','.')
As the example illustrates, copulas integrate naturally with other distribution
fitting functions.
5-139
5 Probability Distributions
5-140
6
Random Number
Generation
• “Generating Random Data” on page 6-2
• “Random Number Generation Functions” on page 6-3
• “Common Generation Methods” on page 6-5
• “Parallel Computing Support for Random Number Generation” on page
6-13
• “Representing Sampling Distributions Using Markov Chain Samplers”
on page 6-15
• “Generating Quasi-Random Numbers” on page 6-17
• “Generating Data Using Flexible Families of Distributions” on page 6-27
6 Random Number Generation
Generating Random Data
Pseudorandom numbers are generated by deterministic algorithms. They are
"random" in the sense that, on average, they pass statistical tests regarding
their distribution and correlation. They differ from true random numbers in
that they are generated by an algorithm, rather than a truly random process.
Random number generators (RNGs) like those in MATLAB are algorithms for
generating pseudorandom numbers with a specified distribution.
For more information on random number generators for supported
distributions, see “Random Number Generators” on page 5-80.
For more information on the GUI for generating random numbers from
supported distributions, see “Visually Exploring Random Number Generation”
on page 5-49.
6-2
Random Number Generation Functions
Random Number Generation Functions
The following table lists the supported distributions and their respective
random number generation functions. For more information on other
functions for each distribution, see “Supported Distributions” on page 5-3.
For more information on random number generators, see “Random Number
Generators” on page 5-80.
Distribution Random Number Generation Function
Beta betarnd, random, randtool
Binomial binornd, random, randtool
Chi-square chi2rnd, random, randtool
Clayton copula copularnd
Exponential exprnd, random, randtool
Extreme value evrnd, random, randtool
F frnd, random, randtool
Frank copula copularnd
Gamma gamrnd, randg, random, randtool
Gaussian copula copularnd
Gaussian mixture random
Generalized extreme gevrnd, random, randtool
value
Generalized Pareto gprnd, random, randtool
Geometric geornd, random, randtool
Gumbel copula copularnd
Hypergeometric hygernd, random
Inverse Wishart iwishrnd
Johnson system johnsrnd
Lognormal lognrnd, random, randtool
Multinomial mnrnd
6-3
6 Random Number Generation
Distribution Random Number Generation Function
Multivariate normal mvnrnd
Multivariate t mvtrnd
Negative binomial nbinrnd, random, randtool
Noncentral chi-square ncx2rnd, random, randtool
Noncentral F ncfrnd, random, randtool
Noncentral t nctrnd, random, randtool
Normal (Gaussian) normrnd, randn, random, randtool
Pearson system pearsrnd
Piecewise random
Poisson poissrnd, random, randtool
Rayleigh raylrnd, random, randtool
Student’s t trnd, random, randtool
t copula copularnd
Uniform (continuous) unifrnd, rand, random
Uniform (discrete) unidrnd, random, randtool
Weibull wblrnd, random
Wishart wishrnd
6-4
Common Generation Methods
Common Generation Methods
In this section...
“Direct Methods” on page 6-5
“Inversion Methods” on page 6-7
“Acceptance-Rejection Methods” on page 6-9
Methods for generating pseudorandom numbers usually start with uniform
random numbers, like the MATLAB rand function produces. The methods
described in this section detail how to produce random numbers from other
distributions.
Direct Methods
Direct methods directly use the definition of the distribution.
For example, consider binomial random numbers. A binomial random number
is the number of heads in N tosses of a coin with probability p of a heads on
any single toss. If you generate N uniform random numbers on the interval
(0,1) and count the number less than p, then the count is a binomial random
number with parameters N and p.
This function is a simple implementation of a binomial RNG using the direct
approach:
function X = directbinornd(N,p,m,n)
X = zeros(m,n); % Preallocate memory
for i = 1:m*n
u = rand(N,1);
X(i) = sum(u < p);
end
For example:
X = directbinornd(100,0.3,1e4,1);
hist(X,101)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
6-5
6 Random Number Generation
The Statistics Toolbox function binornd uses a modified direct method, based
on the definition of a binomial random variable as the sum of Bernoulli
random variables.
You can easily convert the previous method to a random number generator
for the Poisson distribution with parameter λ. The Poisson distribution is
the limiting case of the binomial distribution as N approaches infinity, p
approaches zero, and Np is held fixed at λ. To generate Poisson random
numbers, create a version of the previous generator that inputs λ rather than
N and p, and internally sets N to some large number and p to λ/N.
The Statistics Toolbox function poissrnd actually uses two direct methods:
• A waiting time method for small values of λ
• A method due to Ahrens and Dieter for larger values of λ
6-6
Common Generation Methods
Inversion Methods
Inversion methods are based on the observation that continuous cumulative
distribution functions (cdfs) range uniformly over the interval (0,1). If u is a
uniform random number on (0,1), then using X = F -1(U) generates a random
number X from a continuous distribution with specified cdf F.
For example, the following code generates random numbers from a specific
exponential distribution using the inverse cdf and the MATLAB uniform
random number generator rand:
mu = 1;
X = expinv(rand(1e4,1),mu);
Compare the distribution of the generated random numbers to the pdf of the
specified exponential by scaling the pdf to the area of the histogram used
to display the distribution:
numbins = 50;
hist(X,numbins)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
hold on
[bincounts,binpositions] = hist(X,numbins);
binwidth = binpositions(2) - binpositions(1);
histarea = binwidth*sum(bincounts);
x = binpositions(1):0.001:binpositions(end);
y = exppdf(x,mu);
plot(x,histarea*y,'r','LineWidth',2)
6-7
6 Random Number Generation
Inversion methods also work for discrete distributions. To generate a random
number X from a discrete distribution with probability mass vector P(X=xi) =
pi where x0<x1< x2<..., generate a uniform random number u on (0,1) and then
set X = xi if F(xi–1)<u<F(xi).
For example, the following function implements an inversion method for a
discrete distribution with probability mass vector p:
function X = discreteinvrnd(p,m,n)
X = zeros(m,n); % Preallocate memory
for i = 1:m*n
u = rand;
I = find(u < cumsum(p));
X(i) = min(I);
end
Use the function to generate random numbers from any discrete distribution:
6-8
Common Generation Methods
p = [0.1 0.2 0.3 0.2 0.1 0.1]; % Probability mass vector
X = discreteinvrnd(p,1e4,1);
[n,x] = hist(X,length(p));
bar(1:length(p),n)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
Acceptance-Rejection Methods
The functional form of some distributions makes it difficult or time-consuming
to generate random numbers using direct or inversion methods.
Acceptance-rejection methods provide an alternative in these cases.
Acceptance-rejection methods begin with uniform random numbers, but
require an additional random number generator. If your goal is to generate a
random number from a continuous distribution with pdf f, acceptance-rejection
methods first generate a random number from a continuous distribution with
pdf g satisfying f(x) ≤ cg(x) for some c and all x.
6-9
6 Random Number Generation
A continuous acceptance-rejection RNG proceeds as follows:
1 Chooses a density g.
2 Finds a constant c such that f(x)/g(x)≤c for all x.
3 Generates a uniform random number u.
4 Generates a random number v from g.
5 If cu≤f(v)/g (v), accepts and returns v.
6 Otherwise, rejects v and goes to step 3.
For efficiency, a “cheap” method is necessary for generating random numbers
from g, and the scalar c should be small. The expected number of iterations to
produce a single random number is c.
The following function implements an acceptance-rejection method for
generating random numbers from pdf f, given f, g, the RNG grnd for g, and
the constant c:
6-10
Common Generation Methods
function X = accrejrnd(f,g,grnd,c,m,n)
X = zeros(m,n); % Preallocate memory
for i = 1:m*n
accept = false;
while accept == false
u = rand();
v = grnd();
if c*u <= f(v)/g(v)
X(i) = v;
accept = true;
end
end
end
For example, the function f(x) = xe–x2/2 satisfies the conditions for a pdf on [0,∞)
(nonnegative and integrates to 1). The exponential pdf with mean 1, f(x) = e–x,
dominates g for c greater than about 2.2. Thus, you can use rand and exprnd
to generate random numbers from f:
f = @(x)x.*exp(-(x.^2)/2);
g = @(x)exp(-x);
grnd = @()exprnd(1);
X = accrejrnd(f,g,grnd,2.2,1e4,1);
The pdf f is actually a Rayleigh distribution with shape parameter 1. This
example compares the distribution of random numbers generated by the
acceptance-rejection method with those generated by raylrnd:
Y = raylrnd(1,1e4,1);
hist([X Y])
h = get(gca,'Children');
set(h(1),'FaceColor',[.8 .8 1])
legend('A-R RNG','Rayleigh RNG')
6-11
6 Random Number Generation
The Statistics Toolbox function raylrnd uses a transformation method,
expressing a Rayleigh random variable in terms of a chi-square random
variable, which you compute using randn.
Acceptance-rejection methods also work for discrete distributions. In this case,
the goal is to generate random numbers from a distribution with probability
mass Pp(X = i) = pi, assuming that you have a method for generating random
numbers from a distribution with probability mass Pq(X = i) = qi. The RNG
proceeds as follows:
1 Chooses a density Pq.
2 Finds a constant c such that pi/qi≤c for all i .
3 Generates a uniform random number u.
4 Generates a random number v from Pq.
5 If cu≤pv/qv, accepts and returns v.
6 Otherwise, rejects v and goes to step 3.
6-12
Parallel Computing Support for Random Number Generation
Parallel Computing Support for Random Number
Generation
In this section...
“What is Parallel Computing?” on page 6-13
“Reproducing Computations” on page 6-13
“Assigning Random Number Generators” on page 6-14
What is Parallel Computing?
Parallel computing is the technique of using multiple processors on a single
problem. The primary reason to use parallel computing is to shorten the
computation time.
The following functions use random number generators and support both
parallel and serial computation. They supply two options to control random
number generation, whether in serial and parallel mode.
• bootci
• bootstrp
• TreeBagger
• TreeBagger.growTrees
Reproducing Computations
The previous functions include the 'UseSubstreams' option. This option
provides a quick and easy way to reproduce computations performed using
random number generators. Use this option to rerun a command with
reproducible results, whether using serial or parallel computation. This
option is available only with RandStream types that support substreams. The
default is not to use substreams, since reproducing random number streams
is not commonly desired.
6-13
6 Random Number Generation
Assigning Random Number Generators
The previous functions also include the 'Streams' option. use 'Streams' to
assign specific random number streams to each processor used in the function
evaluation, both in serial and parallel mode. The default is to use the default
random number stream on each processor. The option can be used to control
statistical behavior in your code and in function libraries that you use.
For more information on each of these options, see the function reference
pages.
6-14
Representing Sampling Distributions Using Markov Chain Samplers
Representing Sampling Distributions Using Markov Chain
Samplers
In this section...
“Using the Metropolis-Hastings Algorithm” on page 6-15
“Using Slice Sampling” on page 6-16
The methods in “Common Generation Methods” on page 6-5 might be
inadequate when sampling distributions are difficult to represent in
computations. Such distributions arise, for example, in Bayesian data
analysis and in the large combinatorial problems of Markov chain Monte
Carlo (MCMC) simulations. An alternative is to construct a Markov chain
with a stationary distribution equal to the target sampling distribution, using
the states of the chain to generate random numbers after an initial burn-in
period in which the state distribution converges to the target.
Using the Metropolis-Hastings Algorithm
The Metropolis-Hastings algorithm draws samples from a distribution that
is only known up to a constant. Random numbers are generated from a
distribution with a probability density function that is equal to or proportional
to a proposal function.
To generate random numbers:
1 Assume an initial value x(t).
2 Draw a sample, y(t), from a proposal distribution q(y|x(t)).
3 Accept y(t) as the next sample x(t + 1) with probability r(x(t),y(t)), and keep
x(t) as the next sample x(t + 1) with probability 1 – r(x(t),y(t)), where:
⎧ f ( y) q( x | y) ⎫
r( x, y) = min ⎨ , 1⎬
⎩ f ( x) q( y| x) ⎭
4 Increment t → t+1, and repeat steps 2 and 3 until you get the desired
number of samples.
6-15
6 Random Number Generation
Generate random numbers using the Metropolis-Hastings method with
the mhsample function. To produce quality samples efficiently with the
Metropolis-Hastings algorithm, it is crucial to select a good proposal
distribution. If it is difficult to find an efficient proposal distribution, use
the slice sampling algorithm (slicesample) without explicitly specifying a
proposal distribution.
Using Slice Sampling
In instances where it is difficult to find an efficient Metropolis-Hastings
proposal distribution, the slice sampling algorithm does not require an explicit
specification. The slice sampling algorithm draws samples from the region
under the density function using a sequence of vertical and horizontal steps.
First, it selects a height at random from 0 to the density function f (x). Then,
it selects a new x value at random by sampling from the horizontal “slice” of
the density above the selected height. A similar slice sampling algorithm is
used for a multivariate distribution.
If a function f(x) proportional to the density function is given, then do the
following to generate random numbers:
1 Assume an initial value x(t) within the domain of f(x).
2 Draw a real value y uniformly from (0, f(x(t))), thereby defining a horizontal
“slice” as S = {x: y < f(x)}.
3 Find an interval I = (L, R) around x(t) that contains all, or much of the
“slice” S.
4 Draw the new point x(t+1) within this interval.
5 Increment t → t+1 and repeat steps 2 through 4 until you get the desired
number of samples.
Slice sampling can generate random numbers from a distribution with an
arbitrary form of the density function, provided that an efficient numerical
procedure is available to find the interval I = (L,R), which is the “slice” of
the density.
Generate random numbers using the slice sampling method with the
slicesample function.
6-16
Generating Quasi-Random Numbers
Generating Quasi-Random Numbers
In this section...
“Quasi-Random Sequences” on page 6-17
“Quasi-Random Point Sets” on page 6-18
“Quasi-Random Streams” on page 6-25
Quasi-Random Sequences
Quasi-random number generators (QRNGs) produce highly uniform samples
of the unit hypercube. QRNGs minimize the discrepancy between the
distribution of generated points and a distribution with equal proportions of
points in each sub-cube of a uniform partition of the hypercube. As a result,
QRNGs systematically fill the “holes” in any initial segment of the generated
quasi-random sequence.
Unlike the pseudorandom sequences described in “Common Generation
Methods” on page 6-5, quasi-random sequences fail many statistical tests for
randomness. Approximating true randomness, however, is not their goal.
Quasi-random sequences seek to fill space uniformly, and to do so in such a
way that initial segments approximate this behavior up to a specified density.
QRNG applications include:
• Quasi-Monte Carlo (QMC) integration. Monte Carlo techniques are
often used to evaluate difficult, multi-dimensional integrals without a
closed-form solution. QMC uses quasi-random sequences to improve the
convergence properties of these techniques.
• Space-filling experimental designs. In many experimental settings,
taking measurements at every factor setting is expensive or infeasible.
Quasi-random sequences provide efficient, uniform sampling of the design
space.
• Global optimization. Optimization algorithms typically find a local
optimum in the neighborhood of an initial value. By using a quasi-random
sequence of initial values, searches for global optima uniformly sample the
basins of attraction of all local minima.
6-17
6 Random Number Generation
Example: Using Scramble, Leap, and Skip
Imagine a simple 1-D sequence that produces the integers from 1 to 10. This
is the basic sequence and the first three points are [1,2,3]:
Now look at how Scramble, Leap, and Skip work together:
• Scramble — Scrambling shuffles the points in one of several different
ways. In this example, assume a scramble turns the sequence into
1,3,5,7,9,2,4,6,8,10. The first three points are now [1,3,5]:
• Skip — A Skip value specifies the number of initial points to ignore. In this
example, set the Skip value to 2. The sequence is now 5,7,9,2,4,6,8,10
and the first three points are [5,7,9]:
• Leap — A Leap value specifies the number of points to ignore for each one
you take. Continuing the example with the Skip set to 2, if you set the Leap
to 1, the sequence uses every other point. In this example, the sequence is
now 5,9,4,8 and the first three points are [5,9,4]:
Quasi-Random Point Sets
Statistics Toolbox functions support these quasi-random sequences:
• Halton sequences. Produced by the haltonset function. These sequences
use different prime bases to form successively finer uniform partitions of
the unit interval in each dimension.
6-18
Generating Quasi-Random Numbers
• Sobol sequences. Produced by the sobolset function. These sequences
use a base of 2 to form successively finer uniform partitions of the unit
interval, and then reorder the coordinates in each dimension.
• Latin hypercube sequences. Produced by the lhsdesign function.
Though not quasi-random in the sense of minimizing discrepancy,
these sequences nevertheless produce sparse uniform samples useful in
experimental designs.
Quasi-random sequences are functions from the positive integers to the unit
hypercube. To be useful in application, an initial point set of a sequence must
be generated. Point sets are matrices of size n-by-d, where n is the number of
points and d is the dimension of the hypercube being sampled. The functions
haltonset and sobolset construct point sets with properties of a specified
quasi-random sequence. Initial segments of the point sets are generated by
the net method of the qrandset class (parent class of the haltonset class
and sobolset class), but points can be generated and accessed more generally
using parenthesis indexing.
Because of the way in which quasi-random sequences are generated, they
may contain undesirable correlations, especially in their initial segments, and
especially in higher dimensions. To address this issue, quasi-random point
sets often skip, leap over, or scramble values in a sequence. The haltonset
and sobolset functions allow you to specify both a Skip and a Leap property
of a quasi-random sequence, and the scramble method of the qrandset class
allows you apply a variety of scrambling techniques. Scrambling reduces
correlations while also improving uniformity.
Example: Generate a Quasi-Random Point Set
This example uses haltonset to construct a 2-D Halton point set—an object,
p, of the haltonset class—that skips the first 1000 values of the sequence
and then retains every 101st point:
p = haltonset(2,'Skip',1e3,'Leap',1e2)
p =
Halton point set in 2 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : none
6-19
6 Random Number Generation
The object p encapsulates properties of the specified quasi-random sequence.
The point set is finite, with a length determined by the Skip and Leap
properties and by limits on the size of point set indices (maximum value of 253).
Use scramble to apply reverse-radix scrambling:
p = scramble(p,'RR2')
p =
Halton point set in 2 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : RR2
Use net to generate the first 500 points:
X0 = net(p,500);
This is equivalent to:
X0 = p(1:500,:);
Values of the point set X0 are not generated and stored in memory until you
access p using net or parenthesis indexing.
To appreciate the nature of quasi-random numbers, create a scatter of the
two dimensions in X0:
scatter(X0(:,1),X0(:,2),5,'r')
axis square
title('{\bf Quasi-Random Scatter}')
6-20
Generating Quasi-Random Numbers
Compare this to a scatter of uniform pseudorandom numbers generated by
the MATLAB rand function:
X = rand(500,2);
scatter(X(:,1),X(:,2),5,'b')
axis square
title('{\bf Uniform Random Scatter}')
6-21
6 Random Number Generation
The quasi-random scatter appears more uniform, avoiding the clumping in
the pseudorandom scatter.
In a statistical sense, quasi-random numbers are too uniform to pass
traditional tests of randomness. For example, a Kolmogorov-Smirnov test,
performed by kstest, is used to assess whether or not a point set has a
uniform random distribution. When performed repeatedly on uniform
pseudorandom samples, such as those generated by rand, the test produces
a uniform distribution of p-values:
nTests = 1e5;
sampSize = 50;
PVALS = zeros(nTests,1);
for test = 1:nTests
x = rand(sampSize,1);
[h,pval] = kstest(x,[x,x]);
6-22
Generating Quasi-Random Numbers
PVALS(test) = pval;
end
hist(PVALS,100)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
xlabel('{\it p}-values')
ylabel('Number of Tests')
The results are quite different when the test is performed repeatedly on
uniform quasi-random samples:
p = haltonset(1,'Skip',1e3,'Leap',1e2);
p = scramble(p,'RR2');
nTests = 1e5;
sampSize = 50;
PVALS = zeros(nTests,1);
for test = 1:nTests
6-23
6 Random Number Generation
x = p(test:test+(sampSize-1),:);
[h,pval] = kstest(x,[x,x]);
PVALS(test) = pval;
end
hist(PVALS,100)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
xlabel('{\it p}-values')
ylabel('Number of Tests')
Small p-values call into question the null hypothesis that the data are
uniformly distributed. If the hypothesis is true, about 5% of the p-values are
expected to fall below 0.05. The results are remarkably consistent in their
failure to challenge the hypothesis.
6-24
Generating Quasi-Random Numbers
Quasi-Random Streams
Quasi-random streams, produced by the qrandstream function, are used
to generate sequential quasi-random outputs, rather than point sets of a
specific size. Streams are used like pseudoRNGS, such as rand, when client
applications require a source of quasi-random numbers of indefinite size that
can be accessed intermittently. Properties of a quasi-random stream, such
as its type (Halton or Sobol), dimension, skip, leap, and scramble, are set
when the stream is constructed.
In implementation, quasi-random streams are essentially very large
quasi-random point sets, though they are accessed differently. The state of a
quasi-random stream is the scalar index of the next point to be taken from the
stream. Use the qrand method of the qrandstream class to generate points
from the stream, starting from the current state. Use the reset method to
reset the state to 1. Unlike point sets, streams do not support parenthesis
indexing.
Example: Generate a Quasi-Random Stream
For example, the following code, taken from the example at the end of
“Quasi-Random Point Sets” on page 6-18, uses haltonset to create a
quasi-random point set p, and then repeatedly increments the index into the
point set, test, to generate different samples:
p = haltonset(1,'Skip',1e3,'Leap',1e2);
p = scramble(p,'RR2');
nTests = 1e5;
sampSize = 50;
PVALS = zeros(nTests,1);
for test = 1:nTests
x = p(test:test+(sampSize-1),:);
[h,pval] = kstest(x,[x,x]);
PVALS(test) = pval;
end
The same results are obtained by using qrandstream to construct a
quasi-random stream q based on the point set p and letting the stream take
care of increments to the index:
6-25
6 Random Number Generation
p = haltonset(1,'Skip',1e3,'Leap',1e2);
p = scramble(p,'RR2');
q = qrandstream(p)
nTests = 1e5;
sampSize = 50;
PVALS = zeros(nTests,1);
for test = 1:nTests
X = qrand(q,sampSize);
[h,pval] = kstest(X,[X,X]);
PVALS(test) = pval;
end
6-26
Generating Data Using Flexible Families of Distributions
Generating Data Using Flexible Families of Distributions
In this section...
“Pearson and Johnson Systems” on page 6-27
“Generating Data Using the Pearson System” on page 6-28
“Generating Data Using the Johnson System” on page 6-30
Pearson and Johnson Systems
As described in “Using Probability Distributions” on page 5-2, choosing an
appropriate parametric family of distributions to model your data can be
based on a priori or a posteriori knowledge of the data-producing process,
but the choice is often difficult. The Pearson and Johnson systems can make
such a choice unnecessary. Each system is a flexible parametric family of
distributions that includes a wide range of distribution shapes, and it is often
possible to find a distribution within one of these two systems that provides
a good match to your data.
Data Input
The following parameters define each member of the Pearson and Johnson
systems
• Mean — Estimated by mean
• Standard deviation — Estimated by std
• Skewness — Estimated by skewness
• Kurtosis — Estimated by kurtosis
These statistics can also be computed with the moment function. The Johnson
system, while based on these four parameters, is more naturally described
using quantiles, estimated by the quantile function.
The Statistics Toolbox functions pearsrnd and johnsrnd take input
arguments defining a distribution (parameters or quantiles, respectively) and
return the type and the coefficients of the distribution in the corresponding
system. Both functions also generate random numbers from the specified
distribution.
6-27
6 Random Number Generation
As an example, load the data in carbig.mat, which includes a variable MPG
containing measurements of the gas mileage for each car.
load carbig
MPG = MPG(~isnan(MPG));
[n,x] = hist(MPG,15);
bar(x,n)
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
The following two sections model the distribution with members of the
Pearson and Johnson systems, respectively.
Generating Data Using the Pearson System
The statistician Karl Pearson devised a system, or family, of distributions
that includes a unique distribution corresponding to every valid combination
of mean, standard deviation, skewness, and kurtosis. If you compute sample
6-28
Generating Data Using Flexible Families of Distributions
values for each of these moments from data, it is easy to find the distribution
in the Pearson system that matches these four moments and to generate a
random sample.
The Pearson system embeds seven basic types of distribution together in
a single parametric framework. It includes common distributions such
as the normal and t distributions, simple transformations of standard
distributions such as a shifted and scaled beta distribution and the inverse
gamma distribution, and one distribution—the Type IV—that is not a simple
transformation of any standard distribution.
For a given set of moments, there are distributions that are not in the system
that also have those same first four moments, and the distribution in the
Pearson system may not be a good match to your data, particularly if the
data are multimodal. But the system does cover a wide range of distribution
shapes, including both symmetric and skewed distributions.
To generate a sample from the Pearson distribution that closely matches
the MPG data, simply compute the four sample moments and treat those as
distribution parameters.
moments = {mean(MPG),std(MPG),skewness(MPG),kurtosis(MPG)};
[r,type] = pearsrnd(moments{:},10000,1);
The optional second output from pearsrnd indicates which type of distribution
within the Pearson system matches the combination of moments.
type
type =
1
In this case, pearsrnd has determined that the data are best described with a
Type I Pearson distribution, which is a shifted, scaled beta distribution.
Verify that the sample resembles the original data by overlaying the empirical
cumulative distribution functions.
ecdf(MPG);
[Fi,xi] = ecdf(r);
hold on, stairs(xi,Fi,'r'); hold off
6-29
6 Random Number Generation
Generating Data Using the Johnson System
Statistician Norman Johnson devised a different system of distributions that
also includes a unique distribution for every valid combination of mean,
standard deviation, skewness, and kurtosis. However, since it is more natural
to describe distributions in the Johnson system using quantiles, working with
this system is different than working with the Pearson system.
The Johnson system is based on three possible transformations of a normal
random variable, plus the identity transformation. The three nontrivial cases
are known as SL, SU, and SB, corresponding to exponential, logistic, and
hyperbolic sine transformations. All three can be written as
⎛ ( Ζ-ξ ) ⎞
Χ = γ + δ ⋅ Γ ⎜⎜ ⎟⎟
⎝ λ ⎠
6-30
Generating Data Using Flexible Families of Distributions
where Z is a standard normal random variable, Γ is the transformation, and
γ, δ, ξ, and λ are scale and location parameters. The fourth case, SN, is the
identity transformation.
To generate a sample from the Johnson distribution that matches the MPG
data, first define the four quantiles to which the four evenly spaced standard
normal quantiles of -1.5, -0.5, 0.5, and 1.5 should be transformed. That is, you
compute the sample quantiles of the data for the cumulative probabilities of
0.067, 0.309, 0.691, and 0.933.
probs = normcdf([-1.5 -0.5 0.5 1.5])
probs =
0.066807 0.30854 0.69146 0.93319
quantiles = quantile(MPG,probs)
quantiles =
13.0000 18.0000 27.2000 36.0000
Then treat those quantiles as distribution parameters.
[r1,type] = johnsrnd(quantiles,10000,1);
The optional second output from johnsrnd indicates which type of distribution
within the Johnson system matches the quantiles.
type
type =
SB
You can verify that the sample resembles the original data by overlaying the
empirical cumulative distribution functions.
ecdf(MPG);
[Fi,xi] = ecdf(r1);
hold on, stairs(xi,Fi,'r'); hold off
6-31
6 Random Number Generation
In some applications, it may be important to match the quantiles better in
some regions of the data than in others. To do that, specify four evenly spaced
standard normal quantiles at which you want to match the data, instead
of the default -1.5, -0.5, 0.5, and 1.5. For example, you might care more
about matching the data in the right tail than in the left, and so you specify
standard normal quantiles that emphasizes the right tail.
qnorm = [-.5 .25 1 1.75];
probs = normcdf(qnorm);
qemp = quantile(MPG,probs);
r2 = johnsrnd([qnorm; qemp],10000,1);
However, while the new sample matches the original data better in the right
tail, it matches much worse in the left tail.
[Fj,xj] = ecdf(r2);
hold on, stairs(xj,Fj,'g'); hold off
6-32
Generating Data Using Flexible Families of Distributions
6-33
6 Random Number Generation
6-34
7
Hypothesis Tests
• “Introduction” on page 7-2
• “Hypothesis Test Terminology” on page 7-3
• “Hypothesis Test Assumptions” on page 7-5
• “Example: Hypothesis Testing” on page 7-7
• “Available Hypothesis Tests” on page 7-13
7 Hypothesis Tests
Introduction
Hypothesis testing is a common method of drawing inferences about a
population based on statistical evidence from a sample.
As an example, suppose someone says that at a certain time in the state
of Massachusetts the average price of a gallon of regular unleaded gas was
$1.15. How could you determine the truth of the statement? You could try to
find prices at every gas station in the state at the time. That approach would
be definitive, but it could be time-consuming, costly, or even impossible.
A simpler approach would be to find prices at a small number of randomly
selected gas stations around the state, and then compute the sample average.
Sample averages differ from one another due to chance variability in the
selection process. Suppose your sample average comes out to be $1.18. Is the
$0.03 difference an artifact of random sampling or significant evidence that
the average price of a gallon of gas was in fact greater than $1.15? Hypothesis
testing is a statistical method for making such decisions.
7-2
Hypothesis Test Terminology
Hypothesis Test Terminology
All hypothesis tests share the same basic terminology and structure.
• A null hypothesis is an assertion about a population that you would like to
test. It is “null” in the sense that it often represents a status quo belief,
such as the absence of a characteristic or the lack of an effect. It may be
formalized by asserting that a population parameter, or a combination of
population parameters, has a certain value. In the example given in the
“Introduction” on page 7-2, the null hypothesis would be that the average
price of gas across the state was $1.15. This is written H0: µ = 1.15.
• An alternative hypothesis is a contrasting assertion about the population
that can be tested against the null hypothesis. In the example given in the
“Introduction” on page 7-2, possible alternative hypotheses are:
H1: µ ≠ 1.15 — State average was different from $1.15 (two-tailed test)
H1: µ > 1.15 — State average was greater than $1.15 (right-tail test)
H1: µ < 1.15 — State average was less than $1.15 (left-tail test)
• To conduct a hypothesis test, a random sample from the population is
collected and a relevant test statistic is computed to summarize the sample.
This statistic varies with the type of test, but its distribution under the null
hypothesis must be known (or assumed).
• The p value of a test is the probability, under the null hypothesis, of
obtaining a value of the test statistic as extreme or more extreme than the
value computed from the sample.
• The significance level of a test is a threshold of probability α agreed to before
the test is conducted. A typical value of α is 0.05. If the p value of a test is
less than α, the test rejects the null hypothesis. If the p value is greater
than α, there is insufficient evidence to reject the null hypothesis. Note
that lack of evidence for rejecting the null hypothesis is not evidence for
accepting the null hypothesis. Also note that substantive “significance” of
an alternative cannot be inferred from the statistical significance of a test.
• The significance level α can be interpreted as the probability of rejecting
the null hypothesis when it is actually true—a type I error. The distribution
of the test statistic under the null hypothesis determines the probability
α of a type I error. Even if the null hypothesis is not rejected, it may still
be false—a type II error. The distribution of the test statistic under the
7-3
7 Hypothesis Tests
alternative hypothesis determines the probability β of a type II error. Type
II errors are often due to small sample sizes. The power of a test, 1 – β, is
the probability of correctly rejecting a false null hypothesis.
• Results of hypothesis tests are often communicated with a confidence
interval. A confidence interval is an estimated range of values with a
specified probability of containing the true population value of a parameter.
Upper and lower bounds for confidence intervals are computed from the
sample estimate of the parameter and the known (or assumed) sampling
distribution of the estimator. A typical assumption is that estimates will be
normally distributed with repeated sampling (as dictated by the Central
Limit Theorem). Wider confidence intervals correspond to poor estimates
(smaller samples); narrow intervals correspond to better estimates
(larger samples). If the null hypothesis asserts the value of a population
parameter, the test rejects the null hypothesis when the hypothesized
value lies outside the computed confidence interval for the parameter.
7-4
Hypothesis Test Assumptions
Hypothesis Test Assumptions
Different hypothesis tests make different assumptions about the distribution
of the random variable being sampled in the data. These assumptions must
be considered when choosing a test and when interpreting the results.
For example, the z-test (ztest) and the t-test (ttest) both assume that
the data are independently sampled from a normal distribution. Statistics
Toolbox functions are available for testing this assumption, such as chi2gof,
jbtest, lillietest, and normplot.
Both the z-test and the t-test are relatively robust with respect to departures
from this assumption, so long as the sample size n is large enough. Both
tests compute a sample mean x , which, by the Central Limit Theorem,
has an approximately normal sampling distribution with mean equal to the
population mean μ, regardless of the population distribution being sampled.
The difference between the z-test and the t-test is in the assumption of the
standard deviation σ of the underlying normal distribution. A z-test assumes
that σ is known; a t-test does not. As a result, a t-test must compute an
estimate s of the standard deviation from the sample.
Test statistics for the z-test and the t-test are, respectively,
x−
z=
/ n
x−
t=
s/ n
Under the null hypothesis that the population is distributed with mean μ, the
z-statistic has a standard normal distribution, N(0,1). Under the same null
hypothesis, the t-statistic has Student’s t distribution with n – 1 degrees of
freedom. For small sample sizes, Student’s t distribution is flatter and wider
than N(0,1), compensating for the decreased confidence in the estimate s.
As sample size increases, however, Student’s t distribution approaches the
standard normal distribution, and the two tests become essentially equivalent.
7-5
7 Hypothesis Tests
Knowing the distribution of the test statistic under the null hypothesis allows
for accurate calculation of p-values. Interpreting p-values in the context of
the test assumptions allows for critical analysis of test results.
Assumptions underlying Statistics Toolbox hypothesis tests are given in the
reference pages for implementing functions.
7-6
Example: Hypothesis Testing
Example: Hypothesis Testing
This example uses the gas price data in the file gas.mat. The file contains two
random samples of prices for a gallon of gas around the state of Massachusetts
in 1993. The first sample, price1, contains 20 random observations around
the state on a single day in January. The second sample, price2, contains 20
random observations around the state one month later.
load gas
prices = [price1 price2];
As a first step, you might want to test the assumption that the samples come
from normal distributions.
A normal probability plot gives a quick idea.
7-7
7 Hypothesis Tests
normplot(prices)
Both scatters approximately follow straight lines through the first and third
quartiles of the samples, indicating approximate normal distributions.
The February sample (the right-hand line) shows a slight departure from
normality in the lower tail. A shift in the mean from January to February is
evident.
A hypothesis test is used to quantify the test of normality. Since each sample
is relatively small, a Lilliefors test is recommended.
7-8
Example: Hypothesis Testing
lillietest(price1)
ans =
0
lillietest(price2)
ans =
0
The default significance level of lillietest is 5%. The logical 0 returned by
each test indicates a failure to reject the null hypothesis that the samples are
normally distributed. This failure may reflect normality in the population or
it may reflect a lack of strong evidence against the null hypothesis due to
the small sample size.
Now compute the sample means:
sample_means = mean(prices)
sample_means =
115.1500 118.5000
You might want to test the null hypothesis that the mean price across the
state on the day of the January sample was $1.15. If you know that the
standard deviation in prices across the state has historically, and consistently,
been $0.04, then a z-test is appropriate.
[h,pvalue,ci] = ztest(price1/100,1.15,0.04)
h =
0
pvalue =
0.8668
ci =
1.1340
1.1690
The logical output h = 0 indicates a failure to reject the null hypothesis
at the default significance level of 5%. This is a consequence of the high
probability under the null hypothesis, indicated by the p value, of observing
a value as extreme or more extreme of the z-statistic computed from the
sample. The 95% confidence interval on the mean [1.1340 1.1690] includes
the hypothesized population mean of $1.15.
7-9
7 Hypothesis Tests
Does the later sample offer stronger evidence for rejecting a null hypothesis
of a state-wide average price of $1.15 in February? The shift shown in the
probability plot and the difference in the computed sample means suggest
this. The shift might indicate a significant fluctuation in the market, raising
questions about the validity of using the historical standard deviation. If a
known standard deviation cannot be assumed, a t-test is more appropriate.
[h,pvalue,ci] = ttest(price2/100,1.15)
h =
1
pvalue =
4.9517e-004
ci =
1.1675
1.2025
The logical output h = 1 indicates a rejection of the null hypothesis at the
default significance level of 5%. In this case, the 95% confidence interval on
the mean does not include the hypothesized population mean of $1.15.
You might want to investigate the shift in prices a little more closely.
The function ttest2 tests if two independent samples come from normal
distributions with equal but unknown standard deviations and the same
mean, against the alternative that the means are unequal.
[h,sig,ci] = ttest2(price1,price2)
h =
1
sig =
0.0083
ci =
-5.7845
-0.9155
The null hypothesis is rejected at the default 5% significance level, and
the confidence interval on the difference of means does not include the
hypothesized value of 0.
A notched box plot is another way to visualize the shift.
7-10
Example: Hypothesis Testing
boxplot(prices,1)
set(gca,'XTick',[1 2])
set(gca,'XtickLabel',{'January','February'})
xlabel('Month')
ylabel('Prices ($0.01)')
The plot displays the distribution of the samples around their medians. The
heights of the notches in each box are computed so that the side-by-side
boxes have nonoverlapping notches when their medians are different at a
default 5% significance level. The computation is based on an assumption
of normality in the data, but the comparison is reasonably robust for other
distributions. The side-by-side plots provide a kind of visual hypothesis test,
comparing medians rather than means. The plot above appears to barely
reject the null hypothesis of equal medians.
The nonparametric Wilcoxon rank sum test, implemented by the function
ranksum, can be used to quantify the test of equal medians. It tests if two
independent samples come from identical continuous (not necessarily normal)
distributions with equal medians, against the alternative that they do not
have equal medians.
7-11
7 Hypothesis Tests
[p,h] = ranksum(price1,price2)
p =
0.0095
h =
1
The test rejects the null hypothesis of equal medians at the default 5%
significance level.
7-12
Available Hypothesis Tests
Available Hypothesis Tests
Function Description
ansaribradley Ansari-Bradley test. Tests if two independent samples
come from the same distribution, against the alternative
that they come from distributions that have the same
median and shape but different variances.
chi2gof Chi-square goodness-of-fit test. Tests if a sample comes
from a specified distribution, against the alternative
that it does not come from that distribution.
dwtest Durbin-Watson test. Tests if the residuals from a linear
regression are uncorrelated, against the alternative
that there is autocorrelation among them.
jbtest Jarque-Bera test. Tests if a sample comes from a
normal distribution with unknown mean and variance,
against the alternative that it does not come from a
normal distribution.
kstest One-sample Kolmogorov-Smirnov test. Tests if a sample
comes from a continuous distribution with specified
parameters, against the alternative that it does not
come from that distribution.
kstest2 Two-sample Kolmogorov-Smirnov test. Tests if two
samples come from the same continuous distribution,
against the alternative that they do not come from the
same distribution.
lillietest Lilliefors test. Tests if a sample comes from a
distribution in the normal family, against the
alternative that it does not come from a normal
distribution.
linhyptest Linear hypothesis test. Tests if H*b = c for parameter
estimates b with estimated covariance H and specified
c, against the alternative that H*b c.
7-13
7 Hypothesis Tests
Function Description
ranksum Wilcoxon rank sum test. Tests if two independent
samples come from identical continuous distributions
with equal medians, against the alternative that they
do not have equal medians.
runstest Runs test. Tests if a sequence of values comes in
random order, against the alternative that the ordering
is not random.
signrank One-sample or paired-sample Wilcoxon signed rank test.
Tests if a sample comes from a continuous distribution
symmetric about a specified median, against the
alternative that it does not have that median.
signtest One-sample or paired-sample sign test. Tests if a
sample comes from an arbitrary continuous distribution
with a specified median, against the alternative that it
does not have that median.
ttest One-sample or paired-sample t-test. Tests if a sample
comes from a normal distribution with unknown
variance and a specified mean, against the alternative
that it does not have that mean.
ttest2 Two-sample t-test. Tests if two independent samples
come from normal distributions with unknown but
equal (or, optionally, unequal) variances and the same
mean, against the alternative that the means are
unequal.
vartest One-sample chi-square variance test. Tests if a sample
comes from a normal distribution with specified
variance, against the alternative that it comes from a
normal distribution with a different variance.
vartest2 Two-sample F-test for equal variances. Tests if two
independent samples come from normal distributions
with the same variance, against the alternative that
they come from normal distributions with different
variances.
7-14
Available Hypothesis Tests
Function Description
vartestn Bartlett multiple-sample test for equal variances. Tests
if multiple samples come from normal distributions
with the same variance, against the alternative that
they come from normal distributions with different
variances.
ztest One-sample z-test. Tests if a sample comes from a
normal distribution with known variance and specified
mean, against the alternative that it does not have that
mean.
Note In addition to the previous functions, Statistics Toolbox functions are
available for analysis of variance (ANOVA), which perform hypothesis tests in
the context of linear modeling. These functions are discussed in Chapter 8,
“Analysis of Variance”.
7-15
7 Hypothesis Tests
7-16
8
Analysis of Variance
• “Introduction” on page 8-2
• “ANOVA” on page 8-3
• “MANOVA” on page 8-39
8 Analysis of Variance
Introduction
Analysis of variance (ANOVA) is a procedure for assigning sample variance to
different sources and deciding whether the variation arises within or among
different population groups. Samples are described in terms of variation
around group means and variation of group means around an overall mean. If
variations within groups are small relative to variations between groups, a
difference in group means may be inferred. Chapter 7, “Hypothesis Tests” are
used to quantify decisions.
This chapter treats ANOVA among groups, that is, among categorical
predictors. ANOVA for regression, with continuous predictors, is discussed in
“Tabulating Diagnostic Statistics” on page 9-13.
Multivariate analysis of variance (MANOVA), for data with multiple
measured responses, is also discussed in this chapter.
8-2
ANOVA
ANOVA
In this section...
“One-Way ANOVA” on page 8-3
“Two-Way ANOVA” on page 8-9
“N-Way ANOVA” on page 8-12
“Other ANOVA Models” on page 8-26
“Analysis of Covariance” on page 8-27
“Nonparametric Methods” on page 8-35
One-Way ANOVA
• “Introduction” on page 8-3
• “Example: One-Way ANOVA” on page 8-4
• “Multiple Comparisons” on page 8-6
• “Example: Multiple Comparisons” on page 8-7
Introduction
The purpose of one-way ANOVA is to find out whether data from several
groups have a common mean. That is, to determine whether the groups are
actually different in the measured characteristic.
One-way ANOVA is a simple special case of the linear model. The one-way
ANOVA form of the model is
yij = . j + ij
where:
• yij is a matrix of observations in which each column represents a different
group.
8-3
8 Analysis of Variance
• α.j is a matrix whose columns are the group means. (The “dot j” notation
means that α applies to all rows of column j. That is, the value αij is the
same for all i.)
• εij is a matrix of random disturbances.
The model assumes that the columns of y are a constant plus a random
disturbance. You want to know if the constants are all the same.
Example: One-Way ANOVA
The data below comes from a study by Hogg and Ledolter [48] of bacteria
counts in shipments of milk. The columns of the matrix hogg represent
different shipments. The rows are bacteria counts from cartons of milk chosen
randomly from each shipment. Do some shipments have higher counts than
others?
load hogg
hogg
hogg =
24 14 11 7 19
15 7 9 7 24
21 12 7 4 19
27 17 13 7 15
33 14 12 12 10
23 16 18 18 20
[p,tbl,stats] = anova1(hogg);
p
p =
1.1971e-04
The standard ANOVA table has columns for the sums of squares, degrees of
freedom, mean squares (SS/df), F statistic, and p value.
8-4
ANOVA
You can use the F statistic to do a hypothesis test to find out if the bacteria
counts are the same. anova1 returns the p value from this hypothesis test.
In this case the p value is about 0.0001, a very small value. This is a strong
indication that the bacteria counts from the different shipments are not the
same. An F statistic as extreme as the observed F would occur by chance only
once in 10,000 times if the counts were truly equal.
The p value returned by anova1 depends on assumptions about the random
disturbances εij in the model equation. For the p value to be correct, these
disturbances need to be independent, normally distributed, and have constant
variance.
You can get some graphical assurance that the means are different by
looking at the box plots in the second figure window displayed by anova1.
Note, however, that the notches are used for a comparison of medians, not a
comparison of means. For more information on this display, see “Box Plots”
on page 4-6.
8-5
8 Analysis of Variance
Multiple Comparisons
Sometimes you need to determine not just whether there are any differences
among the means, but specifically which pairs of means are significantly
different. It is tempting to perform a series of t tests, one for each pair of
means, but this procedure has a pitfall.
In a t test, you compute a t statistic and compare it to a critical value. The
critical value is chosen so that when the means are really the same (any
apparent difference is due to random chance), the probability that the t
statistic will exceed the critical value is small, say 5%. When the means
are different, the probability that the statistic will exceed the critical value
is larger.
In this example there are five means, so there are 10 pairs of means to
compare. It stands to reason that if all the means are the same, and if there is
a 5% chance of incorrectly concluding that there is a difference in one pair,
8-6
ANOVA
then the probability of making at least one incorrect conclusion among all 10
pairs is much larger than 5%.
Fortunately, there are procedures known as multiple comparison procedures
that are designed to compensate for multiple tests.
Example: Multiple Comparisons
You can perform a multiple comparison test using the multcompare function
and supplying it with the stats output from anova1.
load hogg
[p,tbl,stats] = anova1(hogg);
[c,m] = multcompare(stats)
c =
1.0000 2.0000 2.4953 10.5000 18.5047
1.0000 3.0000 4.1619 12.1667 20.1714
1.0000 4.0000 6.6619 14.6667 22.6714
1.0000 5.0000 -2.0047 6.0000 14.0047
2.0000 3.0000 -6.3381 1.6667 9.6714
2.0000 4.0000 -3.8381 4.1667 12.1714
2.0000 5.0000 -12.5047 -4.5000 3.5047
3.0000 4.0000 -5.5047 2.5000 10.5047
3.0000 5.0000 -14.1714 -6.1667 1.8381
4.0000 5.0000 -16.6714 -8.6667 -0.6619
m =
23.8333 1.9273
13.3333 1.9273
11.6667 1.9273
9.1667 1.9273
17.8333 1.9273
The first output from multcompare has one row for each pair of groups, with
an estimate of the difference in group means and a confidence interval for that
group. For example, the second row has the values
1.0000 3.0000 4.1619 12.1667 20.1714
indicating that the mean of group 1 minus the mean of group 3 is
estimated to be 12.1667, and a 95% confidence interval for this difference is
8-7
8 Analysis of Variance
[4.1619, 20.1714]. This interval does not contain 0, so you can conclude that
the means of groups 1 and 3 are different.
The second output contains the mean and its standard error for each group.
It is easier to visualize the difference between group means by looking at the
graph that multcompare produces.
There are five groups. The graph instructs you to Click on the group you
want to test. Three groups have slopes significantly different from group one.
The graph shows that group 1 is significantly different from groups 2, 3, and
4. By using the mouse to select group 4, you can determine that it is also
significantly different from group 5. Other pairs are not significantly different.
8-8
ANOVA
Two-Way ANOVA
• “Introduction” on page 8-9
• “Example: Two-Way ANOVA” on page 8-10
Introduction
The purpose of two-way ANOVA is to find out whether data from several
groups have a common mean. One-way ANOVA and two-way ANOVA differ
in that the groups in two-way ANOVA have two categories of defining
characteristics instead of one.
Suppose an automobile company has two factories, and each factory makes
the same three models of car. It is reasonable to ask if the gas mileage in the
cars varies from factory to factory as well as from model to model. There are
two predictors, factory and model, to explain differences in mileage.
There could be an overall difference in mileage due to a difference in the
production methods between factories. There is probably a difference in the
mileage of the different models (irrespective of the factory) due to differences
in design specifications. These effects are called additive.
Finally, a factory might make high mileage cars in one model (perhaps
because of a superior production line), but not be different from the other
factory for other models. This effect is called an interaction. It is impossible
to detect an interaction unless there are duplicate observations for some
combination of factory and car model.
Two-way ANOVA is a special case of the linear model. The two-way ANOVA
form of the model is
yijk = + . j + i. + ij + ijk
where, with respect to the automobile example above:
• yijk is a matrix of gas mileage observations (with row index i, column index
j, and repetition index k).
• μ is a constant matrix of the overall mean gas mileage.
8-9
8 Analysis of Variance
• α.j is a matrix whose columns are the deviations of each car’s gas mileage
(from the mean gas mileage μ) that are attributable to the car’s model. All
values in a given column of α.j are identical, and the values in each row of
α.j sum to 0.
• βi. is a matrix whose rows are the deviations of each car’s gas mileage
(from the mean gas mileage μ) that are attributable to the car’s factory. All
values in a given row of βi. are identical, and the values in each column
of βi. sum to 0.
• γij is a matrix of interactions. The values in each row of γij sum to 0, and the
values in each column of γij sum to 0.
• εijk is a matrix of random disturbances.
Example: Two-Way ANOVA
The purpose of the example is to determine the effect of car model and factory
on the mileage rating of cars.
load mileage
mileage
mileage =
33.3000 34.5000 37.4000
33.4000 34.8000 36.8000
32.9000 33.8000 37.6000
32.6000 33.4000 36.6000
32.5000 33.7000 37.0000
33.0000 33.9000 36.7000
cars = 3;
[p,tbl,stats] = anova2(mileage,cars);
p
p =
0.0000 0.0039 0.8411
There are three models of cars (columns) and two factories (rows). The reason
there are six rows in mileage instead of two is that each factory provides
8-10
ANOVA
three cars of each model for the study. The data from the first factory is in the
first three rows, and the data from the second factory is in the last three rows.
The standard ANOVA table has columns for the sums of squares,
degrees-of-freedom, mean squares (SS/df), F statistics, and p-values.
You can use the F statistics to do hypotheses tests to find out if the mileage is
the same across models, factories, and model-factory pairs (after adjusting for
the additive effects). anova2 returns the p value from these tests.
The p value for the model effect is zero to four decimal places. This is a strong
indication that the mileage varies from one model to another. An F statistic
as extreme as the observed F would occur by chance less than once in 10,000
times if the gas mileage were truly equal from model to model. If you used the
multcompare function to perform a multiple comparison test, you would find
that each pair of the three models is significantly different.
The p value for the factory effect is 0.0039, which is also highly significant.
This indicates that one factory is out-performing the other in the gas mileage
of the cars it produces. The observed p value indicates that an F statistic as
extreme as the observed F would occur by chance about four out of 1000 times
if the gas mileage were truly equal from factory to factory.
There does not appear to be any interaction between factories and models.
The p value, 0.8411, means that the observed result is quite likely (84 out 100
times) given that there is no interaction.
8-11
8 Analysis of Variance
The p-values returned by anova2 depend on assumptions about the random
disturbances εijk in the model equation. For the p-values to be correct these
disturbances need to be independent, normally distributed, and have constant
variance.
In addition, anova2 requires that data be balanced, which in this case means
there must be the same number of cars for each combination of model and
factory. The next section discusses a function that supports unbalanced data
with any number of predictors.
N-Way ANOVA
• “Introduction” on page 8-12
• “N-Way ANOVA with a Small Data Set” on page 8-13
• “N-Way ANOVA with a Large Data Set” on page 8-15
• “ANOVA with Random Effects” on page 8-19
Introduction
You can use N-way ANOVA to determine if the means in a set of data differ
when grouped by multiple factors. If they do differ, you can determine which
factors or combinations of factors are associated with the difference.
N-way ANOVA is a generalization of two-way ANOVA. For three factors, the
model can be written
yijkl = + . j. + i.. + ..k + ( )ij. + ( )i.k + ( ). jk + ( )ijk + ijkl
In this notation parameters with two subscripts, such as (αβ)ij., represent
the interaction effect of two factors. The parameter (αβγ)ijk represents the
three-way interaction. An ANOVA model can have the full set of parameters
or any subset, but conventionally it does not include complex interaction
terms unless it also includes all simpler terms for those factors. For example,
one would generally not include the three-way interaction without also
including all two-way interactions.
8-12
ANOVA
The anovan function performs N-way ANOVA. Unlike the anova1 and anova2
functions, anovan does not expect data in a tabular form. Instead, it expects
a vector of response measurements and a separate vector (or text array)
containing the values corresponding to each factor. This input data format is
more convenient than matrices when there are more than two factors or when
the number of measurements per factor combination is not constant.
N-Way ANOVA with a Small Data Set
Consider the following two-way example using anova2.
m = [23 15 20;27 17 63;43 3 55;41 9 90]
m =
23 15 20
27 17 63
43 3 55
41 9 90
anova2(m,2)
ans =
0.0197 0.2234 0.2663
The factor information is implied by the shape of the matrix m and the number
of measurements at each factor combination (2). Although anova2 does not
actually require arrays of factor values, for illustrative purposes you could
create them as follows.
cfactor = repmat(1:3,4,1)
cfactor =
1 2 3
1 2 3
1 2 3
1 2 3
rfactor = [ones(2,3); 2*ones(2,3)]
rfactor =
1 1 1
8-13
8 Analysis of Variance
1 1 1
2 2 2
2 2 2
The cfactor matrix shows that each column of m represents a different level
of the column factor. The rfactor matrix shows that the top two rows of m
represent one level of the row factor, and bottom two rows of m represent a
second level of the row factor. In other words, each value m(i,j) represents
an observation at column factor level cfactor(i,j) and row factor level
rfactor(i,j).
To solve the above problem with anovan, you need to reshape the matrices m,
cfactor, and rfactor to be vectors.
m = m(:);
cfactor = cfactor(:);
rfactor = rfactor(:);
[m cfactor rfactor]
ans =
23 1 1
27 1 1
43 1 2
41 1 2
15 2 1
17 2 1
3 2 2
9 2 2
20 3 1
63 3 1
55 3 2
90 3 2
anovan(m,{cfactor rfactor},2)
ans =
0.0197
8-14
ANOVA
0.2234
0.2663
N-Way ANOVA with a Large Data Set
The previous example used anova2 to study a small data set measuring car
mileage. This example illustrates how to analyze a larger set of car data with
mileage and other information on 406 cars made between 1970 and 1982.
First, load the data set and look at the variable names.
load carbig
whos
Name Size Bytes Class
Acceleration 406x1 3248 double array
Cylinders 406x1 3248 double array
Displacement 406x1 3248 double array
Horsepower 406x1 3248 double array
MPG 406x1 3248 double array
Model 406x36 29232 char array
Model_Year 406x1 3248 double array
Origin 406x7 5684 char array
Weight 406x1 3248 double array
cyl4 406x5 4060 char array
org 406x7 5684 char array
when 406x5 4060 char array
The example focusses on four variables. MPG is the number of miles per gallon
for each of 406 cars (though some have missing values coded as NaN). The
other three variables are factors: cyl4 (four-cylinder car or not), org (car
originated in Europe, Japan, or the USA), and when (car was built early in the
period, in the middle of the period, or late in the period).
First, fit the full model, requesting up to three-way interactions and Type 3
sums-of-squares.
varnames = {'Origin';'4Cyl';'MfgDate'};
anovan(MPG,{org cyl4 when},3,3,varnames)
ans =
0.0000
8-15
8 Analysis of Variance
NaN
0
0.7032
0.0001
0.2072
0.6990
Note that many terms are marked by a # symbol as not having full rank,
and one of them has zero degrees of freedom and is missing a p value. This
can happen when there are missing factor combinations and the model has
higher-order terms. In this case, the cross-tabulation below shows that there
are no cars made in Europe during the early part of the period with other than
four cylinders, as indicated by the 0 in table(2,1,1).
[table, chi2, p, factorvals] = crosstab(org,when,cyl4)
table(:,:,1) =
82 75 25
0 4 3
3 3 4
8-16
ANOVA
table(:,:,2) =
12 22 38
23 26 17
12 25 32
chi2 =
207.7689
p =
factorvals =
'USA' 'Early' 'Other'
'Europe' 'Mid' 'Four'
'Japan' 'Late' []
Consequently it is impossible to estimate the three-way interaction effects,
and including the three-way interaction term in the model makes the fit
singular.
Using even the limited information available in the ANOVA table, you can see
that the three-way interaction has a p value of 0.699, so it is not significant.
So this time you examine only two-way interactions.
[p,tbl,stats,terms] = anovan(MPG,{org cyl4 when},2,3,varnames);
terms
terms =
1 0 0
0 1 0
0 0 1
1 1 0
1 0 1
0 1 1
8-17
8 Analysis of Variance
Now all terms are estimable. The p-values for interaction term 4
(Origin*4Cyl) and interaction term 6 (4Cyl*MfgDate) are much larger than
a typical cutoff value of 0.05, indicating these terms are not significant. You
could choose to omit these terms and pool their effects into the error term.
The output terms variable returns a matrix of codes, each of which is a bit
pattern representing a term. You can omit terms from the model by deleting
their entries from terms and running anovan again, this time supplying the
resulting vector as the model argument.
terms([4 6],:) = []
terms =
1 0 0
0 1 0
0 0 1
1 0 1
anovan(MPG,{org cyl4 when},terms,3,varnames)
ans =
1.0e-003 *
8-18
ANOVA
0.0000
0
0
0.1140
Now you have a more parsimonious model indicating that the mileage of
these cars seems to be related to all three factors, and that the effect of the
manufacturing date depends on where the car was made.
ANOVA with Random Effects
• “Introduction” on page 8-19
• “Setting Up the Model” on page 8-20
• “Fitting a Random Effects Model” on page 8-21
• “F Statistics for Models with Random Effects” on page 8-22
• “Variance Components” on page 8-24
Introduction. In an ordinary ANOVA model, each grouping variable
represents a fixed factor. The levels of that factor are a fixed set of values.
Your goal is to determine whether different factor levels lead to different
response values. This section presents an example that shows how to use
anovan to fit models where a factor’s levels represent a random selection from
a larger (infinite) set of possible levels.
8-19
8 Analysis of Variance
Setting Up the Model. To set up the example, first load the data, which is
stored in a 6-by-3 matrix, mileage.
load mileage
The anova2 function works only with balanced data, and it infers the values
of the grouping variables from the row and column numbers of the input
matrix. The anovan function, on the other hand, requires you to explicitly
create vectors of grouping variable values. To create these vectors, do the
following steps:
1 Create an array indicating the factory for each value in mileage. This
array is 1 for the first column, 2 for the second, and 3 for the third.
factory = repmat(1:3,6,1);
2 Create an array indicating the car model for each mileage value. This array
is 1 for the first three rows of mileage, and 2 for the remaining three rows.
carmod = [ones(3,3); 2*ones(3,3)];
3 Turn these matrices into vectors and display them.
mileage = mileage(:);
factory = factory(:);
carmod = carmod(:);
[mileage factory carmod]
ans =
33.3000 1.0000 1.0000
33.4000 1.0000 1.0000
32.9000 1.0000 1.0000
32.6000 1.0000 2.0000
32.5000 1.0000 2.0000
33.0000 1.0000 2.0000
34.5000 2.0000 1.0000
34.8000 2.0000 1.0000
33.8000 2.0000 1.0000
33.4000 2.0000 2.0000
33.7000 2.0000 2.0000
8-20
ANOVA
33.9000 2.0000 2.0000
37.4000 3.0000 1.0000
36.8000 3.0000 1.0000
37.6000 3.0000 1.0000
36.6000 3.0000 2.0000
37.0000 3.0000 2.0000
36.7000 3.0000 2.0000
Fitting a Random Effects Model. Continuing the example from the
preceding section, suppose you are studying a few factories but you want
information about what would happen if you build these same car models in
a different factory—either one that you already have or another that you
might construct. To get this information, fit the analysis of variance model,
specifying a model that includes an interaction term and that the factory
factor is random.
[pvals,tbl,stats] = anovan(mileage, {factory carmod}, ...
'model',2, 'random',1,'varnames',{'Factory' 'Car Model'});
In the fixed effects version of this fit, which you get by omitting the inputs
'random',1 in the preceding code, the effect of car model is significant, with a
p value of 0.0039. But in this example, which takes into account the random
variation of the effect of the variable 'Car Model' from one factory to another,
the effect is still significant, but with a higher p value of 0.0136.
8-21
8 Analysis of Variance
F Statistics for Models with Random Effects. The F statistic in a model
having random effects is defined differently than in a model having all fixed
effects. In the fixed effects model, you compute the F statistic for any term by
taking the ratio of the mean square for that term with the mean square for
error. In a random effects model, however, some F statistics use a different
mean square in the denominator.
In the example described in “Setting Up the Model” on page 8-20, the effect
of the variable 'Factory' could vary across car models. In this case, the
interaction mean square takes the place of the error mean square in the F
statistic. The F statistic for factory is:
F = 1.445 / 0.02
F =
72.2500
The degrees of freedom for the statistic are the degrees of freedom for the
numerator (1) and denominator (2) mean squares. Therefore the p value
for the statistic is:
pval = 1 - fcdf(F,1,2)
pval =
0.0136
With random effects, the expected value of each mean square depends not only
on the variance of the error term, but also on the variances contributed by
the random effects. You can see these dependencies by writing the expected
values as linear combinations of contributions from the various model terms.
To find the coefficients of these linear combinations, enter stats.ems, which
returns the ems field of the stats structure:
stats.ems
ans =
6.0000 0.0000 3.0000 1.0000
0.0000 9.0000 3.0000 1.0000
8-22
ANOVA
0.0000 0.0000 3.0000 1.0000
0 0 0 1.0000
To see text representations of the linear combinations, enter
stats.txtems
ans =
'6*V(Factory)+3*V(Factory*Car Model)+V(Error)'
'9*Q(Car Model)+3*V(Factory*Car Model)+V(Error)'
'3*V(Factory*Car Model)+V(Error)'
'V(Error)'
The expected value for the mean square due to car model (second term)
includes contributions from a quadratic function of the car model effects, plus
three times the variance of the interaction term’s effect, plus the variance
of the error term. Notice that if the car model effects were all zero, the
expression would reduce to the expected mean square for the third term (the
interaction term). That is why the F statistic for the car model effect uses the
interaction mean square in the denominator.
In some cases there is no single term whose expected value matches the one
required for the denominator of theFstatistic. In that case, the denominator is
a linear combination of mean squares. The stats structure contains fields
giving the definitions of the denominators for each F statistic. The txtdenom
field, stats.txtdenom, gives a text representation, and the denom field gives
a matrix that defines a linear combination of the variances of terms in the
model. For balanced models like this one, the denom matrix, stats.denom,
contains zeros and ones, because the denominator is just a single term’s mean
square:
stats.txtdenom
ans =
'MS(Factory*Car Model)'
'MS(Factory*Car Model)'
'MS(Error)'
stats.denom
8-23
8 Analysis of Variance
ans =
-0.0000 1.0000 0.0000
0.0000 1.0000 -0.0000
0.0000 0 1.0000
Variance Components. For the model described in “Setting Up the Model”
on page 8-20, consider the mileage for a particular car of a particular model
made at a random factory. The variance of that car is the sum of components,
or contributions, one from each of the random terms.
stats.rtnames
ans =
'Factory'
'Factory*Car Model'
'Error'
You do not know those variances, but you can estimate them from the data.
Recall that the ems field of the stats structure expresses the expected value
of each term’s mean square as a linear combination of unknown variances for
random terms, and unknown quadratic forms for fixed terms. If you take
the expected mean square expressions for the random terms, and equate
those expected values to the computed mean squares, you get a system of
equations that you can solve for the unknown variances. These solutions
are the variance component estimates. The varest field contains a variance
component estimate for each term. The rtnames field contains the names
of the random terms.
stats.varest
ans =
4.4426
-0.0313
0.1139
Under some conditions, the variability attributed to a term is unusually low,
and that term’s variance component estimate is negative. In those cases it
8-24
ANOVA
is common to set the estimate to zero, which you might do, for example, to
create a bar graph of the components.
bar(max(0,stats.varest))
set(gca,'xtick',1:3,'xticklabel',stats.rtnames)
You can also compute confidence bounds for the variance estimate. The
anovan function does this by computing confidence bounds for the variance
expected mean squares, and finding lower and upper limits on each variance
component containing all of these bounds. This procedure leads to a set
of bounds that is conservative for balanced data. (That is, 95% confidence
bounds will have a probability of at least 95% of containing the true variances
if the number of observations for each combination of grouping variables
is the same.) For unbalanced data, these are approximations that are not
guaranteed to be conservative.
[{'Term' 'Estimate' 'Lower' 'Upper'};
stats.rtnames, num2cell([stats.varest stats.varci])]
ans =
'Term' 'Estimate' 'Lower' 'Upper'
8-25
8 Analysis of Variance
'Factory' [ 4.4426] [1.0736] [175.6038]
'Factory*Car Model' [ -0.0313] [ NaN] [ NaN]
'Error' [ 0.1139] [0.0586] [ 0.3103]
Other ANOVA Models
The anovan function also has arguments that enable you to specify two other
types of model terms. First, the 'nested' argument specifies a matrix that
indicates which factors are nested within other factors. A nested factor is one
that takes different values within each level its nested factor.
For example, the mileage data from the previous section assumed that the
two car models produced in each factory were the same. Suppose instead,
each factory produced two distinct car models for a total of six car models, and
we numbered them 1 and 2 for each factory for convenience. Then, the car
model is nested in factory. A more accurate and less ambiguous numbering of
car model would be as follows:
Factory Car Model
1 1
1 2
2 3
2 4
3 5
3 6
However, it is common with nested models to number the nested factor the
same way in each nested factor.
Second, the 'continuous' argument specifies that some factors are to be
treated as continuous variables. The remaining factors are categorical
variables. Although the anovan function can fit models with multiple
continuous and categorical predictors, the simplest model that combines one
predictor of each type is known as an analysis of covariance model. The next
section describes a specialized tool for fitting this model.
8-26
ANOVA
Analysis of Covariance
• “Introduction” on page 8-27
• “Analysis of Covariance Tool” on page 8-27
• “Confidence Bounds” on page 8-32
• “Multiple Comparisons” on page 8-34
Introduction
Analysis of covariance is a technique for analyzing grouped data having a
response (y, the variable to be predicted) and a predictor (x, the variable
used to do the prediction). Using analysis of covariance, you can model y as
a linear function of x, with the coefficients of the line possibly varying from
group to group.
Analysis of Covariance Tool
The aoctool function opens an interactive graphical environment for fitting
and prediction with analysis of covariance (ANOCOVA) models. It fits the
following models for the ith group:
Same mean y=α+ε
Separate means y = (α + αi) + ε
Same line y = α + βx + ε
Parallel lines y = (α + αi) + βx + ε
Separate lines y = (α + αi) + (β + βi)x + ε
For example, in the parallel lines model the intercept varies from one group
to the next, but the slope is the same for each group. In the same mean
model, there is a common intercept and no slope. In order to make the group
coefficients well determined, the tool imposes the constraints
∑ j = ∑ j = 0
The following steps describe the use of aoctool.
8-27
8 Analysis of Variance
1 Load the data. The Statistics Toolbox data set carsmall.mat contains
information on cars from the years 1970, 1976, and 1982. This example
studies the relationship between the weight of a car and its mileage,
and whether this relationship has changed over the years. To start the
demonstration, load the data set.
load carsmall
The Workspace Browser shows the variables in the data set.
You can also use aoctool with your own data.
2 Start the tool. The following command calls aoctool to fit a separate line
to the column vectors Weight and MPG for each of the three model group
defined in Model_Year. The initial fit models the y variable, MPG, as a linear
function of the x variable, Weight.
[h,atab,ctab,stats] = aoctool(Weight,MPG,Model_Year);
8-28
ANOVA
See the aoctool function reference page for detailed information about
calling aoctool.
3 Examine the output. The graphical output consists of a main window
with a plot, a table of coefficient estimates, and an analysis of variance
table. In the plot, each Model_Year group has a separate line. The data
points for each group are coded with the same color and symbol, and the fit
for each group has the same color as the data points.
8-29
8 Analysis of Variance
The coefficients of the three lines appear in the figure titled ANOCOVA
Coefficients. You can see that the slopes are roughly –0.0078, with a small
deviation for each group:
• Model year 1970: y = (45.9798 – 8.5805) + (–0.0078 + 0.002)x + ε
• Model year 1976: y = (45.9798 – 3.8902) + (–0.0078 + 0.0011)x + ε
• Model year 1982: y = (45.9798 + 12.4707) + (–0.0078 – 0.0031)x + ε
Because the three fitted lines have slopes that are roughly similar, you may
wonder if they really are the same. The Model_Year*Weight interaction
expresses the difference in slopes, and the ANOVA table shows a test for
the significance of this term. With an F statistic of 5.23 and a p value of
0.0072, the slopes are significantly different.
8-30
ANOVA
4 Constrain the slopes to be the same. To examine the fits when the
slopes are constrained to be the same, return to the ANOCOVA Prediction
Plot window and use the Model pop-up menu to select a Parallel Lines
model. The window updates to show the following graph.
Though this fit looks reasonable, it is significantly worse than the Separate
Lines model. Use the Model pop-up menu again to return to the original
model.
8-31
8 Analysis of Variance
Confidence Bounds
The example in “Analysis of Covariance Tool” on page 8-27 provides estimates
of the relationship between MPG and Weight for each Model_Year, but how
accurate are these estimates? To find out, you can superimpose confidence
bounds on the fits by examining them one group at a time.
1 In the Model_Year menu at the lower right of the figure, change the
setting from All Groups to 82. The data and fits for the other groups are
dimmed, and confidence bounds appear around the 82 fit.
8-32
ANOVA
The dashed lines form an envelope around the fitted line for model year 82.
Under the assumption that the true relationship is linear, these bounds
provide a 95% confidence region for the true line. Note that the fits for the
other model years are well outside these confidence bounds for Weight
values between 2000 and 3000.
2 Sometimes it is more valuable to be able to predict the response value for
a new observation, not just estimate the average response value. Use the
aoctool function Bounds menu to change the definition of the confidence
bounds from Line to Observation. The resulting wider intervals reflect
the uncertainty in the parameter estimates as well as the randomness
of a new observation.
8-33
8 Analysis of Variance
Like the polytool function, the aoctool function has cross hairs that you
can use to manipulate the Weight and watch the estimate and confidence
bounds along the y-axis update. These values appear only when a single
group is selected, not when All Groups is selected.
Multiple Comparisons
You can perform a multiple comparison test by using the stats output
structure from aoctool as input to the multcompare function. The
multcompare function can test either slopes, intercepts, or population
marginal means (the predicted MPG of the mean weight for each group). The
example in “Analysis of Covariance Tool” on page 8-27 shows that the slopes
are not all the same, but could it be that two are the same and only the other
one is different? You can test that hypothesis.
multcompare(stats,0.05,'on','','s')
ans =
1.0000 2.0000 -0.0012 0.0008 0.0029
1.0000 3.0000 0.0013 0.0051 0.0088
2.0000 3.0000 0.0005 0.0042 0.0079
This matrix shows that the estimated difference between the intercepts of
groups 1 and 2 (1970 and 1976) is 0.0008, and a confidence interval for the
difference is [–0.0012, 0.0029]. There is no significant difference between the
two. There are significant differences, however, between the intercept for
1982 and each of the other two. The graph shows the same information.
8-34
ANOVA
Note that the stats structure was created in the initial call to the aoctool
function, so it is based on the initial model fit (typically a separate-lines
model). If you change the model interactively and want to base your multiple
comparisons on the new model, you need to run aoctool again to get another
stats structure, this time specifying your new model as the initial model.
Nonparametric Methods
• “Introduction” on page 8-36
8-35
8 Analysis of Variance
• “Kruskal-Wallis Test” on page 8-36
• “Friedman’s Test” on page 8-37
Introduction
Statistics Toolbox functions include nonparametric versions of one-way and
two-way analysis of variance. Unlike classical tests, nonparametric tests
make only mild assumptions about the data, and are appropriate when the
distribution of the data is non-normal. On the other hand, they are less
powerful than classical methods for normally distributed data.
Both of the nonparametric functions described here will return a stats
structure that can be used as an input to the multcompare function for
multiple comparisons.
Kruskal-Wallis Test
The example “Example: One-Way ANOVA” on page 8-4 uses one-way
analysis of variance to determine if the bacteria counts of milk varied from
shipment to shipment. The one-way analysis rests on the assumption that
the measurements are independent, and that each has a normal distribution
with a common variance and with a mean that was constant in each column.
You can conclude that the column means were not all the same. The following
example repeats that analysis using a nonparametric procedure.
The Kruskal-Wallis test is a nonparametric version of one-way analysis of
variance. The assumption behind this test is that the measurements come
from a continuous distribution, but not necessarily a normal distribution. The
test is based on an analysis of variance using the ranks of the data values, not
the data values themselves. Output includes a table similar to an ANOVA
table, and a box plot.
You can run this test as follows:
load hogg
p = kruskalwallis(hogg)
p =
0.0020
8-36
ANOVA
The low p value means the Kruskal-Wallis test results agree with the one-way
analysis of variance results.
Friedman’s Test
“Example: Two-Way ANOVA” on page 8-10 uses two-way analysis of variance
to study the effect of car model and factory on car mileage. The example
tests whether either of these factors has a significant effect on mileage, and
whether there is an interaction between these factors. The conclusion of
the example is there is no interaction, but that each individual factor has
a significant effect. The next example examines whether a nonparametric
analysis leads to the same conclusion.
Friedman’s test is a nonparametric test for data having a two-way layout (data
grouped by two categorical factors). Unlike two-way analysis of variance,
Friedman’s test does not treat the two factors symmetrically and it does not
test for an interaction between them. Instead, it is a test for whether the
columns are different after adjusting for possible row differences. The test is
based on an analysis of variance using the ranks of the data across categories
of the row factor. Output includes a table similar to an ANOVA table.
You can run Friedman’s test as follows.
load mileage
p = friedman(mileage,3)
p =
7.4659e-004
Recall the classical analysis of variance gave a p value to test column effects,
row effects, and interaction effects. This p value is for column effects. Using
either this p value or the p value from ANOVA (p < 0.0001), you conclude that
there are significant column effects.
In order to test for row effects, you need to rearrange the data to swap the
roles of the rows in columns. For a data matrix x with no replications, you
could simply transpose the data and type
p = friedman(x')
With replicated data it is slightly more complicated. A simple way is to
transform the matrix into a three-dimensional array with the first dimension
8-37
8 Analysis of Variance
representing the replicates, swapping the other two dimensions, and restoring
the two-dimensional shape.
x = reshape(mileage, [3 2 3]);
x = permute(x,[1 3 2]);
x = reshape(x,[9 2])
x =
33.3000 32.6000
33.4000 32.5000
32.9000 33.0000
34.5000 33.4000
34.8000 33.7000
33.8000 33.9000
37.4000 36.6000
36.8000 37.0000
37.6000 36.7000
friedman(x,3)
ans =
0.0082
Again, the conclusion is similar to that of the classical analysis of variance.
Both this p value and the one from ANOVA (p = 0.0039) lead you to conclude
that there are significant row effects.
You cannot use Friedman’s test to test for interactions between the row and
column factors.
8-38
MANOVA
MANOVA
In this section...
“Introduction” on page 8-39
“ANOVA with Multiple Responses” on page 8-39
Introduction
The analysis of variance technique in “Example: One-Way ANOVA” on
page 8-4 takes a set of grouped data and determine whether the mean of a
variable differs significantly among groups. Often there are multiple response
variables, and you are interested in determining whether the entire set of
means is different from one group to the next. There is a multivariate version
of analysis of variance that can address the problem.
ANOVA with Multiple Responses
The carsmall data set has measurements on a variety of car models from
the years 1970, 1976, and 1982. Suppose you are interested in whether the
characteristics of the cars have changed over time.
First, load the data.
load carsmall
whos
Name Size Bytes Class
Acceleration 100x1 800 double array
Cylinders 100x1 800 double array
Displacement 100x1 800 double array
Horsepower 100x1 800 double array
MPG 100x1 800 double array
Model 100x36 7200 char array
Model_Year 100x1 800 double array
Origin 100x7 1400 char array
Weight 100x1 800 double array
Four of these variables (Acceleration, Displacement, Horsepower, and
MPG) are continuous measurements on individual car models. The variable
8-39
8 Analysis of Variance
Model_Year indicates the year in which the car was made. You can create a
grouped plot matrix of these variables using the gplotmatrix function.
x = [MPG Horsepower Displacement Weight];
gplotmatrix(x,[],Model_Year,[],'+xo')
(When the second argument of gplotmatrix is empty, the function graphs
the columns of the x argument against each other, and places histograms
along the diagonals. The empty fourth argument produces a graph with the
default colors. The fifth argument controls the symbols used to distinguish
between groups.)
It appears the cars do differ from year to year. The upper right plot, for
example, is a graph of MPG versus Weight. The 1982 cars appear to have
higher mileage than the older cars, and they appear to weigh less on average.
But as a group, are the three years significantly different from one another?
The manova1 function can answer that question.
[d,p,stats] = manova1(x,Model_Year)
8-40
MANOVA
d =
2
p =
1.0e-006 *
0
0.1141
stats =
W: [4x4 double]
B: [4x4 double]
T: [4x4 double]
dfW: 90
dfB: 2
dfT: 92
lambda: [2x1 double]
chisq: [2x1 double]
chisqdf: [2x1 double]
eigenval: [4x1 double]
eigenvec: [4x4 double]
canon: [100x4 double]
mdist: [100x1 double]
gmdist: [3x3 double]
The manova1 function produces three outputs:
• The first output, d, is an estimate of the dimension of the group means. If
the means were all the same, the dimension would be 0, indicating that the
means are at the same point. If the means differed but fell along a line,
the dimension would be 1. In the example the dimension is 2, indicating
that the group means fall in a plane but not along a line. This is the largest
possible dimension for the means of three groups.
• The second output, p, is a vector of p-values for a sequence of tests. The
first p value tests whether the dimension is 0, the next whether the
dimension is 1, and so on. In this case both p-values are small. That’s
why the estimated dimension is 2.
• The third output, stats, is a structure containing several fields, described
in the following section.
8-41
8 Analysis of Variance
The Fields of the stats Structure
The W, B, and T fields are matrix analogs to the within, between, and total sums
of squares in ordinary one-way analysis of variance. The next three fields are
the degrees of freedom for these matrices. Fields lambda, chisq, and chisqdf
are the ingredients of the test for the dimensionality of the group means. (The
p-values for these tests are the first output argument of manova1.)
The next three fields are used to do a canonical analysis. Recall that in
principal components analysis (“Principal Component Analysis” on page
10-31) you look for the combination of the original variables that has the
largest possible variation. In multivariate analysis of variance, you instead
look for the linear combination of the original variables that has the largest
separation between groups. It is the single variable that would give the most
significant result in a univariate one-way analysis of variance. Having found
that combination, you next look for the combination with the second highest
separation, and so on.
The eigenvec field is a matrix that defines the coefficients of the linear
combinations of the original variables. The eigenval field is a vector
measuring the ratio of the between-group variance to the within-group
variance for the corresponding linear combination. The canon field is a matrix
of the canonical variable values. Each column is a linear combination of the
mean-centered original variables, using coefficients from the eigenvec matrix.
A grouped scatter plot of the first two canonical variables shows more
separation between groups then a grouped scatter plot of any pair of original
variables. In this example it shows three clouds of points, overlapping but
with distinct centers. One point in the bottom right sits apart from the others.
By using the gname function, you can see that this is the 20th point.
c1 = stats.canon(:,1);
c2 = stats.canon(:,2);
gscatter(c2,c1,Model_Year,[],'oxs')
gname
8-42
MANOVA
Roughly speaking, the first canonical variable, c1, separates the 1982 cars
(which have high values of c1) from the older cars. The second canonical
variable, c2, reveals some separation between the 1970 and 1976 cars.
The final two fields of the stats structure are Mahalanobis distances. The
mdist field measures the distance from each point to its group mean. Points
with large values may be outliers. In this data set, the largest outlier is the
one in the scatter plot, the Buick Estate station wagon. (Note that you could
have supplied the model name to the gname function above if you wanted to
label the point with its model name rather than its row number.)
max(stats.mdist)
ans =
31.5273
find(stats.mdist == ans)
ans =
8-43
8 Analysis of Variance
20
Model(20,:)
ans =
buick_estate_wagon_(sw)
The gmdist field measures the distances between each pair of group means.
The following commands examine the group means and their distances:
grpstats(x, Model_Year)
ans =
1.0e+003 *
0.0177 0.1489 0.2869 3.4413
0.0216 0.1011 0.1978 3.0787
0.0317 0.0815 0.1289 2.4535
stats.gmdist
ans =
0 3.8277 11.1106
3.8277 0 6.1374
11.1106 6.1374 0
As might be expected, the multivariate distance between the extreme years
1970 and 1982 (11.1) is larger than the difference between more closely
spaced years (3.8 and 6.1). This is consistent with the scatter plots, where the
points seem to follow a progression as the year changes from 1970 through
1976 to 1982. If you had more groups, you might find it instructive to use
the manovacluster function to draw a diagram that presents clusters of the
groups, formed using the distances between their means.
8-44
9
Regression Analysis
• “Introduction” on page 9-2
• “Linear Regression” on page 9-3
• “Nonlinear Regression” on page 9-58
9 Regression Analysis
Introduction
Regression is the process of fitting models to data. The process depends on the
model. If a model is parametric, regression estimates the parameters from the
data. If a model is linear in the parameters, estimation is based on methods
from linear algebra that minimize the norm of a residual vector. If a model
is nonlinear in the parameters, estimation is based on search methods from
optimization that minimize the norm of a residual vector. Nonparametric
models, like “Regression Trees” on page 9-94, use methods all their own.
This chapter considers data and models with continuous predictors and
responses. Categorical predictors are the subject of Chapter 8, “Analysis
of Variance”. Categorical responses are the subject of Chapter 12,
“Classification”.
9-2
Linear Regression
Linear Regression
In this section...
“Linear Regression Models” on page 9-3
“Multiple Linear Regression” on page 9-8
“Robust Regression” on page 9-14
“Stepwise Regression” on page 9-19
“Ridge Regression” on page 9-29
“Partial Least Squares” on page 9-32
“Polynomial Models” on page 9-37
“Response Surface Models” on page 9-45
“Generalized Linear Models” on page 9-52
“Multivariate Regression” on page 9-57
Linear Regression Models
In statistics, linear regression models often take the form of something like
this:
y = 0 + 1 x1 + 2 x2 + 3 x1 x2 + 4 x12 + 5 x22 +
Here a response variable y is modeled as a combination of constant, linear,
interaction, and quadratic terms formed from two predictor variables x1 and
x2. Uncontrolled factors and experimental errors are modeled by ε. Given data
on x1, x2, and y, regression estimates the model parameters βj (j = 1, ..., 5).
More general linear regression models represent the relationship between a
continuous response y and a continuous or categorical predictor x in the form:
y = 1 f1 ( x) + ... + p f p ( x) +
The response is modeled as a linear combination of (not necessarily linear)
functions of the predictor, plus a random error ε. The expressions fj(x) (j = 1,
..., p) are the terms of the model. The βj (j = 1, ..., p) are the coefficients. Errors
9-3
9 Regression Analysis
ε are assumed to be uncorrelated and distributed with mean 0 and constant
(but unknown) variance.
Examples of linear regression models with a scalar predictor variable x
include:
• Linear additive (straight-line) models — Terms are f1(x) = 1 and f2(x) = x.
• Polynomial models — Terms are f1(x) = 1, f2(x) = x, …, fp(x) = xp–1.
• Chebyshev orthogonal polynomial models — Terms are f1(x) = 1, f2(x) = x,
…, fp(x) = 2xfp–1(x) – fp–2(x).
• Fourier trigonometric polynomial models — Terms are f1(x) = 1/2 and sines
and cosines of different frequencies.
Examples of linear regression models with a vector of predictor variables x
= (x1, ..., xN) include:
• Linear additive (hyperplane) models — Terms are f1(x) = 1 and fk+1(x) =
xk (k = 1, ..., N).
• Pairwise interaction models — Terms are linear additive terms plus gk1k2(x)
= xk1xk2 (k1, k2 = 1, ..., N, k1 ≠ k2).
• Quadratic models — Terms are pairwise interaction terms plus hk(x) =
xk2 (k = 1, ..., N).
• Pure quadratic models — Terms are quadratic terms minus the gk1k2(x)
terms.
Whether or not the predictor x is a vector of predictor variables, multivariate
regression refers to the case where the response y = (y1, ..., yM) is a vector of
M response variables. See “Multivariate Regression” on page 9-57 for more
on multivariate regression models.
Given n independent observations (x1, y1), …, (xn, yn) of the predictor x and the
response y, the linear regression model becomes an n-by-p system of equations:
9-4
Linear Regression
⎛ y1 ⎞ ⎛ f1 ( x1 ) f p ( x1 ) ⎞ ⎛ 1 ⎞ ⎛ 1 ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ + ⎜ ⎟
⎜ y ⎟ ⎜ f (x ) … f p ( xn ) ⎟⎠ ⎜⎝ p ⎟⎠ ⎜⎝ n ⎟⎠
⎝ n⎠ ⎝ 1 n
y X
X is the design matrix of the system. The columns of X are the terms of the
model evaluated at the predictors. To fit the model to the data, the system
must be solved for the p coefficient values in β = (β1, …, βp)T.
The MATLAB backslash operator \ (mldivide) solves systems of linear
equations. Ignoring the unknown error ε, MATLAB estimates model
coefficients in y = Xβ using
betahat = X\y
where X is the design matrix and y is the vector of observed responses.
MATLAB returns the least-squares solution to the system; betahat minimizes
the norm of the residual vector y-X*beta over all beta. If the system is
consistent, the norm is 0 and the solution is exact. In this case, the regression
model interpolates the data. In more typical regression cases where n > p and
the system is overdetermined, the least-squares solution estimates model
coefficients obscured by the error ε.
The least-squares estimator betahat has several important statistical
properties. First, it is unbiased, with expected value β. Second, by the
Gauss-Markov theorem, it has minimum variance among all unbiased
estimators formed from linear combinations of the response data. Under the
additional assumption that ε is normally distributed, betahat is a maximum
likelihood estimator. The assumption also implies that the estimates
themselves are normally distributed, which is useful for computing confidence
intervals. Even without the assumption, by the Central Limit theorem, the
estimates have an approximate normal distribution if the sample size is large
enough.
Visualize the least-squares estimator as follows.
9-5
9 Regression Analysis
For betahat to minimize norm(y-X*beta), y-X*betahat must be
perpendicular to the column space of X, which contains all linear combinations
of the model terms. This requirement is summarized in the normal equations,
which express vanishing inner products between y-X*betahat and the
columns of X:
( )
X T y − X ˆ = 0
or
X T X ˆ = X T y
If X is n-by-p, the normal equations are a p-by-p square system with solution
betahat = inv(X'*X)*X'*y, where inv is the MATLAB inverse operator.
The matrix inv(X'*X)*X' is the pseudoinverse of X, computed by the
MATLAB function pinv.
The normal equations are often badly conditioned relative to the original
system y = Xβ (the coefficient estimates are much more sensitive to the model
error ε), so the MATLAB backslash operator avoids solving them directly.
9-6
Linear Regression
Instead, a QR (orthogonal, triangular) decomposition of X is used to create a
simpler, more stable triangular system:
X T X ˆ = XT y
(QR)T (QR) ˆ = (QR)T y
RT QT QRˆ = RT QT y
RT Rˆ = RT QT y
Rˆ = QT y
Statistics Toolbox functions like regress and regstats call the MATLAB
backslash operator to perform linear regression. The QR decomposition is also
used for efficient computation of confidence intervals.
Once betahat is computed, the model can be evaluated at the predictor data:
yhat = X*betahat
or
yhat = X*inv(X'*X)*X'*y
H = X*inv(X'*X)*X' is the hat matrix. It is a square, symmetric n-by-n
matrix determined by the predictor data. The diagonal elements H(i,i)
(i = 1, ..., n) give the leverage of the ith observation. Since yhat = H*y,
leverage values determine the influence of the observed response y(i) on
the predicted response yhat(i). For leverage values near 1, the predicted
response approximates the observed response. The Statistics Toolbox function
leverage computes leverage values from a QR decomposition of X.
Component residual values in y-yhat are useful for detecting failures in
model assumptions. Like the errors in ε, residuals have an expected value
of 0. Unlike the errors, however, residuals are correlated, with nonconstant
variance. Residuals may be “Studentized” (scaled by an estimate of their
standard deviation) for comparison. Studentized residuals are used by
Statistics Toolbox functions like regress and robustfit to identify outliers
in the data.
9-7
9 Regression Analysis
Multiple Linear Regression
• “Introduction” on page 9-8
• “Programmatic Multiple Linear Regression” on page 9-9
• “Interactive Multiple Linear Regression” on page 9-11
• “Tabulating Diagnostic Statistics” on page 9-13
Introduction
The system of linear equations
⎛ y1 ⎞ ⎛ f1 ( x1 ) f p ( x1 ) ⎞ ⎛ 1 ⎞ ⎛ 1 ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ + ⎜ ⎟
⎜ y ⎟ ⎜ f (x ) … f p ( xn ) ⎟⎠ ⎜⎝ p ⎟⎠ ⎜⎝ n ⎟⎠
⎝ n⎠ ⎝ 1 n
y X
in “Linear Regression Models” on page 9-3 expresses a response y as a linear
combination of model terms fj(x) (j = 1, ..., p) at each of the observations
(x1, y1), …, (xn, yn).
If the predictor x is multidimensional, so are the functions fj that form the
terms of the model. For example, if the predictor is x = (x1, x2), terms for the
model might include f1(x) = x1 (a linear term), f2(x) = x12 (a quadratic term),
and f3(x) = x1x2 (a pairwise interaction term). Typically, the function f(x) = 1 is
included among the fj, so that the design matrix X contains a column of 1s and
the model contains a constant (y-intercept) term.
Multiple linear regression models are useful for:
• Understanding which terms fj(x) have greatest effect on the response
(coefficients βj with greatest magnitude)
• Finding the direction of the effects (signs of the βj)
• Predicting unobserved values of the response (y(x) for new x)
The Statistics Toolbox functions regress and regstats are used for multiple
linear regression analysis.
9-8
Linear Regression
Programmatic Multiple Linear Regression
For example, the file moore.mat contains the 20-by-6 data matrix moore. The
first five columns are measurements of biochemical oxygen demand on five
predictor variables. The final column contains the observed responses. Use
regress to find coefficient estimates betahat for a linear additive model as
follows. Before using regress give the design matrix X1 a first column of 1s to
include a constant term in the model, betahat(1).
load moore
X1 = [ones(size(moore,1),1) moore(:,1:5)];
y = moore(:,6);
betahat = regress(y,X1)
betahat =
-2.1561
-0.0000
0.0013
0.0001
0.0079
0.0001
The MATLAB backslash (mldivide) operator, which regress calls, obtains
the same result:
betahat = X1\y
betahat =
-2.1561
-0.0000
0.0013
0.0001
0.0079
0.0001
The advantage of working with regress is that it allows for additional inputs
and outputs relevant to statistical analysis of the regression. For example:
alpha = 0.05;
[betahat,Ibeta,res,Ires,stats] = regress(y,X1,alpha);
returns not only the coefficient estimates in betahat, but also
9-9
9 Regression Analysis
• Ibeta — A p-by-2 matrix of 95% confidence intervals for the coefficient
estimates, using a 100*(1-alpha)% confidence level. The first column
contains lower confidence bounds for each of the p coefficient estimates; the
second column contains upper confidence bounds.
• res — An n-by-1 vector of residuals.
• Ires — An n-by-2 matrix of intervals that can be used to diagnose outliers.
If the interval Ires(i,:) for observation i does not contain zero, the
corresponding residual is larger than expected in 100*(1-alpha)% of new
observations, suggesting an outlier.
• stats — A 1-by-4 vector that contains, in order, the R2 statistic, the
F statistic and its p value, and an estimate of the error variance. The
statistics are computed assuming the model contains a constant term, and
are incorrect otherwise.
Visualize the residuals, in case (row number) order, with the rcoplot
function:
rcoplot(res,Ires)
9-10
Linear Regression
The interval around the first residual, shown in red when plotted, does not
contain zero. This indicates that the residual is larger than expected in 95%
of new observations, and suggests the data point is an outlier.
Outliers in regression appear for a variety of reasons:
1 If there is sufficient data, 5% of the residuals, by the definition of rint,
are too big.
2 If there is a systematic error in the model (that is, if the model is not
appropriate for generating the data under model assumptions), the mean
of the residuals is not zero.
3 If the errors in the model are not normally distributed, the distributions
of the residuals may be skewed or leptokurtic (with heavy tails and more
outliers).
When errors are normally distributed, Ires(i,:) is a confidence interval for
the mean of res(i) and checking if the interval contains zero is a test of the
null hypothesis that the residual has zero mean.
Interactive Multiple Linear Regression
The function regstats also performs multiple linear regression, but computes
more statistics than regress. By default, regstats automatically adds a first
column of 1s to the design matrix (necessary for computing the F statistic
and its p value), so a constant term should not be included explicitly as for
regress. For example:
X2 = moore(:,1:5);
stats = regstats(y,X2);
creates a structure stats with fields containing regression statistics. An
optional input argument allows you to specify which statistics are computed.
To interactively specify the computed statistics, call regstats without an
output argument. For example:
regstats(y,X2)
opens the following interface.
9-11
9 Regression Analysis
Select the check boxes corresponding to the statistics you want to compute and
click OK. Selected statistics are returned to the MATLAB workspace. Names
9-12
Linear Regression
of container variables for the statistics appear on the right-hand side of the
interface, where they can be changed to any valid MATLAB variable name.
Tabulating Diagnostic Statistics
The regstats function computes statistics that are typically used in
regression diagnostics. Statistics can be formatted into standard tabular
displays in a variety of ways. For example, the tstat field of the stats output
structure of regstats is itself a structure containing statistics related to the
estimated coefficients of the regression. Dataset arrays (see “Dataset Arrays”
on page 2-23) provide a natural tabular format for the information:
t = stats.tstat;
CoeffTable = dataset({t.beta,'Coef'},{t.se,'StdErr'}, ...
{t.t,'tStat'},{t.pval,'pVal'})
CoeffTable =
Coef StdErr tStat pVal
-2.1561 0.91349 -2.3603 0.0333
-9.0116e-006 0.00051835 -0.017385 0.98637
0.0013159 0.0012635 1.0415 0.31531
0.0001278 7.6902e-005 1.6618 0.11876
0.0078989 0.014 0.56421 0.58154
0.00014165 7.3749e-005 1.9208 0.075365
The MATLAB function fprintf gives you control over tabular formatting.
For example, the fstat field of the stats output structure of regstats is a
structure with statistics related to the analysis of variance (ANOVA) of the
regression. The following commands produce a standard regression ANOVA
table:
f = stats.fstat;
fprintf('\n')
fprintf('Regression ANOVA');
fprintf('\n\n')
fprintf('%6s','Source');
fprintf('%10s','df','SS','MS','F','P');
fprintf('\n')
fprintf('%6s','Regr');
9-13
9 Regression Analysis
fprintf('%10.4f',f.dfr,f.ssr,f.ssr/f.dfr,f.f,f.pval);
fprintf('\n')
fprintf('%6s','Resid');
fprintf('%10.4f',f.dfe,f.sse,f.sse/f.dfe);
fprintf('\n')
fprintf('%6s','Total');
fprintf('%10.4f',f.dfe+f.dfr,f.sse+f.ssr);
fprintf('\n')
The result looks like this:
Regression ANOVA
Source df SS MS F P
Regr 5.0000 4.1084 0.8217 11.9886 0.0001
Resid 14.0000 0.9595 0.0685
Total 19.0000 5.0679
Robust Regression
• “Introduction” on page 9-14
• “Programmatic Robust Regression” on page 9-15
• “Interactive Robust Regression” on page 9-16
Introduction
The models described in “Linear Regression Models” on page 9-3 are based on
certain assumptions, such as a normal distribution of errors in the observed
responses. If the distribution of errors is asymmetric or prone to outliers,
model assumptions are invalidated, and parameter estimates, confidence
intervals, and other computed statistics become unreliable. The Statistics
Toolbox function robustfit is useful in these cases. The function implements
a robust fitting method that is less sensitive than ordinary least squares to
large changes in small parts of the data.
Robust regression works by assigning a weight to each data point. Weighting
is done automatically and iteratively using a process called iteratively
9-14
Linear Regression
reweighted least squares. In the first iteration, each point is assigned equal
weight and model coefficients are estimated using ordinary least squares. At
subsequent iterations, weights are recomputed so that points farther from
model predictions in the previous iteration are given lower weight. Model
coefficients are then recomputed using weighted least squares. The process
continues until the values of the coefficient estimates converge within a
specified tolerance.
Programmatic Robust Regression
The example in “Multiple Linear Regression” on page 9-8 shows an outlier
when ordinary least squares is used to model the response variable as a linear
combination of the five predictor variables. To determine the influence of the
outlier, compare the coefficient estimates computed by regress:
load moore
X1 = [ones(size(moore,1),1) moore(:,1:5)];
y = moore(:,6);
betahat = regress(y,X1)
betahat =
-2.1561
-0.0000
0.0013
0.0001
0.0079
0.0001
to those computed by robustfit:
X2 = moore(:,1:5);
robustbeta = robustfit(X2,y)
robustbeta =
-1.7516
0.0000
0.0009
0.0002
0.0060
0.0001
By default, robustfit automatically adds a first column of 1s to the design
matrix, so a constant term does not have to be included explicitly as for
9-15
9 Regression Analysis
regress. In addition, the order of inputs is reversed for robustfit and
regress.
To understand the difference in the coefficient estimates, look at the final
weights given to the data points in the robust fit:
[robustbeta,stats] = robustfit(X2,y);
stats.w'
ans =
Columns 1 through 5
0.0246 0.9986 0.9763 0.9323 0.9704
Columns 6 through 10
0.8597 0.9180 0.9992 0.9590 0.9649
Columns 11 through 15
0.9769 0.9868 0.9999 0.9976 0.8122
Columns 16 through 20
0.9733 0.9892 0.9988 0.8974 0.6774
The first data point has a very low weight compared to the other data points,
and so is effectively ignored in the robust regression.
Interactive Robust Regression
The robustdemo function shows the difference between ordinary least squares
and robust fitting for data with a single predictor. You can use data provided
with the demo, or you can supply your own data. The following steps show
you how to use robustdemo.
1 Start the demo. To begin using robustdemo with the built-in data, simply
type the function name:
robustdemo
9-16
Linear Regression
The resulting figure shows a scatter plot with two fitted lines. The red line
is the fit using ordinary least-squares regression. The green line is the
fit using robust regression. At the bottom of the figure are the equations
for the fitted lines, together with the estimated root mean squared errors
for each fit.
2 View leverages and robust weights. Right-click on any data point to
see its least-squares leverage and robust weight.
9-17
9 Regression Analysis
In the built-in data, the right-most point has a relatively high leverage of
0.35. The point exerts a large influence on the least-squares fit, but its
small robust weight shows that it is effectively excluded from the robust fit.
3 See how changes in the data affect the fits. With the left mouse
button, click and hold on any data point and drag it to a new location.
When you release the mouse button, the displays update.
9-18
Linear Regression
Bringing the right-most data point closer to the least-squares line makes
the two fitted lines nearly identical. The adjusted right-most data point
has significant weight in the robust fit.
Stepwise Regression
• “Introduction” on page 9-19
• “Programmatic Stepwise Regression” on page 9-21
• “Interactive Stepwise Regression” on page 9-27
Introduction
Multiple linear regression models, as described in “Multiple Linear
Regression” on page 9-8, are built from a potentially large number of
predictive terms. The number of interaction terms, for example, increases
exponentially with the number of predictor variables. If there is no theoretical
9-19
9 Regression Analysis
basis for choosing the form of a model, and no assessment of correlations
among terms, it is possible to include redundant terms in a model that confuse
the identification of significant effects.
Stepwise regression is a systematic method for adding and removing terms
from a multilinear model based on their statistical significance in a regression.
The method begins with an initial model and then compares the explanatory
power of incrementally larger and smaller models. At each step, the p value of
an F-statistic is computed to test models with and without a potential term. If
a term is not currently in the model, the null hypothesis is that the term would
have a zero coefficient if added to the model. If there is sufficient evidence to
reject the null hypothesis, the term is added to the model. Conversely, if a
term is currently in the model, the null hypothesis is that the term has a zero
coefficient. If there is insufficient evidence to reject the null hypothesis, the
term is removed from the model. The method proceeds as follows:
1 Fit the initial model.
2 If any terms not in the model have p-values less than an entrance tolerance
(that is, if it is unlikely that they would have zero coefficient if added to
the model), add the one with the smallest p value and repeat this step;
otherwise, go to step 3.
3 If any terms in the model have p-values greater than an exit tolerance (that
is, if it is unlikely that the hypothesis of a zero coefficient can be rejected),
remove the one with the largest p value and go to step 2; otherwise, end.
Depending on the terms included in the initial model and the order in which
terms are moved in and out, the method may build different models from the
same set of potential terms. The method terminates when no single step
improves the model. There is no guarantee, however, that a different initial
model or a different sequence of steps will not lead to a better fit. In this
sense, stepwise models are locally optimal, but may not be globally optimal.
Statistics Toolbox functions for stepwise regression are:
• stepwisefit — A function that proceeds automatically from a specified
initial model and entrance/exit tolerances
• stepwise — An interactive tool that allows you to explore individual steps
in the regression
9-20
Linear Regression
Programmatic Stepwise Regression
For example, load the data in hald.mat, which contains observations of the
heat of reaction of various cement mixtures:
load hald
whos
Name Size Bytes Class Attributes
Description 22x58 2552 char
hald 13x5 520 double
heat 13x1 104 double
ingredients 13x4 416 double
The response (heat) depends on the quantities of the four predictors (the
columns of ingredients).
Use stepwisefit to carry out the stepwise regression algorithm, beginning
with no terms in the model and using entrance/exit tolerances of 0.05/0.10
on the p-values:
stepwisefit(ingredients,heat,...
'penter',0.05,'premove',0.10);
Initial columns included: none
Step 1, added column 4, p=0.000576232
Step 2, added column 1, p=1.10528e-006
Final columns included: 1 4
'Coeff' 'Std.Err.' 'Status' 'P'
[ 1.4400] [ 0.1384] 'In' [1.1053e-006]
[ 0.4161] [ 0.1856] 'Out' [ 0.0517]
[-0.4100] [ 0.1992] 'Out' [ 0.0697]
[-0.6140] [ 0.0486] 'In' [1.8149e-007]
stepwisefit automatically includes an intercept term in the model, so you do
not add it explicitly to ingredients as you would for regress. For terms not
in the model, coefficient estimates and their standard errors are those that
result if the term is added.
The inmodel parameter is used to specify terms in an initial model:
initialModel = ...
[false true false false]; % Force in 2nd term
9-21
9 Regression Analysis
stepwisefit(ingredients,heat,...
'inmodel',initialModel,...
'penter',.05,'premove',0.10);
Initial columns included: 2
Step 1, added column 1, p=2.69221e-007
Final columns included: 1 2
'Coeff' 'Std.Err.' 'Status' 'P'
[ 1.4683] [ 0.1213] 'In' [2.6922e-007]
[ 0.6623] [ 0.0459] 'In' [5.0290e-008]
[ 0.2500] [ 0.1847] 'Out' [ 0.2089]
[-0.2365] [ 0.1733] 'Out' [ 0.2054]
The preceding two models, built from different initial models, use different
subsets of the predictive terms. Terms 2 and 4, swapped in the two models,
are highly correlated:
term2 = ingredients(:,2);
term4 = ingredients(:,4);
R = corrcoef(term2,term4)
R =
1.0000 -0.9730
-0.9730 1.0000
To compare the models, use the stats output of stepwisefit:
[betahat1,se1,pval1,inmodel1,stats1] = ...
stepwisefit(ingredients,heat,...
'penter',.05,'premove',0.10,...
'display','off');
[betahat2,se2,pval2,inmodel2,stats2] = ...
stepwisefit(ingredients,heat,...
'inmodel',initialModel,...
'penter',.05,'premove',0.10,...
'display','off');
RMSE1 = stats1.rmse
RMSE1 =
2.7343
RMSE2 = stats2.rmse
RMSE2 =
2.4063
9-22
Linear Regression
The second model has a lower Root Mean Square Error (RMSE).
An added variable plot is used to determine the unique effect of adding a new
term to a model. The plot shows the relationship between the part of the
response unexplained by terms already in the model and the part of the new
term unexplained by terms already in the model. The “unexplained” parts
are measured by the residuals of the respective regressions. A scatter of the
residuals from the two regressions forms the added variable plot.
For example, suppose you want to add term2 to a model that already contains
the single term term1. First, consider the ability of term2 alone to explain
the response:
load hald
term2 = ingredients(:,2);
[b2,Ib2,res2] = regress(heat,[ones(size(term2)) term2]);
scatter(term2,heat)
xlabel('Term 2')
ylabel('Heat')
hold on
x2 = 20:80;
y2 = b2(1) + b2(2)*x2;
plot(x2,y2,'r')
title('{\bf Response Explained by Term 2: Ignoring Term 1}')
9-23
9 Regression Analysis
Next, consider the following regressions involving the model term term1:
term1 = ingredients(:,1);
[b1,Ib1,res1] = regress(heat,[ones(size(term1)) term1]);
[b21,Ib21,res21] = regress(term2,[ones(size(term1)) term1]);
bres = regress(res1,[ones(size(res21)) res21]);
A scatter of the residuals res1 vs. the residuals res12 forms the added
variable plot:
figure
scatter(res21,res1)
xlabel('Residuals: Term 2 on Term 1')
ylabel('Residuals: Heat on Term 1')
hold on
9-24
Linear Regression
xres = -30:30;
yres = bres(1) + bres(2)*xres;
plot(xres,yres,'r')
title('{\bf Response Explained by Term 2: Adjusted for Term 1}')
Since the plot adjusted for term1 shows a stronger relationship (less variation
along the fitted line) than the plot ignoring term1, the two terms act jointly to
explain extra variation. In this case, adding term2 to a model consisting of
term1 would reduce the RMSE.
The Statistics Toolbox function addedvarplot produces added variable plots.
The previous plot is essentially the one produced by the following:
figure
9-25
9 Regression Analysis
inmodel = [true false false false];
addedvarplot(ingredients,heat,2,inmodel)
In addition to the scatter of residuals, the plot shows 95% confidence intervals
on predictions from the fitted line. The fitted line has intercept zero because,
under the assumptions outlined in “Linear Regression Models” on page 9-3,
both of the plotted variables have mean zero. The slope of the fitted line is the
coefficient that term2 would have if it was added to the model with term1.
The addevarplot function is useful for considering the unique effect of adding
a new term to an existing model with any number of terms.
9-26
Linear Regression
Interactive Stepwise Regression
The stepwise interface provides interactive features that allow you to
investigate individual steps in a stepwise regression, and to build models
from arbitrary subsets of the predictive terms. To open the interface with
data from hald.mat:
load hald
stepwise(ingredients,heat)
9-27
9 Regression Analysis
The upper left of the interface displays estimates of the coefficients for all
potential terms, with horizontal bars indicating 90% (colored) and 95% (grey)
confidence intervals. The red color indicates that, initially, the terms are not
in the model. Values displayed in the table are those that would result if
the terms were added to the model.
The middle portion of the interface displays summary statistics for the entire
model. These statistics are updated with each step.
The lower portion of the interface, Model History, displays the RMSE for
the model. The plot tracks the RMSE from step to step, so you can compare
the optimality of different models. Hover over the blue dots in the history to
see which terms were in the model at a particular step. Click on a blue dot
in the history to open a copy of the interface initialized with the terms in
the model at that step.
Initial models, as well as entrance/exit tolerances for the p-values of
F-statistics, are specified using additional input arguments to stepwise.
Defaults are an initial model with no terms, an entrance tolerance of 0.05,
and an exit tolerance of 0.10.
To center and scale the input data (compute z-scores) to improve conditioning
of the underlying least-squares problem, select Scale Inputs from the
Stepwise menu.
You proceed through a stepwise regression in one of two ways:
1 Click Next Step to select the recommended next step. The recommended
next step either adds the most significant term or removes the least
significant term. When the regression reaches a local minimum of RMSE,
the recommended next step is “Move no terms.” You can perform all of the
recommended steps at once by clicking All Steps.
2 Click a line in the plot or in the table to toggle the state of the corresponding
term. Clicking a red line, corresponding to a term not currently in the
model, adds the term to the model and changes the line to blue. Clicking
a blue line, corresponding to a term currently in the model, removes the
term from the model and changes the line to red.
9-28
Linear Regression
To call addedvarplot and produce an added variable plot from the stepwise
interface, select Added Variable Plot from the Stepwise menu. A list of
terms is displayed. Select the term you want to add, and then click OK.
Click Export to display a dialog box that allows you to select information
from the interface to save to the MATLAB workspace. Check the information
you want to export and, optionally, change the names of the workspace
variables to be created. Click OK to export the information.
Ridge Regression
• “Introduction” on page 9-29
• “Example: Ridge Regression” on page 9-30
Introduction
Coefficient estimates for the models described in “Multiple Linear Regression”
on page 9-8 rely on the independence of the model terms. When terms are
correlated and the columns of the design matrix X have an approximate
linear dependence, the matrix (XTX)–1 becomes close to singular. As a result,
the least-squares estimate
ˆ = ( X T X )−1 X T y
becomes highly sensitive to random errors in the observed response y,
producing a large variance. This situation of multicollinearity can arise, for
example, when data are collected without an experimental design.
Ridge regression addresses the problem by estimating regression coefficients
using
ˆ = ( X T X + kI )−1 X T y
where k is the ridge parameter and I is the identity matrix. Small positive
values of k improve the conditioning of the problem and reduce the variance
of the estimates. While biased, the reduced variance of ridge estimates
often result in a smaller mean square error when compared to least-squares
estimates.
9-29
9 Regression Analysis
The Statistics Toolbox function ridge carries out ridge regression.
Example: Ridge Regression
For example, load the data in acetylene.mat, with observations of the
predictor variables x1, x2, x3, and the response variable y:
load acetylene
Plot the predictor variables against each other:
subplot(1,3,1)
plot(x1,x2,'.')
xlabel('x1'); ylabel('x2'); grid on; axis square
subplot(1,3,2)
plot(x1,x3,'.')
xlabel('x1'); ylabel('x3'); grid on; axis square
subplot(1,3,3)
plot(x2,x3,'.')
xlabel('x2'); ylabel('x3'); grid on; axis square
Note the correlation between x1 and the other two predictor variables.
Use ridge and x2fx to compute coefficient estimates for a multilinear model
with interaction terms, for a range of ridge parameters:
X = [x1 x2 x3];
D = x2fx(X,'interaction');
9-30
Linear Regression
D(:,1) = []; % No constant term
k = 0:1e-5:5e-3;
betahat = ridge(y,D,k);
Plot the ridge trace:
figure
plot(k,betahat,'LineWidth',2)
ylim([-100 100])
grid on
xlabel('Ridge Parameter')
ylabel('Standardized Coefficient')
title('{\bf Ridge Trace}')
legend('x1','x2','x3','x1x2','x1x3','x2x3')
9-31
9 Regression Analysis
The estimates stabilize to the right of the plot. Note that the coefficient of
the x2x3 interaction term changes sign at a value of the ridge parameter
≈ 5 × 10–4.
Partial Least Squares
• “Introduction” on page 9-33
9-32
Linear Regression
• “Example: Partial Least Squares” on page 9-33
Introduction
Partial least-squares (PLS) regression is a technique used with data that
contain correlated predictor variables. This technique constructs new
predictor variables, known as components, as linear combinations of the
original predictor variables. PLS constructs these components while
considering the observed response values, leading to a parsimonious model
with reliable predictive power.
The technique is something of a cross between multiple linear regression
and principal component analysis:
• Multiple linear regression finds a combination of the predictors that best fit
a response.
• Principal component analysis finds combinations of the predictors with
large variance, reducing correlations. The technique makes no use of
response values.
• PLS finds combinations of the predictors that have a large covariance with
the response values.
PLS therefore combines information about the variances of both the predictors
and the responses, while also considering the correlations among them.
PLS shares characteristics with other regression and feature transformation
techniques. It is similar to ridge regression in that it is used in situations with
correlated predictors. It is similar to stepwise regression (or more general
feature selection techniques) in that it can be used to select a smaller set of
model terms. PLS differs from these methods, however, by transforming the
original predictor space into the new component space.
The Statistics Toolbox function plsregress carries out PLS regression.
Example: Partial Least Squares
For example, consider the data on biochemical oxygen demand in moore.mat,
padded with noisy versions of the predictors to introduce correlations:
load moore
9-33
9 Regression Analysis
y = moore(:,6); % Response
X0 = moore(:,1:5); % Original predictors
X1 = X0+10*randn(size(X0)); % Correlated predictors
X = [X0,X1];
Use plsregress to perform PLS regression with the same number of
components as predictors, then plot the percentage variance explained in the
response as a function of the number of components:
[XL,yl,XS,YS,beta,PCTVAR] = plsregress(X,y,10);
plot(1:10,cumsum(100*PCTVAR(2,:)),'-bo');
xlabel('Number of PLS components');
ylabel('Percent Variance Explained in y');
Choosing the number of components in a PLS model is a critical step. The plot
gives a rough indication, showing nearly 80% of the variance in y explained
9-34
Linear Regression
by the first component, with as many as five additional components making
significant contributions.
The following computes the six-component model:
[XL,yl,XS,YS,beta,PCTVAR,MSE,stats] = plsregress(X,y,6);
yfit = [ones(size(X,1),1) X]*beta;
plot(y,yfit,'o')
The scatter shows a reasonable correlation between fitted and observed
responses, and this is confirmed by the R2 statistic:
TSS = sum((y-mean(y)).^2);
RSS = sum((y-yfit).^2);
Rsquared = 1 - RSS/TSS
Rsquared =
0.8421
9-35
9 Regression Analysis
A plot of the weights of the ten predictors in each of the six components shows
that two of the components (the last two computed) explain the majority of
the variance in X:
plot(1:10,stats.W,'o-');
legend({'c1','c2','c3','c4','c5','c6'},'Location','NW')
xlabel('Predictor');
ylabel('Weight');
A plot of the mean-squared errors suggests that as few as two components
may provide an adequate model:
[axes,h1,h2] = plotyy(0:6,MSE(1,:),0:6,MSE(2,:));
set(h1,'Marker','o')
set(h2,'Marker','o')
legend('MSE Predictors','MSE Response')
xlabel('Number of Components')
9-36
Linear Regression
The calculation of mean-squared errors by plsregress is controlled by
optional parameter name/value pairs specifying cross-validation type and the
number of Monte Carlo repetitions.
Polynomial Models
• “Introduction” on page 9-37
• “Programmatic Polynomial Regression” on page 9-38
• “Interactive Polynomial Regression” on page 9-43
Introduction
Polynomial models are a special case of the linear models discussed in “Linear
Regression Models” on page 9-3. Polynomial models have the advantages of
being simple, familiar in their properties, and reasonably flexible for following
9-37
9 Regression Analysis
data trends. They are also robust with respect to changes in the location and
scale of the data (see “Conditioning Polynomial Fits” on page 9-41). However,
polynomial models may be poor predictors of new values. They oscillate
between data points, especially as the degree is increased to improve the fit.
Asymptotically, they follow power functions, leading to inaccuracies when
extrapolating other long-term trends. Choosing a polynomial model is often a
trade-off between a simple description of overall data trends and the accuracy
of predictions made from the model.
Programmatic Polynomial Regression
• “Functions for Polynomial Fitting” on page 9-38
• “Displaying Polynomial Fits” on page 9-40
• “Conditioning Polynomial Fits” on page 9-41
Functions for Polynomial Fitting. To fit polynomials to data, MATLAB
and Statistics Toolbox software offer a number of dedicated functions. The
MATLAB function polyfit computes least-squares coefficient estimates for
polynomials of arbitrary degree. For example:
x = 0:5; % x data
y = [2 1 4 4 3 2]; % y data
p = polyfit(x,y,3) % Degree 3 fit
p =
-0.1296 0.6865 -0.1759 1.6746
Polynomial coefficients in p are listed from highest to lowest degree, so p(x)
≈ –0.13 x3 + 0.69 x2 – 0.18 x + 1.67. For convenience, polyfit sets up the
Vandermonde design matrix (vander) and calls backslash (mldivide) to
perform the fit.
Once the coefficients of a polynomial are collected in a vector p, use the
MATLAB function polyval to evaluate the polynomial at arbitrary inputs.
For example, the following plots the data and the fit over a range of inputs:
plot(x,y,'ro','LineWidth',2) % Plot data
hold on
xfit = -1:0.01:6;
yfit = polyval(p,xfit);
9-38
Linear Regression
plot(xfit,yfit,'LineWidth',2) % Plot fit
ylim([0,5])
grid on
Use the MATLAB function roots to find the roots of p:
r = roots(p)
r =
5.4786
-0.0913 + 1.5328i
-0.0913 - 1.5328i
The MATLAB function poly solves the inverse problem, finding a polynomial
with specified roots. poly is the inverse of roots up to ordering, scaling, and
round-off error.
9-39
9 Regression Analysis
An optional output from polyfit is passed to polyval or to the Statistics
Toolbox function polyconf to compute prediction intervals for the fit.
For example, the following computes 95% prediction intervals for new
observations at each value of the predictor x:
[p,S] = polyfit(x,y,3);
[yhat,delta] = polyconf(p,x,S);
PI = [yhat-delta;yhat+delta]'
PI =
-5.3022 8.6514
-4.2068 8.3179
-2.9899 9.0534
-2.1963 9.8471
-2.6036 9.9211
-5.2229 8.7308
Optional input arguments to polyconf allow you to compute prediction
intervals for estimated values (yhat) as well as new observations, and to
compute the bounds simultaneously for all x instead of nonsimultaneously
(the default). The confidence level for the intervals can also be set.
Displaying Polynomial Fits. The documentation example function
polydemo combines the functions polyfit, polyval, roots, and polyconf to
produce a formatted display of data with a polynomial fit.
Note Statistics Toolbox documentation example files are located in the
\help\toolbox\stats\examples subdirectory of your MATLAB root folder
(matlabroot). This subdirectory is not on the MATLAB path at installation.
To use the files in this subdirectory, either add the subdirectory to the
MATLAB path (addpath) or make the subdirectory your current working
folder (cd).
For example, the following uses polydemo to produce a display of simulated
data with a quadratic trend, a fitted polynomial, and 95% prediction intervals
for new observations:
x = -5:5;
y = x.^2 - 5*x - 3 + 5*randn(size(x));
p = polydemo(x,y,2,0.05)
9-40
Linear Regression
p =
0.8107 -4.5054 -1.1862
polydemo calls the documentation example function polystr to convert the
coefficient vector p into a string for the polynomial expression displayed in the
figure title.
Conditioning Polynomial Fits. If x and y data are on very different
scales, polynomial fits may be badly conditioned, in the sense that coefficient
estimates are very sensitive to random errors in the data. For example,
using polyfit to estimate coefficients of a cubic fit to the U.S. census data in
census.mat produces the following warning:
load census
x = cdate;
9-41
9 Regression Analysis
y = pop;
p = polyfit(x,y,3);
Warning: Polynomial is badly conditioned.
Add points with distinct X values,
reduce the degree of the polynomial,
or try centering and scaling as
described in HELP POLYFIT.
The following implements the suggested centering and scaling, and
demonstrates the robustness of polynomial fits under these transformations:
plot(x,y,'ro') % Plot data
hold on
z = zscore(x); % Compute z-scores of x data
zfit = linspace(z(1),z(end),100);
pz = polyfit(z,y,3); % Compute conditioned fit
yfit = polyval(pz,zfit);
xfit = linspace(x(1),x(end),100);
plot(xfit,yfit,'b-') % Plot conditioned fit vs. x data
grid on
9-42
Linear Regression
Interactive Polynomial Regression
The functions polyfit, polyval, and polyconf are interactively applied to
data using two graphical interfaces for polynomial fitting:
• “The Basic Fitting Tool” on page 9-43
• “The Polynomial Fitting Tool” on page 9-44
The Basic Fitting Tool. The Basic Fitting Tool is a MATLAB interface,
discussed in “Interactive Fitting” in the MATLAB documentation. The tool
allows you to:
• Fit interpolants and polynomials of degree ≤ 10
• Plot residuals and compute their norm
• Interpolate or extrapolate values from the fit
9-43
9 Regression Analysis
• Save results to the MATLAB workspace
The Polynomial Fitting Tool. The Statistics Toolbox function polytool
opens the Polynomial Fitting Tool. For example, the following opens the
interface using simulated data with a quadratic trend and displays a fitted
polynomial with 95% prediction intervals for new observations:
x = -5:5;
y = x.^2 - 5*x - 3 + 5*randn(size(x));
polytool(x,y,2,0.05)
The tool allows you to:
9-44
Linear Regression
• Interactively change the degree of the fit. Change the value in the Degree
text box at the top of the figure.
• Evaluate the fit and the bounds using a movable crosshair. Click, hold, and
drag the crosshair to change its position.
• Export estimated coefficients, predicted values, prediction intervals, and
residuals to the MATLAB workspace. Click Export to a open a dialog box
with choices for exporting the data.
Options for the displayed bounds and the fitting method are available through
menu options at the top of the figure:
• The Bounds menu lets you choose between bounds on new observations
(the default) and bounds on estimated values. It also lets you choose
between nonsimultaneous (the default) and simultaneous bounds. See
polyconf for a description of these options.
• The Method menu lets you choose between ordinary least-squares
regression and robust regression, as described in “Robust Regression” on
page 9-14.
Response Surface Models
• “Introduction” on page 9-45
• “Programmatic Response Surface Methodology” on page 9-46
• “Interactive Response Surface Methodology” on page 9-51
Introduction
Polynomial models are generalized to any number of predictor variables xi (i
= 1, ..., N) as follows:
N N N
y( x) = a0 + ∑ ai xi + ∑ aij xi x j + ∑ aii xi2 + …
i=0 i< j i=0
The model includes, from left to right, an intercept, linear terms, quadratic
interaction terms, and squared terms. Higher order terms would follow, as
necessary.
9-45
9 Regression Analysis
Response surface models are multivariate polynomial models. They typically
arise in the design of experiments (see Chapter 14, “Design of Experiments”),
where they are used to determine a set of design variables that optimize a
response. Linear terms alone produce models with response surfaces that
are hyperplanes. The addition of interaction terms allows for warping of
the hyperplane. Squared terms produce the simplest models in which the
response surface has a maximum or minimum, and so an optimal response.
Response surface methodology (RSM) is the process of adjusting predictor
variables to move the response in a desired direction and, iteratively, to an
optimum. The method generally involves a combination of both computation
and visualization. The use of quadratic response surface models makes the
method much simpler than standard nonlinear techniques for determining
optimal designs.
Programmatic Response Surface Methodology
The file reaction.mat contains simulated data on the rate of a chemical
reaction:
load reaction
The variables include:
• rate — A 13-by-1 vector of observed reaction rates
• reactants — A 13-by-3 matrix of reactant concentrations
• xn — The names of the three reactants
• yn — The name of the response
In “Nonlinear Regression” on page 9-58, the nonlinear Hougen-Watson model
is fit to the data using nlinfit. However, there may be no theoretical basis
for choosing a particular model to fit the data. A quadratic response surface
model provides a simple way to determine combinations of reactants that
lead to high reaction rates.
As described in “Multiple Linear Regression” on page 9-8, the regress and
regstats functions fit linear models—including response surface models—to
data using a design matrix of model terms evaluated at predictor data. The
9-46
Linear Regression
x2fx function converts predictor data to design matrices for quadratic models.
The regstats function calls x2fx when instructed to do so.
For example, the following fits a quadratic response surface model to the
data in reaction.mat:
stats = regstats(rate,reactants,'quadratic','beta');
b = stats.beta; % Model coefficients
The 10-by-1 vector b contains, in order, a constant term and then the
coefficients for the model terms x1, x2, x3, x1x2, x1x3, x2x3, x12, x22, and x32, where
x1, x2, and x3 are the three columns of reactants. The order of coefficients for
quadratic models is described in the reference page for x2fx.
Since the model involves only three predictors, it is possible to visualize the
entire response surface using a color dimension for the reaction rate:
x1 = reactants(:,1);
x2 = reactants(:,2);
x3 = reactants(:,3);
xx1 = linspace(min(x1),max(x1),25);
xx2 = linspace(min(x2),max(x2),25);
xx3 = linspace(min(x3),max(x3),25);
[X1,X2,X3] = meshgrid(xx1,xx2,xx3);
RATE = b(1) + b(2)*X1 + b(3)*X2 + b(4)*X3 + ...
b(5)*X1.*X2 + b(6)*X1.*X3 + b(7)*X2.*X3 + ...
b(8)*X1.^2 + b(9)*X2.^2 + b(10)*X3.^2;
hmodel = scatter3(X1(:),X2(:),X3(:),5,RATE(:),'filled');
hold on
hdata = scatter3(x1,x2,x3,'ko','filled');
axis tight
xlabel(xn(1,:))
ylabel(xn(2,:))
zlabel(xn(3,:))
hbar = colorbar;
ylabel(hbar,yn);
title('{\bf Quadratic Response Surface Model}')
9-47
9 Regression Analysis
legend(hdata,'Data','Location','NE')
The plot show a general increase in model response, within the space of
the observed data, as the concentration of n-pentane increases and the
concentrations of hydrogen and isopentane decrease.
Before trying to determine optimal values of the predictors, perhaps by
collecting more data in the direction of increased reaction rate indicated by
the plot, it is helpful to evaluate the geometry of the response surface. If x
= (x1, x2, x3)T is the vector of predictors, and H is the matrix such that xTHx
gives the quadratic terms of the model, the model has a unique optimum if
and only if H is positive definite. For the data in this example, the model does
not have a unique optimum:
9-48
Linear Regression
H = [b(8),b(5)/2,b(6)/2; ...
b(5)/2,b(9),b(7)/2; ...
b(6)/2,b(7)/2,b(10)];
lambda = eig(H)
lambda =
1.0e-003 *
-0.1303
0.0412
0.4292
The negative eigenvalue shows a lack of positive definiteness. The saddle in
the model is visible if the range of the predictors in the plot (xx1, xx2, and
xx3) is expanded:
9-49
9 Regression Analysis
When the number of predictors makes it impossible to visualize the entire
response surface, 3-, 2-, and 1-dimensional slices provide local views. The
MATLAB function slice displays 2-dimensional contours of the data at fixed
values of the predictors:
delete(hmodel)
X2slice = 200; % Fix n-Pentane concentration
slice(X1,X2,X3,RATE,[],X2slice,[])
One-dimensional contours are displayed by the Response Surface Tool,
rstool, described in the next section.
9-50
Linear Regression
Interactive Response Surface Methodology
The Statistics Toolbox function rstool opens a GUI for interactively
investigating simultaneous one-dimensional contours of multidimensional
response surface models. For example, the following opens the interface with
a quadratic response surface fit to the data in reaction.mat:
load reaction
alpha = 0.01; % Significance level
rstool(reactants,rate,'quadratic',alpha,xn,yn)
A sequence of plots is displayed, each showing a contour of the response
surface against a single predictor, with all other predictors held fixed.
Confidence intervals for new observations are shown as dashed red curves
above and below the response. Predictor values are displayed in the text
boxes on the horizontal axis and are marked by vertical dashed blue lines
9-51
9 Regression Analysis
in the plots. Predictor values are changed by editing the text boxes or by
dragging the dashed blue lines. When you change the value of a predictor, all
plots update to show the new point in predictor space.
Note The Statistics Toolbox demonstration function rsmdemo generates
simulated data for experimental settings specified by either the user or by
a D-optimal design generated by cordexch. It uses the rstool interface to
visualize response surface models fit to the data, and it uses the nlintool
interface to visualize a nonlinear model fit to the data.
Generalized Linear Models
• “Introduction” on page 9-52
• “Example: Generalized Linear Models” on page 9-53
Introduction
Linear regression models describe a linear relationship between a response
and one or more predictive terms. Many times, however, a nonlinear
relationship exists. “Nonlinear Regression” on page 9-58 describes general
nonlinear models. A special class of nonlinear models, known as generalized
linear models, makes use of linear methods.
Recall that linear models have the following characteristics:
• At each set of values for the predictors, the response has a normal
distribution with mean μ.
• A coefficient vector b defines a linear combination Xb of the predictors X.
• The model is μ = Xb.
In generalized linear models, these characteristics are generalized as follows:
• At each set of values for the predictors, the response has a distribution
that may be normal, binomial, Poisson, gamma, or inverse Gaussian, with
parameters including a mean μ.
• A coefficient vector b defines a linear combination Xb of the predictors X.
9-52
Linear Regression
• A link function f defines the model as f(μ) = Xb.
Example: Generalized Linear Models
The following data are derived from carbig.mat, which contains
measurements of large cars of various weights. Each weight in w has a
corresponding number of cars in total and a corresponding number of
poor-mileage cars in poor:
w = [2100 2300 2500 2700 2900 3100 ...
3300 3500 3700 3900 4100 4300]';
total = [48 42 31 34 31 21 23 23 21 16 17 21]';
poor = [1 2 0 3 8 8 14 17 19 15 17 21]';
A plot shows that the proportion of poor-mileage cars follows an S-shaped
sigmoid:
plot(w,poor./total,'x','LineWidth',2)
grid on
xlabel('Weight')
ylabel('Proportion of Poor-Mileage Cars')
9-53
9 Regression Analysis
The logistic model is useful for proportion data. It defines the relationship
between the proportion p and the weight w by:
log[p/(1 – p)] = b1 + b2w
Some of the proportions in the data are 0 and 1, making the left-hand side of
this equation undefined. To keep the proportions within range, add relatively
small perturbations to the poor and total values. A semi-log plot then shows
a nearly linear relationship, as predicted by the model:
p_adjusted = (poor+.5)./(total+1);
semilogy(w,p_adjusted./(1-p_adjusted),'x','LineWidth',2)
grid on
xlabel('Weight')
ylabel('Adjusted p / (1 - p)')
9-54
Linear Regression
It is reasonable to assume that the values of poor follow binomial
distributions, with the number of trials given by total and the percentage
of successes depending on w. This distribution can be accounted for in the
context of a logistic model by using a generalized linear model with link
function log(µ/(1 – µ)) = Xb.
Use the glmfit function to carry out the associated regression:
b = glmfit(w,[poor total],'binomial','link','logit')
b =
-13.3801
0.0042
To use the coefficients in b to compute fitted proportions, invert the logistic
relationship:
p = 1/(1 + exp(–b1 – b2w))
9-55
9 Regression Analysis
Use the glmval function to compute the fitted values:
x = 2100:100:4500;
y = glmval(b,x,'logit');
plot(w,poor./total,'x','LineWidth',2)
hold on
plot(x,y,'r-','LineWidth',2)
grid on
xlabel('Weight')
ylabel('Proportion of Poor-Mileage Cars')
The previous is an example of logistic regression. For an example of a kind
of stepwise logistic regression, analogous to stepwise regression for linear
models, see “Sequential Feature Selection” on page 10-23.
9-56
Linear Regression
Multivariate Regression
Whether or not the predictor x is a vector of predictor variables, multivariate
regression refers to the case where the response y = (y1, ..., yM) is a vector of
M response variables.
The Statistics Toolbox functions mvregress and mvregresslike are used
for multivariate regression analysis.
9-57
9 Regression Analysis
Nonlinear Regression
In this section...
“Nonlinear Regression Models” on page 9-58
“Parametric Models” on page 9-59
“Mixed-Effects Models” on page 9-64
“Regression Trees” on page 9-94
Nonlinear Regression Models
The models described in “Linear Regression Models” on page 9-3 are often
called empirical models, because they are based solely on observed data.
Model parameters typically have no relationship to any mechanism producing
the data. To increase the accuracy of a linear model within the range of
observations, the number of terms is simply increased.
Nonlinear models, on the other hand, typically involve parameters with
specific physical interpretations. While they require a priori assumptions
about the data-producing process, they are often more parsimonious than
linear models, and more accurate outside the range of observed data.
Parametric nonlinear models represent the relationship between a continuous
response variable and one or more predictor variables (either continuous or
categorical) in the form y = f(X, β) + ε, where
• y is an n-by-1 vector of observations of the response variable.
• X is an n-by-p design matrix determined by the predictors.
• β is a p-by-1 vector of unknown parameters to be estimated.
• f is any function of X and β.
• ε is an n-by-1 vector of independent, identically distributed random
disturbances.
Nonparametric models do not attempt to characterize the relationship
between predictors and response with model parameters. Descriptions are
often graphical, as in the case of “Regression Trees” on page 9-94.
9-58
Nonlinear Regression
Parametric Models
• “A Parametric Nonlinear Model” on page 9-59
• “Confidence Intervals for Parameter Estimates” on page 9-61
• “Confidence Intervals for Predicted Responses” on page 9-61
• “Interactive Nonlinear Parametric Regression” on page 9-62
A Parametric Nonlinear Model
The Hougen-Watson model (Bates and Watts, [2], pp. 271–272) for reaction
kinetics is an example of a parametric nonlinear model. The form of the
model is
1 x2 − x3 / 5
rate =
1 + 2 x1 + 3 x2 + 4 x3
where rate is the reaction rate, x1, x2, and x3 are concentrations of hydrogen,
n-pentane, and isopentane, respectively, and β1, β2, ... , β5 are the unknown
parameters.
The file reaction.mat contains simulated reaction data:
load reaction
The variables are:
• rate — A 13-by-1 vector of observed reaction rates
• reactants — A 13-by-3 matrix of reactant concentrations
• beta — A 5-by-1 vector of initial parameter estimates
• model — The name of a function file for the model
• xn — The names of the reactants
• yn — The name of the response
The function for the model is hougen, which looks like this:
type hougen
9-59
9 Regression Analysis
function yhat = hougen(beta,x)
%HOUGEN Hougen-Watson model for reaction kinetics.
% YHAT = HOUGEN(BETA,X) gives the predicted values of the
% reaction rate, YHAT, as a function of the vector of
% parameters, BETA, and the matrix of data, X.
% BETA must have five elements and X must have three
% columns.
%
% The model form is:
% y = (b1*x2 - x3/b5)./(1+b2*x1+b3*x2+b4*x3)
b1 = beta(1);
b2 = beta(2);
b3 = beta(3);
b4 = beta(4);
b5 = beta(5);
x1 = x(:,1);
x2 = x(:,2);
x3 = x(:,3);
yhat = (b1*x2 - x3/b5)./(1+b2*x1+b3*x2+b4*x3);
The function nlinfit is used to find least-squares parameter estimates
for nonlinear models. It uses the Gauss-Newton algorithm with
Levenberg-Marquardt modifications for global convergence.
nlinfit requires the predictor data, the responses, and an initial guess of the
unknown parameters. It also requires a function handle to a function that
takes the predictor data and parameter estimates and returns the responses
predicted by the model.
To fit the reaction data, call nlinfit using the following syntax:
load reaction
betahat = nlinfit(reactants,rate,@hougen,beta)
betahat =
1.2526
0.0628
9-60
Nonlinear Regression
0.0400
0.1124
1.1914
The output vector betahat contains the parameter estimates.
The function nlinfit has robust options, similar to those for robustfit, for
fitting nonlinear models to data with outliers.
Confidence Intervals for Parameter Estimates
To compute confidence intervals for the parameter estimates, use the function
nlparci, together with additional outputs from nlinfit:
[betahat,resid,J] = nlinfit(reactants,rate,@hougen,beta);
betaci = nlparci(betahat,resid,J)
betaci =
-0.7467 3.2519
-0.0377 0.1632
-0.0312 0.1113
-0.0609 0.2857
-0.7381 3.1208
The columns of the output betaci contain the lower and upper bounds,
respectively, of the (default) 95% confidence intervals for each parameter.
Confidence Intervals for Predicted Responses
The function nlpredci is used to compute confidence intervals for predicted
responses:
[yhat,delta] = nlpredci(@hougen,reactants,betahat,resid,J);
opd = [rate yhat delta]
opd =
8.5500 8.4179 0.2805
3.7900 3.9542 0.2474
4.8200 4.9109 0.1766
0.0200 -0.0110 0.1875
2.7500 2.6358 0.1578
14.3900 14.3402 0.4236
2.5400 2.5662 0.2425
9-61
9 Regression Analysis
4.3500 4.0385 0.1638
13.0000 13.0292 0.3426
8.5000 8.3904 0.3281
0.0500 -0.0216 0.3699
11.3200 11.4701 0.3237
3.1300 3.4326 0.1749
The output opd contains the observed rates in the first column and the
predicted rates in the second column. The (default) 95% simultaneous
confidence intervals on the predictions are the values in the second column ±
the values in the third column. These are not intervals for new observations
at the predictors, even though most of the confidence intervals do contain the
original observations.
Interactive Nonlinear Parametric Regression
Calling nlintool opens a graphical user interface (GUI) for interactive
exploration of multidimensional nonlinear functions, and for fitting
parametric nonlinear models. The GUI calls nlinfit, and requires the same
inputs. The interface is analogous to polytool and rstool for polynomial
models.
Open nlintool with the reaction data and the hougen model by typing
load reaction
nlintool(reactants,rate,@hougen,beta,0.01,xn,yn)
9-62
Nonlinear Regression
You see three plots. The response variable for all plots is the reaction rate,
plotted in green. The red lines show confidence intervals on predicted
responses. The first plot shows hydrogen as the predictor, the second shows
n-pentane, and the third shows isopentane.
Each plot displays the fitted relationship of the reaction rate to one predictor
at a fixed value of the other two predictors. The fixed values are in the text
boxes below each predictor axis. Change the fixed values by typing in a new
value or by dragging the vertical lines in the plots to new positions. When
you change the value of a predictor, all plots update to display the model
at the new point in predictor space.
9-63
9 Regression Analysis
While this example uses only three predictors, nlintool can accommodate
any number of predictors.
Note The Statistics Toolbox demonstration function rsmdemo generates
simulated data for experimental settings specified by either the user or by
a D-optimal design generated by cordexch. It uses the rstool interface to
visualize response surface models fit to the data, and it uses the nlintool
interface to visualize a nonlinear model fit to the data.
Mixed-Effects Models
• “Introduction” on page 9-64
• “Mixed-Effects Model Hierarchy” on page 9-65
• “Specifying Mixed-Effects Models” on page 9-67
• “Specifying Covariate Models” on page 9-70
• “Choosing nlmefit or nlmefitsa” on page 9-71
• “Using Output Functions with Mixed-Effects Models” on page 9-74
• “Example: Mixed-Effects Models Using nlmefit and nlmefitsa” on page 9-80
Introduction
In statistics, an effect is anything that influences the value of a response
variable at a particular setting of the predictor variables. Effects are
translated into model parameters. In linear models, effects become
coefficients, representing the proportional contributions of model terms. In
nonlinear models, effects often have specific physical interpretations, and
appear in more general nonlinear combinations.
Fixed effects represent population parameters, assumed to be the same each
time data is collected. Estimating fixed effects is the traditional domain of
regression modeling. Random effects, by comparison, are sample-dependent
random variables. In modeling, random effects act like additional error terms,
and their distributions and covariances must be specified.
9-64
Nonlinear Regression
For example, consider a model of the elimination of a drug from the
bloodstream. The model uses time t as a predictor and the concentration
of the drug C as the response. The nonlinear model term C0e–rt combines
parameters C0 and r, representing, respectively, an initial concentration
and an elimination rate. If data is collected across multiple individuals, it
is reasonable to assume that the elimination rate is a random variable ri
depending on individual i, varying around a population mean r . The term
C0e–rt becomes
C0 e−[ r +( ri − r )]t = C0 e− ( + bi )t ,
where β = r is a fixed effect and bi = ri − r is a random effect.
Random effects are useful when data falls into natural groups. In the drug
elimination model, the groups are simply the individuals under study. More
sophisticated models might group data by an individual’s age, weight, diet,
etc. Although the groups are not the focus of the study, adding random effects
to a model extends the reliability of inferences beyond the specific sample of
individuals.
Mixed-effects models account for both fixed and random effects. As with
all regression models, their purpose is to describe a response variable as a
function of the predictor variables. Mixed-effects models, however, recognize
correlations within sample subgroups. In this way, they provide a compromise
between ignoring data groups entirely and fitting each group with a separate
model.
Mixed-Effects Model Hierarchy
Suppose data for a nonlinear regression model falls into one of m distinct
groups i = 1, ..., m. To account for the groups in a model, write response j
in group i as:
yij = f ( , xij ) + ij
yij is the response, xij is a vector of predictors, φ is a vector of model
parameters, and εij is the measurement or process error. The index j ranges
from 1 to ni, where ni is the number of observations in group i. The function
f specifies the form of the model. Often, xij is simply an observation time tij.
9-65
9 Regression Analysis
The errors are usually assumed to be independent and identically, normally
distributed, with constant variance.
Estimates of the parameters in φ describe the population, assuming those
estimates are the same for all groups. If, however, the estimates vary by
group, the model becomes
yij = f ( i , xij ) + ij
In a mixed-effects model, φi may be a combination of a fixed and a random
effect:
i = + bi
The random effects bi are usually described as multivariate normally
distributed, with mean zero and covariance Ψ. Estimating the fixed effects
β and the covariance of the random effects Ψ provides a description of the
population that does not assume the parameters φi are the same across
groups. Estimating the random effects bi also gives a description of specific
groups within the data.
Model parameters do not have to be identified with individual effects. In
general, design matrices A and B are used to identify parameters with linear
combinations of fixed and random effects:
i = A + Bbi
If the design matrices differ among groups, the model becomes
i = Ai + Bibi
If the design matrices also differ among observations, the model becomes
ij = Aij + Bij bi
yij = f ( ij , xij ) + ij
9-66
Nonlinear Regression
Some of the group-specific predictors in xij may not change with observation j.
Calling those vi, the model becomes
yij = f ( ij , xij , vi ) + ij
Specifying Mixed-Effects Models
Suppose data for a nonlinear regression model falls into one of m distinct
groups i = 1, ..., m. (Specifically, suppose that the groups are not nested.) To
specify a general nonlinear mixed-effects model for this data:
1 Define group-specific model parameters φi as linear combinations of fixed
effects β and random effects bi.
2 Define response values yi as a nonlinear function f of the parameters and
group-specific predictor variables Xi.
The model is:
i = Ai + Bi bi
yi = f (i , X i ) + i
bi ∼ N (0, Ψ)
i ∼ N (0, 2 )
This formulation of the nonlinear mixed-effects model uses the following
notation:
φi A vector of group-specific model parameters
β A vector of fixed effects, modeling population parameters
bi A vector of multivariate normally distributed group-specific
random effects
Ai A group-specific design matrix for combining fixed effects
Bi A group-specific design matrix for combining random effects
Xi A data matrix of group-specific predictor values
yi A data vector of group-specific response values
9-67
9 Regression Analysis
f A general, real-valued function of φi and Xi
εi A vector of group-specific errors, assumed to be independent,
identically, normally distributed, and independent of bi
Ψ A covariance matrix for the random effects
σ 2
The error variance, assumed to be constant across observations
For example, consider a model of the elimination of a drug from the
bloodstream. The model incorporates two overlapping phases:
• An initial phase p during which drug concentrations reach equilibrium
with surrounding tissues
• A second phase q during which the drug is eliminated from the bloodstream
For data on multiple individuals i, the model is
− rpitij − rqitij
yij = C pi e + Cqi e + ij ,
where yij is the observed concentration in individual i at time tij. The model
allows for different sampling times and different numbers of observations for
different individuals.
The elimination rates rpi and rqi must be positive to be physically meaningful.
Enforce this by introducing the log rates Rpi = log(rpi) and Rqi = log(rqi) and
reparametrizing the model:
− exp( Rpi )tij − exp( Rqi )tij
yij = C pi e + Cqi e + ij
Choosing which parameters to model with random effects is an important
consideration when building a mixed-effects model. One technique is to add
random effects to all parameters, and use estimates of their variances to
determine their significance in the model. An alternative is to fit the model
separately to each group, without random effects, and look at the variation
of the parameter estimates. If an estimate varies widely across groups, or if
confidence intervals for each group have minimal overlap, the parameter is a
good candidate for a random effect.
9-68
Nonlinear Regression
To introduce fixed effects β and random effects bi for all model parameters,
reexpress the model as follows:
− exp[ Rp + ( Rpi − Rp )]tij
yij = [C p + (C pi − C p )]e +
− exp[ Rq + ( Rqi − Rq )]tij
[Cq + (Cqi − Cq )]e + ij
− exp( 2 + b2 i )tij
= (1 + b1i ) e +
− exp( 4 + b4 i )tij
( 3 + b3i ) e + ij
In the notation of the general model:
⎛ 1 ⎞ ⎛ bi1 ⎞ ⎛ yi1 ⎞ ⎛ ti1 ⎞
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
= ⎜ ⎟ , bi = ⎜ ⎟ , yi = ⎜ ⎟ , X i = ⎜ ⎟ ,
⎜ ⎟ ⎜b ⎟ ⎜y ⎟ ⎜t ⎟
⎝ 4⎠ ⎝ i4 ⎠ ⎝ ini ⎠ ⎝ ini ⎠
where ni is the number of observations of individual i. In this case, the design
matrices Ai and Bi are, at least initially, 4-by-4 identity matrices. Design
matrices may be altered, as necessary, to introduce weighting of individual
effects, or time dependency.
Fitting the model and estimating the covariance matrix Ψ often leads to
further refinements. A relatively small estimate for the variance of a random
effect suggests that it can be removed from the model. Likewise, relatively
small estimates for covariances among certain random effects suggests that a
full covariance matrix is unnecessary. Since random effects are unobserved,
Ψ must be estimated indirectly. Specifying a diagonal or block-diagonal
covariance pattern for Ψ can improve convergence and efficiency of the fitting
algorithm.
Statistics Toolbox functions nlmefit and nlmefitsa fit the general nonlinear
mixed-effects model to data, estimating the fixed and random effects. The
functions also estimate the covariance matrix Ψ for the random effects.
Additional diagnostic outputs allow you to assess tradeoffs between the
number of model parameters and the goodness of fit.
9-69
9 Regression Analysis
Specifying Covariate Models
If the model in “Specifying Mixed-Effects Models” on page 9-67 assumes a
group-dependent covariate such as weight (w) the model becomes:
⎛ 1 ⎞
⎛ 1 ⎞ ⎛ 1 0 0 wi ⎞ ⎜ ⎟ ⎛ 1 0 0 ⎞ ⎛ b1 ⎞
⎜ ⎟ ⎜ ⎟ ⎜ 2 ⎟ ⎜ ⎟⎜ ⎟
⎜ 2 ⎟ = ⎜ 0 1 0 0 ⎟ ⎜ ⎟ + ⎜ 0 1 0 ⎟ ⎜ b2 ⎟
⎜ ⎟ ⎜ 0 0 1 0 ⎟ ⎜ 3 ⎟ ⎜ 0 0 1 ⎟ ⎜ b ⎟
⎝ 3⎠ ⎝ ⎠⎜ ⎟ ⎝ ⎠⎝ 3 ⎠
⎝ 4⎠
Thus, the parameter φi for any individual in the ith group is:
⎛ 1 ⎞ ⎛ + * w ⎞ ⎛ b1 ⎞
⎜ i ⎟ ⎜ 1 4 i
⎟ ⎜⎜
i
⎟
⎜ 2 ⎟=⎜ 2 ⎟ + ⎜ b2i ⎟
⎜ i ⎟ ⎜ ⎟ ⎜ ⎟
⎜ 3 ⎟ ⎝ 3 ⎠ ⎝ b3i ⎟
⎝ i ⎠ ⎠
To specify a covariate model, use the 'FEGroupDesign' option.
'FEGroupDesign' is a p-by-q-by-m array specifying a different p-by-q
fixed-effects design matrix for each of the m groups. Using the previous
example, the array resembles the following:
1 Create the array.
% Number of parameters in the model (Phi)
num_params = 3;
9-70
Nonlinear Regression
% Number of covariates
num_cov = 1;
% Assuming number of groups in the data set is 7
num_groups = 7;
% Array of covariate values
covariates = [75; 52; 66; 55; 70; 58; 62 ];
A = repmat(eye(num_params, num_params+num_cov),...
[1,1,num_groups]);
A(1,num_params+1,1:num_groups) = covariates(:,1)
2 Create a struct with the specified design matrix.
options.FEGroupDesign = A;
3 Specify the arguments for nlmefit (or nlmefitsa) as shown in “Example:
Mixed-Effects Models Using nlmefit and nlmefitsa” on page 9-80.
Choosing nlmefit or nlmefitsa
Statistics Toolbox provides two functions, nlmefit and nlmefitsa for fitting
non-linear mixed-effects models. Each function provides different capabilities,
which may help you decide which to use.
• “Approximation Methods” on page 9-71
• “Parameters Specific to nlmefitsa” on page 9-72
• “Model and Data Requirements” on page 9-73
Approximation Methods. nlmefit provides the following four
approximation methods for fitting non-linear mixed-effects models:
• 'LME' — Use the likelihood for the linear mixed-effects model at the
current conditional estimates of beta and B. This is the default.
• 'RELME' — Use the restricted likelihood for the linear mixed-effects model
at the current conditional estimates of beta and B.
• 'FO' — First-order Laplacian approximation without random effects.
• 'FOCE' — First-order Laplacian approximation at the conditional estimates
of B.
9-71
9 Regression Analysis
nlmefitsa provides an additional approximation method, Stochastic
Approximation Expectation-Maximization (SAEM) [24] with three steps :
1 Simulation: Generate simulated values of the random effects b from the
posterior density p(b|Σ) given the current parameter estimates.
2 Stochastic approximation: Update the expected value of the log likelihood
function by taking its value from the previous step, and moving part
way toward the average value of the log likelihood calculated from the
simulated random effects.
3 Maximization step: Choose new parameter estimates to maximize the log
likelihood function given the simulated values of the random effects.
Both nlmefit and nlmefitsa attempt to find parameter estimates to
maximize a likelihood function, which is difficult to compute. nlmefit deals
with the problem by approximating the likelihood function in various ways,
and maximizing the approximate function. It uses traditional optimization
techniques that depend on things like convergence criteria and iteration
limits.
nlmefitsa, on the other hand, simulates random values of the parameters in
such a way that in the long run they converge to the values that maximize
the exact likelihood function. The results are random, and traditional
convergence tests don’t apply. Therefore nlmefitsa provides options to plot
the results as the simulation progresses, and to re-start the simulation
multiple times. You can use these features to judge whether the results have
converged to the accuracy you desire.
Parameters Specific to nlmefitsa. The following parameters are specific to
nlmefitsa. Most control the stochastic algorithm.
• ErrorModel — A string specifying the form of the error term. Default
is 'constant'. Each model defines the error using a standard normal
(Gaussian) variable e, the function value f, and one or two parameters
a and b.
• Cov0 — Initial value for the covariance matrix PSI. Must be an r-by-r
positive definite matrix. If empty, the default value depends on the values
of BETA0.
9-72
Nonlinear Regression
• ComputeStdErrors — true to compute standard errors for the coefficient
estimates and store them in the output STATS structure, or false (default)
to omit this computation.
• LogLikMethod — Specifies the method for approximating the log likelihood.
• NBurnIn — Number of initial burn-in iterations during which the
parameter estimates are not recomputed. Default is 5.
• NIterations — Controls how many iterations are performed for each of
three phases of the algorithm.
• NMCMCIterations — Number of Markov Chain Monte Carlo (MCMC)
iterations.
Model and Data Requirements. There are some differences in the
capabilities of nlmefit and nlmefitsa. Therefore some data and models
are usable with either function, but some may require you to choose just
one of them.
• Error models — nlmefitsa supports a variety of error models. For
example, the standard deviation of the response can be constant,
proportional to the function value, or a combination of the two. nlmefit fits
models under the assumption that the standard deviation of the response
is constant. One of the error models, 'exponential', specifies that the log
of the response has a constant standard deviation. You can fit such models
using nlmefit by providing the log response as input, and by re-writing the
model function to produce the log of the nonlinear function value.
• Random effects — Both functions fit data to a nonlinear function with
parameters, and the parameters may be simple scalar values or linear
functions of covariates. nlmefit allows any coefficients of the linear
functions to have both fixed and random effects. nlmefitsa supports
random effects only for the constant (intercept) coefficient of the linear
functions, but not for slope coefficients. So in the example in “Specifying
Covariate Models” on page 9-70, nlmefitsa can treat only the first three
beta values as random effects.
• Model form — nlmefit supports a very general model specification, with
few restrictions on the design matrices that relate the fixed coefficients and
the random effects to the model parameters. nlmefitsa is more restrictive:
9-73
9 Regression Analysis
- The fixed effect design must be constant in every group (for every
individual), so an observation-dependent design is not supported.
- The random effect design must be constant for the entire data set, so
neither an observation-dependent design nor a group-dependent design
is supported.
- As mentioned under Random Effects, the random effect design must
not specify random effects for slope coefficients. This implies that the
design must consist of zeros and ones.
- The random effect design must not use the same random effect for
multiple coefficients, and cannot use more than one random effect for
any single coefficient.
- The fixed effect design must not use the same coefficient for multiple
parameters. This implies that it can have at most one non-zero value
in each column.
If you want to use nlmefitsa for data in which the covariate effects are
random, include the covariates directly in the nonlinear model expression.
Don’t include the covariates in the fixed or random effect design matrices.
• Convergence — As described in the Model form, nlmefit and nlmefitsa
have different approaches to measuring convergence. nlmefit uses
traditional optimization measures, and nlmefitsa provides diagnostics to
help you judge the convergence of a random simulation.
In practice, nlmefitsa tends to be more robust, and less likely to fail on
difficult problems. However, nlmefit may converge faster on problems where
it converges at all. Some problems may benefit from a combined strategy,
for example by running nlmefitsa for a while to get reasonable parameter
estimates, and using those as a starting point for additional iterations using
nlmefit.
Using Output Functions with Mixed-Effects Models
The Outputfcn field of the options structure specifies one or more functions
that the solver calls after each iteration. Typically, you might use an output
function to plot points at each iteration or to display optimization quantities
from the algorithm. To set up an output function:
1 Write the output function as a MATLAB file function or subfunction.
9-74
Nonlinear Regression
2 Use statset to set the value of Outputfcn to be a function handle, that is,
the name of the function preceded by the @ sign. For example, if the output
function is outfun.m, the command
options = statset('OutputFcn', @outfun);
specifies OutputFcn to be the handle to outfun. To specify multiple output
functions, use the syntax:
options = statset('OutputFcn',{@outfun, @outfun2});
3 Call the optimization function with options as an input argument.
For an example of an output function, see “Sample Output Function” on page
9-80.
Structure of the Output Function. The function definition line of the
output function has the following form:
stop = outfun(beta,status,state)
where
• beta is the current fixed effects.
• status is a structure containing data from the current iteration. “Fields in
status” on page 9-75 describes the structure in detail.
• state is the current state of the algorithm. “States of the Algorithm” on
page 9-78 lists the possible values.
• stop is a flag that is true or false depending on whether the optimization
routine should quit or continue. See “Stop Flag” on page 9-78 for more
information.
The solver passes the values of the input arguments to outfun at each
iteration.
Fields in status. The following table lists the fields of the status structure:
9-75
9 Regression Analysis
Field Description
procedure • 'ALT' — alternating algorithm for the optimization of
the linear mixed effects or restricted linear mixed effects
approximations
• 'LAP' — optimization of the Laplacian approximation for
first order or first order conditional estimation
iteration An integer starting from 0.
inner A structure describing the status of the inner iterations
within the ALT and LAP procedures, with the fields:
• procedure — When procedure is 'ALT':
- 'PNLS' (penalized non-linear least squares)
- 'LME' (linear mixed-effects estimation)
- 'none'
When procedure is 'LAP',
- 'PNLS' (penalized non-linear least squares)
- 'PLM' (profiled likelihood maximization)
- 'none'
• state — one of the following:
- 'init'
- 'iter'
- 'done'
- 'none'
• iteration — an integer starting from 0, or NaN. For
nlmefitsa with burn-in iterations, the output function is
called after each of those iterations with a negative value
for STATUS.iteration.
fval The current log likelihood
Psi The current random-effects covariance matrix
9-76
Nonlinear Regression
Field Description
theta The current parameterization of Psi
mse The current error variance
9-77
9 Regression Analysis
States of the Algorithm. The following table lists the possible values for
state:
state Description
'init' The algorithm is in the initial state before the first
iteration.
'iter' The algorithm is at the end of an iteration.
'done' The algorithm is in the final state after the last iteration.
The following code illustrates how the output function might use the value of
state to decide which tasks to perform at the current iteration:
switch state
case 'iter'
% Make updates to plot or guis as needed
case 'init'
% Setup for plots or guis
case 'done'
% Cleanup of plots, guis, or final plot
otherwise
end
Stop Flag. The output argument stop is a flag that is true or false.
The flag tells the solver whether it should quit or continue. The following
examples show typical ways to use the stop flag.
Stopping an Optimization Based on Intermediate Results
The output function can stop the estimation at any iteration based on the
values of arguments passed into it. For example, the following code sets stop
to true based on the value of the log likelihood stored in the 'fval'field of
the status structure:
stop = outfun(beta,status,state)
stop = false;
% Check if loglikelihood is more than 132.
if status.fval > -132
stop = true;
9-78
Nonlinear Regression
end
Stopping an Iteration Based on GUI Input
If you design a GUI to perform nlmefit iterations, you can make the output
function stop when a user clicks a Stop button on the GUI. For example, the
following code implements a dialog to cancel calculations:
function retval = stop_outfcn(beta,str,status)
persistent h stop;
if isequal(str.inner.state,'none')
switch(status)
case 'init'
% Initialize dialog
stop = false;
h = msgbox('Press STOP to cancel calculations.',...
'NLMEFIT: Iteration 0 ');
button = findobj(h,'type','uicontrol');
set(button,'String','STOP','Callback',@stopper)
pos = get(h,'Position');
pos(3) = 1.1 * pos(3);
set(h,'Position',pos)
drawnow
case 'iter'
% Display iteration number in the dialog title
set(h,'Name',sprintf('NLMEFIT: Iteration %d',...
str.iteration))
drawnow;
case 'done'
% Delete dialog
delete(h);
end
end
if stop
% Stop if the dialog button has been pressed
delete(h)
end
retval = stop;
function stopper(varargin)
9-79
9 Regression Analysis
% Set flag to stop when button is pressed
stop = true;
disp('Calculation stopped.')
end
end
Sample Output Function. nmlefitoutputfcn is the sample Statistics
Toolbox output function for nlmefit and nlmefitsa. It initializes or updates
a plot with the fixed-effects (BETA) and variance of the random effects
(diag(STATUS.Psi)). For nlmefit, the plot also includes the log-likelihood
(STATUS.fval).
nlmefitoutputfcn is the default output function for nlmefitsa. To use it
with nlmefit, specify a function handle for it in the options structure:
opt = statset('OutputFcn', @nlmefitoutputfcn, )
beta = nlmefit( , 'Options', opt, )
To prevent nlmefitsa from using of this function, specify an empty value for
the output function:
opt = statset('OutputFcn', [], )
beta = nlmefitsa( , 'Options', opt, )
nlmefitoutputfcn stops nlmefit or nlmefitsa if you close the figure that
it produces.
Example: Mixed-Effects Models Using nlmefit and nlmefitsa
The following example also works with nlmefitsa in place of nlmefit.
The data in indomethacin.mat records concentrations of the drug
indomethacin in the bloodstream of six subjects over eight hours:
load indomethacin
gscatter(time,concentration,subject)
xlabel('Time (hours)')
ylabel('Concentration (mcg/ml)')
title('{\bf Indomethacin Elimination}')
hold on
9-80
Nonlinear Regression
“Specifying Mixed-Effects Models” on page 9-67 discusses a useful model for
this type of data. Construct the model via an anonymous function as follows:
model = @(phi,t)(phi(1)*exp(-exp(phi(2))*t) + ...
phi(3)*exp(-exp(phi(4))*t));
Use the nlinfit function to fit the model to all of the data, ignoring
subject-specific effects:
phi0 = [1 1 1 1];
[phi,res] = nlinfit(time,concentration,model,phi0);
numObs = length(time);
numParams = 4;
df = numObs-numParams;
mse = (res'*res)/df
9-81
9 Regression Analysis
mse =
0.0304
tplot = 0:0.01:8;
plot(tplot,model(phi,tplot),'k','LineWidth',2)
hold off
A boxplot of residuals by subject shows that the boxes are mostly above or
below zero, indicating that the model has failed to account for subject-specific
effects:
colors = 'rygcbm';
h = boxplot(res,subject,'colors',colors,'symbol','o');
set(h(~isnan(h)),'LineWidth',2)
hold on
boxplot(res,subject,'colors','k','symbol','ko')
9-82
Nonlinear Regression
grid on
xlabel('Subject')
ylabel('Residual')
hold off
To account for subject-specific effects, fit the model separately to the data
for each subject:
phi0 = [1 1 1 1];
PHI = zeros(4,6);
RES = zeros(11,6);
for I = 1:6
tI = time(subject == I);
cI = concentration(subject == I);
[PHI(:,I),RES(:,I)] = nlinfit(tI,cI,model,phi0);
end
9-83
9 Regression Analysis
PHI
PHI =
0.1915 0.4989 1.6757 0.2545 3.5661 0.9685
-1.7878 -1.6354 -0.4122 -1.6026 1.0408 -0.8731
2.0293 2.8277 5.4683 2.1981 0.2915 3.0023
0.5794 0.8013 1.7498 0.2423 -1.5068 1.0882
numParams = 24;
df = numObs-numParams;
mse = (RES(:)'*RES(:))/df
mse =
0.0057
gscatter(time,concentration,subject)
xlabel('Time (hours)')
ylabel('Concentration (mcg/ml)')
title('{\bf Indomethacin Elimination}')
hold on
for I = 1:6
plot(tplot,model(PHI(:,I),tplot),'Color',colors(I))
end
axis([0 8 0 3.5])
hold off
9-84
Nonlinear Regression
PHI gives estimates of the four model parameters for each of the six subjects.
The estimates vary considerably, but taken as a 24-parameter model of the
data, the mean-squared error of 0.0057 is a significant reduction from 0.0304
in the original four-parameter model.
A boxplot of residuals by subject shows that the larger model accounts for
most of the subject-specific effects:
h = boxplot(RES,'colors',colors,'symbol','o');
set(h(~isnan(h)),'LineWidth',2)
hold on
boxplot(RES,'colors','k','symbol','ko')
grid on
xlabel('Subject')
ylabel('Residual')
9-85
9 Regression Analysis
hold off
The spread of the residuals (the vertical scale of the boxplot) is much smaller
than in the previous boxplot, and the boxes are now mostly centered on zero.
While the 24-parameter model successfully accounts for variations due
to the specific subjects in the study, it does not consider the subjects as
representatives of a larger population. The sampling distribution from which
the subjects are drawn is likely more interesting than the sample itself. The
purpose of mixed-effects models is to account for subject-specific variations
more broadly, as random effects varying around population means.
Use the nlmefit function to fit a mixed-effects model to the data.
The following anonymous function, nlme_model, adapts the four-parameter
model used by nlinfit to the calling syntax of nlmefit by allowing separate
9-86
Nonlinear Regression
parameters for each individual. By default, nlmefit assigns random effects
to all the model parameters. Also by default, nlmefit assumes a diagonal
covariance matrix (no covariance among the random effects) to avoid
overparametrization and related convergence issues.
nlme_model = @(PHI,t)(PHI(:,1).*exp(-exp(PHI(:,2)).*t) + ...
PHI(:,3).*exp(-exp(PHI(:,4)).*t));
phi0 = [1 1 1 1];
[phi,PSI,stats] = nlmefit(time,concentration,subject, ...
[],nlme_model,phi0)
phi =
0.4606
-1.3459
2.8277
0.7729
PSI =
0.0124 0 0 0
0 0.0000 0 0
0 0 0.3264 0
0 0 0 0.0250
stats =
logl: 54.5884
mse: 0.0066
aic: -91.1767
bic: -71.4698
sebeta: NaN
dfe: 57
The mean-squared error of 0.0066 is comparable to the 0.0057 of the
24-parameter model without random effects, and significantly better than the
0.0304 of the four-parameter model without random effects.
The estimated covariance matrix PSI shows that the variance of the second
random effect is essentially zero, suggesting that you can remove it to simplify
the model. To do this, use the REParamsSelect parameter to specify the
indices of the parameters to be modeled with random effects in nlmefit:
[phi,PSI,stats] = nlmefit(time,concentration,subject, ...
[],nlme_model,phi0, ...
'REParamsSelect',[1 3 4])
9-87
9 Regression Analysis
phi =
0.4606
-1.3460
2.8277
0.7729
PSI =
0.0124 0 0
0 0.3270 0
0 0 0.0250
stats =
logl: 54.5876
mse: 0.0066
aic: -93.1752
bic: -75.6580
sebeta: NaN
dfe: 58
The log-likelihood logl is almost identical to what it was with random effects
for all of the parameters, the Akaike information criterion aic is reduced
from -91.1767 to -93.1752, and the Bayesian information criterion bic is
reduced from -71.4698 to -75.6580. These measures support the decision to
drop the second random effect.
Refitting the simplified model with a full covariance matrix allows for
identification of correlations among the random effects. To do this, use the
CovPattern parameter to specify the pattern of nonzero elements in the
covariance matrix:
[phi,PSI,stats] = nlmefit(time,concentration,subject, ...
[],nlme_model,phi0, ...
'REParamsSelect',[1 3 4], ...
'CovPattern',ones(3))
phi =
0.5613
-1.1407
2.8148
0.8293
PSI =
0.0236 0.0500 0.0032
0.0500 0.4768 0.1152
9-88
Nonlinear Regression
0.0032 0.1152 0.0321
stats =
logl: 58.4731
mse: 0.0061
aic: -94.9462
bic: -70.8600
sebeta: NaN
dfe: 55
The estimated covariance matrix PSI shows that the random effects on the
last two parameters have a relatively strong correlation, and both have a
relatively weak correlation with the first random effect. This structure in
the covariance matrix is more apparent if you convert PSI to a correlation
matrix using corrcov:
RHO = corrcov(PSI)
RHO =
1.0000 0.4707 0.1179
0.4707 1.0000 0.9316
0.1179 0.9316 1.0000
clf; imagesc(RHO)
set(gca,'XTick',[1 2 3],'YTick',[1 2 3])
title('{\bf Random Effect Correlation}')
h = colorbar;
set(get(h,'YLabel'),'String','Correlation');
9-89
9 Regression Analysis
Incorporate this structure into the model by changing the specification of the
covariance pattern to block-diagonal:
P = [1 0 0;0 1 1;0 1 1] % Covariance pattern
P =
1 0 0
0 1 1
0 1 1
[phi,PSI,stats,b] = nlmefit(time,concentration,subject, ...
[],nlme_model,phi0, ...
'REParamsSelect',[1 3 4], ...
'CovPattern',P)
phi =
0.5850
9-90
Nonlinear Regression
-1.1087
2.8056
0.8476
PSI =
0.0331 0 0
0 0.4793 0.1069
0 0.1069 0.0294
stats =
logl: 57.4996
mse: 0.0061
aic: -96.9992
bic: -77.2923
sebeta: NaN
dfe: 57
b =
-0.2438 0.0723 0.2014 0.0592 -0.2181 0.1289
-0.8500 -0.1237 0.9538 -0.7267 0.5895 0.1571
-0.1591 0.0033 0.1568 -0.2144 0.1834 0.0300
The block-diagonal covariance structure reduces aic from -94.9462 to
-96.9992 and bic from -70.8600 to -77.2923 without significantly affecting
the log-likelihood. These measures support the covariance structure used in
the final model.
The output b gives predictions of the three random effects for each of the six
subjects. These are combined with the estimates of the fixed effects in phi
to produce the mixed-effects model.
Use the following commands to plot the mixed-effects model for each of the six
subjects. For comparison, the model without random effects is also shown.
PHI = repmat(phi,1,6) + ... % Fixed effects
[b(1,:);zeros(1,6);b(2,:);b(3,:)]; % Random effects
RES = zeros(11,6); % Residuals
colors = 'rygcbm';
for I = 1:6
fitted_model = @(t)(PHI(1,I)*exp(-exp(PHI(2,I))*t) + ...
PHI(3,I)*exp(-exp(PHI(4,I))*t));
tI = time(subject == I);
9-91
9 Regression Analysis
cI = concentration(subject == I);
RES(:,I) = cI - fitted_model(tI);
subplot(2,3,I)
scatter(tI,cI,20,colors(I),'filled')
hold on
plot(tplot,fitted_model(tplot),'Color',colors(I))
plot(tplot,model(phi,tplot),'k')
axis([0 8 0 3.5])
xlabel('Time (hours)')
ylabel('Concentration (mcg/ml)')
legend(num2str(I),'Subject','Fixed')
end
9-92
Nonlinear Regression
If obvious outliers in the data (visible in previous box plots) are ignored, a
normal probability plot of the residuals shows reasonable agreement with
model assumptions on the errors:
clf; normplot(RES(:))
9-93
9 Regression Analysis
Regression Trees
Introduction
Parametric models specify the form of the relationship between predictors and
a response, as in the Hougen-Watson model described in “Parametric Models”
on page 9-59. In many cases, the form of the relationship is unknown, and
a parametric model requires assumptions and simplifications. Regression
trees offer a nonparametric alternative. When response data are categorical,
classification trees are a natural modification.
9-94
Nonlinear Regression
Note This section demonstrates methods for objects of the classregtree
class. These methods supersede the functions treefit, treedisp, treeval,
treeprune, and treetest, which are maintained in Statistics Toolbox
software only for backwards compatibility.
Algorithm Reference. The algorithms used by Statistics Toolbox
classification and regression tree functions are based on those in Breiman,
L., et al., Classification and Regression Trees, Chapman & Hall, Boca Raton,
1993.
Example: Regression Trees
This example uses the data on cars in carsmall.mat to create a regression
tree for predicting mileage using measurements of weight and the number
of cylinders as predictors. Note that, in this case, one predictor (weight) is
continuous and the other (cylinders) is categorical. The response (mileage) is
continuous.
Load the data and use the classregtree constructor of the classregtree
class to create the regression tree:
load carsmall
t = classregtree([Weight, Cylinders],MPG,...
'cat',2,'splitmin',20,...
'names',{'W','C'})
t =
Decision tree for regression
1 if W<3085.5 then node 2 elseif W>=3085.5 then node 3 else 23.7181
2 if W<2371 then node 4 elseif W>=2371 then node 5 else 28.7931
3 if C=8 then node 6 elseif C in {4 6} then node 7 else 15.5417
4 if W<2162 then node 8 elseif W>=2162 then node 9 else 32.0741
5 if C=6 then node 10 elseif C=4 then node 11 else 25.9355
6 if W<4381 then node 12 elseif W>=4381 then node 13 else 14.2963
7 fit = 19.2778
8 fit = 33.3056
9-95
9 Regression Analysis
9 fit = 29.6111
10 fit = 23.25
11 if W<2827.5 then node 14 elseif W>=2827.5 then node 15 else 27.2143
12 if W<3533.5 then node 16 elseif W>=3533.5 then node 17 else 14.8696
13 fit = 11
14 fit = 27.6389
15 fit = 24.6667
16 fit = 16.6
17 fit = 14.3889
t is a classregtree object and can be operated on with any of the methods
of the class.
Use the type method of the classregtree class to show the type of the tree:
treetype = type(t)
treetype =
regression
classregtree creates a regression tree because MPG is a numerical vector,
and the response is assumed to be continuous.
To view the tree, use the view method of the classregtree class:
view(t)
9-96
Nonlinear Regression
The tree predicts the response values at the circular leaf nodes based on a
series of questions about the car at the triangular branching nodes. A true
answer to any question follows the branch to the left; a false follows the
branch to the right.
Use the tree to predict the mileage for a 2000-pound car with either 4, 6, or
8 cylinders:
mileage2K = t([2000 4; 2000 6; 2000 8])
mileage2K =
33.3056
33.3056
33.3056
Note that the object allows for functional evaluation, of the form t(X). This is
a shorthand way of calling the eval method of the classregtree class.
The predicted responses computed above are all the same. This is because they
follow a series of splits in the tree that depend only on weight, terminating
at the left-most leaf node in the view above. A 4000-pound car, following the
right branch from the top of the tree, leads to different predicted responses:
mileage4K = t([4000 4; 4000 6; 4000 8])
mileage4K =
19.2778
19.2778
14.3889
You can use a variety of other methods of the classregtree class, such as
cutvar, cuttype, and cutcategories, to get more information about the split
at node 3 that distinguishes the 8-cylinder car:
var3 = cutvar(t,3) % What variable determines the split?
var3 =
'C'
type3 = cuttype(t,3) % What type of split is it?
type3 =
'categorical'
c = cutcategories(t,3) % Which classes are sent to the left
9-97
9 Regression Analysis
% child node, and which to the right?
c =
[8] [1x2 double]
c{1}
ans =
8
c{2}
ans =
4 6
Regression trees fit the original (training) data well, but may do a poor job of
predicting new values. Lower branches, especially, may be strongly affected
by outliers. A simpler tree often avoids over-fitting. To find the best regression
tree, employing the techniques of resubstitution and cross-validation, use the
test method of the classregtree class.
9-98
10
Multivariate Methods
• “Introduction” on page 10-2
• “Multidimensional Scaling” on page 10-3
• “Procrustes Analysis” on page 10-14
• “Feature Selection” on page 10-23
• “Feature Transformation” on page 10-28
10 Multivariate Methods
Introduction
Large, high-dimensional data sets are common in the modern era
of computer-based instrumentation and electronic data storage.
High-dimensional data present many challenges for statistical visualization,
analysis, and modeling.
Data visualization, of course, is impossible beyond a few dimensions. As a
result, pattern recognition, data preprocessing, and model selection must
rely heavily on numerical methods.
A fundamental challenge in high-dimensional data analysis is the so-called
curse of dimensionality. Observations in a high-dimensional space are
necessarily sparser and less representative than those in a low-dimensional
space. In higher dimensions, data over-represent the edges of a sampling
distribution, because regions of higher-dimensional space contain the majority
of their volume near the surface. (A d-dimensional spherical shell has a
volume, relative to the total volume of the sphere, that approaches 1 as d
approaches infinity.) In high dimensions, typical data points at the interior of
a distribution are sampled less frequently.
Often, many of the dimensions in a data set—the measured features—are
not useful in producing a model. Features may be irrelevant or redundant.
Regression and classification algorithms may require large amounts of
storage and computation time to process raw data, and even if the algorithms
are successful the resulting models may contain an incomprehensible number
of terms.
Because of these challenges, multivariate statistical methods often begin with
some type of dimension reduction, in which data are approximated by points
in a lower-dimensional space. Dimension reduction is the goal of the methods
presented in this chapter. Dimension reduction often leads to simpler models
and fewer measured variables, with consequent benefits when measurements
are expensive and visualization is important.
10-2
Multidimensional Scaling
Multidimensional Scaling
In this section...
“Introduction” on page 10-3
“Classical Multidimensional Scaling” on page 10-3
“Nonclassical Multidimensional Scaling” on page 10-8
“Nonmetric Multidimensional Scaling” on page 10-10
Introduction
One of the most important goals in visualizing data is to get a sense of how
near or far points are from each other. Often, you can do this with a scatter
plot. However, for some analyses, the data that you have might not be in
the form of points at all, but rather in the form of pairwise similarities or
dissimilarities between cases, observations, or subjects. There are no points
to plot.
Even if your data are in the form of points rather than pairwise distances,
a scatter plot of those data might not be useful. For some kinds of data,
the relevant way to measure how near two points are might not be their
Euclidean distance. While scatter plots of the raw data make it easy to
compare Euclidean distances, they are not always useful when comparing
other kinds of inter-point distances, city block distance for example, or even
more general dissimilarities. Also, with a large number of variables, it is very
difficult to visualize distances unless the data can be represented in a small
number of dimensions. Some sort of dimension reduction is usually necessary.
Multidimensional scaling (MDS) is a set of methods that address all these
problems. MDS allows you to visualize how near points are to each other
for many kinds of distance or dissimilarity metrics and can produce a
representation of your data in a small number of dimensions. MDS does not
require raw data, but only a matrix of pairwise distances or dissimilarities.
Classical Multidimensional Scaling
• “Introduction” on page 10-4
10-3
10 Multivariate Methods
• “Example: Multidimensional Scaling” on page 10-6
Introduction
The function cmdscale performs classical (metric) multidimensional scaling,
also known as principal coordinates analysis. cmdscale takes as an input a
matrix of inter-point distances and creates a configuration of points. Ideally,
those points are in two or three dimensions, and the Euclidean distances
between them reproduce the original distance matrix. Thus, a scatter plot
of the points created by cmdscale provides a visual representation of the
original distances.
As a very simple example, you can reconstruct a set of points from only their
inter-point distances. First, create some four dimensional points with a small
component in their fourth coordinate, and reduce them to distances.
X = [ normrnd(0,1,10,3), normrnd(0,.1,10,1) ];
D = pdist(X,'euclidean');
Next, use cmdscale to find a configuration with those inter-point distances.
cmdscale accepts distances as either a square matrix, or, as in this example,
in the vector upper-triangular form produced by pdist.
[Y,eigvals] = cmdscale(D);
cmdscale produces two outputs. The first output, Y, is a matrix containing the
reconstructed points. The second output, eigvals, is a vector containing the
sorted eigenvalues of what is often referred to as the “scalar product matrix,”
which, in the simplest case, is equal to Y*Y'. The relative magnitudes of those
eigenvalues indicate the relative contribution of the corresponding columns of
Y in reproducing the original distance matrix D with the reconstructed points.
format short g
[eigvals eigvals/max(abs(eigvals))]
ans =
12.623 1
4.3699 0.34618
1.9307 0.15295
0.025884 0.0020505
1.7192e-015 1.3619e-016
6.8727e-016 5.4445e-017
10-4
Multidimensional Scaling
4.4367e-017 3.5147e-018
-9.2731e-016 -7.3461e-017
-1.327e-015 -1.0513e-016
-1.9232e-015 -1.5236e-016
If eigvals contains only positive and zero (within round-off error) eigenvalues,
the columns of Y corresponding to the positive eigenvalues provide an exact
reconstruction of D, in the sense that their inter-point Euclidean distances,
computed using pdist, for example, are identical (within round-off) to the
values in D.
maxerr4 = max(abs(D - pdist(Y))) % exact reconstruction
maxerr4 =
2.6645e-015
If two or three of the eigenvalues in eigvals are much larger than the rest,
then the distance matrix based on the corresponding columns of Y nearly
reproduces the original distance matrix D. In this sense, those columns
form a lower-dimensional representation that adequately describes the
data. However it is not always possible to find a good low-dimensional
reconstruction.
% good reconstruction in 3D
maxerr3 = max(abs(D - pdist(Y(:,1:3))))
maxerr3 =
0.029728
% poor reconstruction in 2D
maxerr2 = max(abs(D - pdist(Y(:,1:2))))
maxerr2 =
0.91641
The reconstruction in three dimensions reproduces D very well, but the
reconstruction in two dimensions has errors that are of the same order of
magnitude as the largest values in D.
max(max(D))
ans =
3.4686
10-5
10 Multivariate Methods
Often, eigvals contains some negative eigenvalues, indicating that the
distances in D cannot be reproduced exactly. That is, there might not be any
configuration of points whose inter-point Euclidean distances are given by
D. If the largest negative eigenvalue is small in magnitude relative to the
largest positive eigenvalues, then the configuration returned by cmdscale
might still reproduce D well.
Example: Multidimensional Scaling
Given only the distances between 10 US cities, cmdscale can construct a map
of those cities. First, create the distance matrix and pass it to cmdscale.
In this example,D is a full distance matrix: it is square and symmetric, has
positive entries off the diagonal, and has zeros on the diagonal.
cities = ...
{'Atl','Chi','Den','Hou','LA','Mia','NYC','SF','Sea','WDC'};
D = [ 0 587 1212 701 1936 604 748 2139 2182 543;
587 0 920 940 1745 1188 713 1858 1737 597;
1212 920 0 879 831 1726 1631 949 1021 1494;
701 940 879 0 1374 968 1420 1645 1891 1220;
1936 1745 831 1374 0 2339 2451 347 959 2300;
604 1188 1726 968 2339 0 1092 2594 2734 923;
748 713 1631 1420 2451 1092 0 2571 2408 205;
2139 1858 949 1645 347 2594 2571 0 678 2442;
2182 1737 1021 1891 959 2734 2408 678 0 2329;
543 597 1494 1220 2300 923 205 2442 2329 0];
[Y,eigvals] = cmdscale(D);
Next, look at the eigenvalues returned by cmdscale. Some of these are
negative, indicating that the original distances are not Euclidean. This is
because of the curvature of the earth.
format short g
[eigvals eigvals/max(abs(eigvals))]
ans =
9.5821e+006 1
1.6868e+006 0.17604
8157.3 0.0008513
1432.9 0.00014954
508.67 5.3085e-005
25.143 2.624e-006
10-6
Multidimensional Scaling
5.3394e-010 5.5722e-017
-897.7 -9.3685e-005
-5467.6 -0.0005706
-35479 -0.0037026
However, in this case, the two largest positive eigenvalues are much larger
in magnitude than the remaining eigenvalues. So, despite the negative
eigenvalues, the first two coordinates of Y are sufficient for a reasonable
reproduction of D.
Dtriu = D(find(tril(ones(10),-1)))';
maxrelerr = max(abs(Dtriu-pdist(Y(:,1:2))))./max(Dtriu)
maxrelerr =
0.0075371
Here is a plot of the reconstructed city locations as a map. The orientation of
the reconstruction is arbitrary. In this case, it happens to be close to, although
not exactly, the correct orientation.
plot(Y(:,1),Y(:,2),'.')
text(Y(:,1)+25,Y(:,2),cities)
xlabel('Miles')
ylabel('Miles')
10-7
10 Multivariate Methods
Nonclassical Multidimensional Scaling
The function mdscale performs nonclassical multidimensional scaling. As
with cmdcale, you use mdscale either to visualize dissimilarity data for which
no “locations” exist, or to visualize high-dimensional data by reducing its
dimensionality. Both functions take a matrix of dissimilarities as an input
and produce a configuration of points. However, mdscale offers a choice of
different criteria to construct the configuration, and allows missing data and
weights.
For example, the cereal data include measurements on 10 variables describing
breakfast cereals. You can use mdscale to visualize these data in two
dimensions. First, load the data. For clarity, this example code selects a
subset of 22 of the observations.
load cereal.mat
X = [Calories Protein Fat Sodium Fiber ...
Carbo Sugars Shelf Potass Vitamins];
10-8
Multidimensional Scaling
X = X(strmatch('G',Mfg),:); % Take a subset from a
% single manufacturer
size(X)
ans =
22 10
Then use pdist to transform the 10-dimensional data into dissimilarities.
The output from pdist is a symmetric dissimilarity matrix, stored as a vector
containing only the (23*22/2) elements in its upper triangle.
dissimilarities = pdist(zscore(X),'cityblock');
size(dissimilarities)
ans =
1 231
This example code first standardizes the cereal data, and then uses city block
distance as a dissimilarity. The choice of transformation to dissimilarities is
application-dependent, and the choice here is only for simplicity. In some
applications, the original data are already in the form of dissimilarities.
Next, use mdscale to perform metric MDS. Unlike cmdscale, you must
specify the desired number of dimensions, and the method to use to construct
the output configuration. For this example, use two dimensions. The metric
STRESS criterion is a common method for computing the output; for other
choices, see the mdscale reference page in the online documentation. The
second output from mdscale is the value of that criterion evaluated for the
output configuration. It measures the how well the inter-point distances of
the output configuration approximate the original input dissimilarities:
[Y,stress] =...
mdscale(dissimilarities,2,'criterion','metricstress');
stress
stress =
0.1856
A scatterplot of the output from mdscale represents the original
10-dimensional data in two dimensions, and you can use the gname function to
label selected points:
plot(Y(:,1),Y(:,2),'o','LineWidth',2);
gname(Name(strmatch('G',Mfg)))
10-9
10 Multivariate Methods
Nonmetric Multidimensional Scaling
Metric multidimensional scaling creates a configuration of points whose
inter-point distances approximate the given dissimilarities. This is sometimes
too strict a requirement, and non-metric scaling is designed to relax it a bit.
Instead of trying to approximate the dissimilarities themselves, non-metric
scaling approximates a nonlinear, but monotonic, transformation of them.
Because of the monotonicity, larger or smaller distances on a plot of the
output will correspond to larger or smaller dissimilarities, respectively.
However, the nonlinearity implies that mdscale only attempts to preserve the
ordering of dissimilarities. Thus, there may be contractions or expansions of
distances at different scales.
You use mdscale to perform nonmetric MDS in much the same way as for
metric scaling. The nonmetric STRESS criterion is a common method for
computing the output; for more choices, see the mdscale reference page in
the online documentation. As with metric scaling, the second output from
10-10
Multidimensional Scaling
mdscale is the value of that criterion evaluated for the output configuration.
For nonmetric scaling, however, it measures the how well the inter-point
distances of the output configuration approximate the disparities. The
disparities are returned in the third output. They are the transformed values
of the original dissimilarities:
[Y,stress,disparities] = ...
mdscale(dissimilarities,2,'criterion','stress');
stress
stress =
0.1562
To check the fit of the output configuration to the dissimilarities, and to
understand the disparities, it helps to make a Shepard plot:
distances = pdist(Y);
[dum,ord] = sortrows([disparities(:) dissimilarities(:)]);
plot(dissimilarities,distances,'bo', ...
dissimilarities(ord),disparities(ord),'r.-', ...
[0 25],[0 25],'k-')
xlabel('Dissimilarities')
ylabel('Distances/Disparities')
legend({'Distances' 'Disparities' '1:1 Line'},...
'Location','NorthWest');
10-11
10 Multivariate Methods
This plot shows that mdscale has found a configuration of points in two
dimensions whose inter-point distances approximates the disparities, which
in turn are a nonlinear transformation of the original dissimilarities. The
concave shape of the disparities as a function of the dissimilarities indicates
that fit tends to contract small distances relative to the corresponding
dissimilarities. This may be perfectly acceptable in practice.
mdscale uses an iterative algorithm to find the output configuration, and
the results can often depend on the starting point. By default, mdscale
uses cmdscale to construct an initial configuration, and this choice often
leads to a globally best solution. However, it is possible for mdscale to
stop at a configuration that is a local minimum of the criterion. Such
cases can be diagnosed and often overcome by running mdscale multiple
times with different starting points. You can do this using the 'start'
and 'replicates' parameters. The following code runs five replicates of
10-12
Multidimensional Scaling
MDS, each starting at a different randomly-chosen initial configuration.
The criterion value is printed out for each replication; mdscale returns the
configuration with the best fit.
opts = statset('Display','final');
[Y,stress] =...
mdscale(dissimilarities,2,'criterion','stress',...
'start','random','replicates',5,'Options',opts);
90 iterations, Final stress criterion = 0.156209
100 iterations, Final stress criterion = 0.195546
116 iterations, Final stress criterion = 0.156209
85 iterations, Final stress criterion = 0.156209
106 iterations, Final stress criterion = 0.17121
Notice that mdscale finds several different local solutions, some of which
do not have as low a stress value as the solution found with the cmdscale
starting point.
10-13
10 Multivariate Methods
Procrustes Analysis
In this section...
“Comparing Landmark Data” on page 10-14
“Data Input” on page 10-14
“Preprocessing Data for Accurate Results” on page 10-15
“Example: Comparing Handwritten Shapes” on page 10-16
Comparing Landmark Data
The procrustes function analyzes the distribution of a set of shapes using
Procrustes analysis. This analysis method matches landmark data (geometric
locations representing significant features in a given shape) to calculate the
best shape-preserving Euclidian transformations. These transformations
minimize the differences in location between compared landmark data.
Procrustes analysis is also useful in conjunction with multidimensional
scaling. In “Example: Multidimensional Scaling” on page 10-6 there is an
observation that the orientation of the reconstructed points is arbitrary. Two
different applications of multidimensional scaling could produce reconstructed
points that are very similar in principle, but that look different because they
have different orientations. The procrustes function transforms one set of
points to make them more comparable to the other.
Data Input
The procrustes function takes two matrices as input:
• The target shape matrix X has dimension n × p, where n is the number
of landmarks in the shape and p is the number of measurements per
landmark.
• The comparison shape matrix Y has dimension n × q with q ≤ p. If there
are fewer measurements per landmark for the comparison shape than
the target shape (q < p), the function adds columns of zeros to Y, yielding
an n × p matrix.
The equation to obtain the transformed shape, Z, is
10-14
Procrustes Analysis
Z = bYT + c (10-1)
where:
• b is a scaling factor that stretches (b > 1) or shrinks (b < 1) the points.
• T is the orthogonal rotation and reflection matrix.
• c is a matrix with constant values in each column, used to shift the points.
The procrustes function chooses b, T, and c to minimize the distance between
the target shape X and the transformed shape Z as measured by the least
squares criterion:
n p
∑∑(X
i =1 j =1
ij − Z ij ) 2
Preprocessing Data for Accurate Results
Procrustes analysis is appropriate when all p measurement dimensions have
similar scales. The analysis would be inaccurate, for example, if the columns
of Z had different scales:
• The first column is measured in milliliters ranging from 2,000 to 6,000.
• The second column is measured in degrees Celsius ranging from 10 to 25.
• The third column is measured in kilograms ranging from 50 to 230.
In such cases, standardize your variables by:
1 Subtracting the sample mean from each variable.
2 Dividing each resultant variable by its sample standard deviation.
Use the zscore function to perform this standardization.
10-15
10 Multivariate Methods
Example: Comparing Handwritten Shapes
In this example, use Procrustes analysis to compare two handwritten number
threes. Visually and analytically explore the effects of forcing size and
reflection changes as follows:
• “Step 1: Load and Display the Original Data” on page 10-16
• “Step 2: Calculate the Best Transformation” on page 10-17
• “Step 3: Examine the Similarity of the Two Shapes” on page 10-18
• “Step 4: Restrict the Form of the Transformations” on page 10-20
Step 1: Load and Display the Original Data
Input landmark data for two handwritten number threes:
A = [11 39;17 42;25 42;25 40;23 36;19 35;30 34;35 29;...
30 20;18 19];
B = [15 31;20 37;30 40;29 35;25 29;29 31;31 31;35 20;...
29 10;25 18];
Create X and Y from A and B, moving B to the side to make each shape more
visible:
X = A;
Y = B + repmat([25 0], 10,1);
Plot the shapes, using letters to designate the landmark points. Lines in the
figure join the points to indicate the drawing path of each shape.
plot(X(:,1), X(:,2),'r-', Y(:,1), Y(:,2),'b-');
text(X(:,1), X(:,2),('abcdefghij')')
text(Y(:,1), Y(:,2),('abcdefghij')')
legend('X = Target','Y = Comparison','location','SE')
set(gca,'YLim',[0 55],'XLim',[0 65]);
10-16
Procrustes Analysis
Step 2: Calculate the Best Transformation
Use Procrustes analysis to find the transformation that minimizes distances
between landmark data points.
Call procrustes as follows:
[d, Z, tr] = procrustes(X,Y);
The outputs of the function are:
• d – A standardized dissimilarity measure.)
• Z – A matrix of the transformed landmarks.
• tr – A structure array of the computed transformation with fields T, b, and
c which correspond to the transformation equation, Equation 10-1.
10-17
10 Multivariate Methods
Visualize the transformed shape, Z, using a dashed blue line:
plot(X(:,1), X(:,2),'r-', Y(:,1), Y(:,2),'b-',...
Z(:,1),Z(:,2),'b:');
text(X(:,1), X(:,2),('abcdefghij')')
text(Y(:,1), Y(:,2),('abcdefghij')')
text(Z(:,1), Z(:,2),('abcdefghij')')
legend('X = Target','Y = Comparison',...
'Z = Transformed','location','SW')
set(gca,'YLim',[0 55],'XLim',[0 65]);
Step 3: Examine the Similarity of the Two Shapes
Use two different numerical values to assess the similarity of the target shape
and the transformed shape.
10-18
Procrustes Analysis
Dissimilarity Measure d. The dissimilarity measure d gives a number
between 0 and 1 describing the difference between the target shape and the
transformed shape. Values near 0 imply more similar shapes, while values
near 1 imply dissimilarity. For this example:
d =
0.1502
The small value of d in this case shows that the two shapes are similar.
procrustes calculates d by comparing the sum of squared deviations between
the set of points with the sum of squared deviations of the original points from
their column means:
numerator = sum(sum((X-Z).^2))
numerator =
166.5321
denominator = sum(sum(bsxfun(@minus,X,mean(X)).^2))
denominator =
1.1085e+003
ratio = numerator/denominator
ratio =
0.1502
Note The resulting measure d is independent of the scale of the size of
the shapes and takes into account only the similarity of landmark data.
“Examining the Scaling Measure b” on page 10-19 shows how to examine the
size similarity of the shapes.
Examining the Scaling Measure b. The target and comparison threes in
the previous figure visually show that the two numbers are of a similar size.
The closeness of calculated value of the scaling factor b to 1 supports this
observation as well:
10-19
10 Multivariate Methods
tr.b
ans =
0.9291
The sizes of the target and comparison shapes appear similar. This visual
impression is reinforced by the value of b = 0.93, which implies that the best
transformation results in shrinking the comparison shape by a factor .93
(only 7%).
Step 4: Restrict the Form of the Transformations
Explore the effects of manually adjusting the scaling and reflection
coefficients.
Fixing the Scaling Factor b = 1. Force b to equal 1 (set 'Scaling' to
false) to examine the amount of dissimilarity in size of the target and
transformed figures:
ds = procrustes(X,Y,'Scaling',false)
ds =
0.1552
In this case, setting 'Scaling' to false increases the calculated value of
d only 0.0049, which further supports the similarity in the size of the two
number threes. A larger increase in d would have indicated a greater size
discrepancy.
Forcing a Reflection in the Transformation. This example requires only a
rotation, not a reflection, to align the shapes. You can show this by observing
that the determinant of the matrix T is 1 in this analysis:
det(tr.T)
ans =
1.0000
If you need a reflection in the transformation, the determinant of T is -1. You
can force a reflection into the transformation as follows:
[dr,Zr,trr] = procrustes(X,Y,'Reflection',true);
dr
dr =
10-20
Procrustes Analysis
0.8130
The d value increases dramatically, indicating that a forced reflection leads
to a poor transformation of the landmark points. A plot of the transformed
shape shows a similar result:
• The landmark data points are now further away from their target
counterparts.
• The transformed three is now an undesirable mirror image of the target
three.
plot(X(:,1), X(:,2),'r-', Y(:,1), Y(:,2),'b-',...
Zr(:,1),Zr(:,2),'b:');
text(X(:,1), X(:,2),('abcdefghij')')
text(Y(:,1), Y(:,2),('abcdefghij')')
text(Zr(:,1), Zr(:,2),('abcdefghij')')
legend('X = Target','Y = Comparison',...
'Z = Transformed','location','SW')
set(gca,'YLim',[0 55],'XLim',[0 65]);
10-21
10 Multivariate Methods
It appears that the shapes might be better matched if you flipped the
transformed shape upside down. Flipping the shapes would make the
transformation even worse, however, because the landmark data points
would be further away from their target counterparts. From this example,
it is clear that manually adjusting the scaling and reflection parameters is
generally not optimal.
10-22
Feature Selection
Feature Selection
In this section...
“Introduction” on page 10-23
“Sequential Feature Selection” on page 10-23
Introduction
Feature selection reduces the dimensionality of data by selecting only a subset
of measured features (predictor variables) to create a model. Selection criteria
usually involve the minimization of a specific measure of predictive error for
models fit to different subsets. Algorithms search for a subset of predictors
that optimally model measured responses, subject to constraints such as
required or excluded features and the size of the subset.
Feature selection is preferable to feature transformation when the original
units and meaning of features are important and the modeling goal is to
identify an influential subset. When categorical features are present, and
numerical transformations are inappropriate, feature selection becomes the
primary means of dimension reduction.
Sequential Feature Selection
• “Introduction” on page 10-23
• “Example: Sequential Feature Selection” on page 10-24
Introduction
A common method of feature selection is sequential feature selection. This
method has two components:
• An objective function, called the criterion, which the method seeks to
minimize over all feasible feature subsets. Common criteria are mean
squared error (for regression models) and misclassification rate (for
classification models).
• A sequential search algorithm, which adds or removes features from a
candidate subset while evaluating the criterion. Since an exhaustive
10-23
10 Multivariate Methods
comparison of the criterion value at all 2n subsets of an n-feature data set
is typically infeasible (depending on the size of n and the cost of objective
calls), sequential searches move in only one direction, always growing or
always shrinking the candidate set.
The method has two variants:
• Sequential forward selection (SFS), in which features are sequentially
added to an empty candidate set until the addition of further features does
not decrease the criterion.
• Sequential backward selection (SBS), in which features are sequentially
removed from a full candidate set until the removal of further features
increase the criterion.
Stepwise regression is a sequential feature selection technique designed
specifically for least-squares fitting. The functions stepwise and stepwisefit
make use of optimizations that are only possible with least-squares criteria.
Unlike generalized sequential feature selection, stepwise regression may
remove features that have been added or add features that have been removed.
The Statistics Toolbox function sequentialfs carries out sequential feature
selection. Input arguments include predictor and response data and a function
handle to a file implementing the criterion function. Optional inputs allow
you to specify SFS or SBS, required or excluded features, and the size of the
feature subset. The function calls cvpartition and crossval to evaluate the
criterion at different candidate sets.
Example: Sequential Feature Selection
For example, consider a data set with 100 observations of 10 predictors.
As described in “Example: Generalized Linear Models” on page 9-53, the
following generates random data from a logistic model, with a binomial
distribution of responses at each set of values for the predictors. Some
coefficients are set to zero so that not all of the predictors affect the response:
n = 100;
m = 10;
X = rand(n,m);
b = [1 0 0 2 .5 0 0 0.1 0 1];
Xb = X*b';
10-24
Feature Selection
p = 1./(1+exp(-Xb));
N = 50;
y = binornd(N,p);
The glmfit function fits a logistic model to the data:
Y = [y N*ones(size(y))];
[b0,dev0,stats0] = glmfit(X,Y,'binomial');
% Display coefficient estimates and their standard errors:
model0 = [b0 stats0.se]
model0 =
0.3115 0.2596
0.9614 0.1656
-0.1100 0.1651
-0.2165 0.1683
1.9519 0.1809
0.5683 0.2018
-0.0062 0.1740
0.0651 0.1641
-0.1034 0.1685
0.0017 0.1815
0.7979 0.1806
% Display the deviance of the fit:
dev0
dev0 =
101.2594
This is the full model, using all of the features (and an initial constant term).
Sequential feature selection searches for a subset of the features in the full
model with comparative predictive power.
First, you must specify a criterion for selecting the features. The following
function, which calls glmfit and returns the deviance of the fit (a
generalization of the residual sum of squares) is a useful criterion in this case:
function dev = critfun(X,Y)
[b,dev] = glmfit(X,Y,'binomial');
10-25
10 Multivariate Methods
You should create this function as a file on the MATLAB path.
The function sequentialfs performs feature selection, calling the criterion
function via a function handle:
maxdev = chi2inv(.95,1);
opt = statset('display','iter',...
'TolFun',maxdev,...
'TolTypeFun','abs');
inmodel = sequentialfs(@critfun,X,Y,...
'cv','none',...
'nullmodel',true,...
'options',opt,...
'direction','forward');
Start forward sequential feature selection:
Initial columns included: none
Columns that can not be included: none
Step 1, used initial columns, criterion value 309.118
Step 2, added column 4, criterion value 180.732
Step 3, added column 1, criterion value 138.862
Step 4, added column 10, criterion value 114.238
Step 5, added column 5, criterion value 103.503
Final columns included: 1 4 5 10
The iterative display shows a decrease in the criterion value as each new
feature is added to the model. The final result is a reduced model with only
four of the original ten features: columns 1, 4, 5, and 10 of X. These features
are indicated in the logical vector inmodel returned by sequentialfs.
The deviance of the reduced model is higher than for the full model, but
the addition of any other single feature would not decrease the criterion
by more than the absolute tolerance, maxdev, set in the options structure.
Adding a feature with no effect reduces the deviance by an amount that has
a chi-square distribution with one degree of freedom. Adding a significant
feature results in a larger change. By setting maxdev to chi2inv(.95,1), you
instruct sequentialfs to continue adding features so long as the change in
deviance is more than would be expected by random chance.
10-26
Feature Selection
The reduced model (also with an initial constant term) is:
[b,dev,stats] = glmfit(X(:,in),Y,'binomial');
% Display coefficient estimates and their standard errors:
model = [b stats.se]
model =
0.0784 0.1642
1.0040 0.1592
1.9459 0.1789
0.6134 0.1872
0.8245 0.1730
10-27
10 Multivariate Methods
Feature Transformation
In this section...
“Introduction” on page 10-28
“Nonnegative Matrix Factorization” on page 10-28
“Principal Component Analysis” on page 10-31
“Factor Analysis” on page 10-45
Introduction
Feature transformation is a group of methods that create new features
(predictor variables). The methods are useful for dimension reduction when
the transformed features have a descriptive power that is more easily ordered
than the original features. In this case, less descriptive features can be
dropped from consideration when building models.
Feature transformation methods are contrasted with the methods presented
in “Feature Selection” on page 10-23, where dimension reduction is achieved
by computing an optimal subset of predictive features measured in the
original data.
The methods presented in this section share some common methodology.
Their goals, however, are essentially different:
• Nonnegative matrix factorization is used when model terms must represent
nonnegative quantities, such as physical quantities.
• Principal component analysis is used to summarize data in fewer
dimensions, for example, to visualize it.
• Factor analysis is used to build explanatory models of data correlations.
Nonnegative Matrix Factorization
• “Introduction” on page 10-29
• “Example: Nonnegative Matrix Factorization” on page 10-29
10-28
Feature Transformation
Introduction
Nonnegative matrix factorization (NMF) is a dimension-reduction technique
based on a low-rank approximation of the feature space. Besides providing
a reduction in the number of features, NMF guarantees that the features
are nonnegative, producing additive models that respect, for example, the
nonnegativity of physical quantities.
Given a nonnegative m-by-n matrix X and a positive integer k < min(m,n),
NMF finds nonnegative m-by-k and k-by-n matrices W and H, respectively,
that minimize the norm of the difference X – WH. W and H are thus
approximate nonnegative factors of X.
The k columns of W represent transformations of the variables in X; the k
rows of H represent the coefficients of the linear combinations of the original
n variables in X that produce the transformed variables in W. Since k is
generally smaller than the rank of X, the product WH provides a compressed
approximation of the data in X. A range of possible values for k is often
suggested by the modeling context.
The Statistics Toolbox function nnmf carries out nonnegative matrix
factorization. nnmf uses one of two iterative algorithms that begin with
random initial values for W and H. Because the norm of the residual X
– WH may have local minima, repeated calls to nnmf may yield different
factorizations. Sometimes the algorithm converges to a solution of lower rank
than k, which may indicate that the result is not optimal.
Example: Nonnegative Matrix Factorization
For example, consider the five predictors of biochemical oxygen demand in the
data set moore.mat:
load moore
X = moore(:,1:5);
The following uses nnmf to compute a rank-two approximation of X with a
multiplicative update algorithm that begins from five random initial values
for W and H:
opt = statset('MaxIter',10,'Display','final');
[W0,H0] = nnmf(X,2,'replicates',5,...
'options',opt,...
10-29
10 Multivariate Methods
'algorithm','mult');
rep iteration rms resid |delta x|
1 10 358.296 0.00190554
2 10 78.3556 0.000351747
3 10 230.962 0.0172839
4 10 326.347 0.00739552
5 10 361.547 0.00705539
Final root mean square residual = 78.3556
The 'mult' algorithm is sensitive to initial values, which makes it a good
choice when using 'replicates' to find W and H from multiple random
starting values.
Now perform the factorization using an alternating least-squares algorithm,
which converges faster and more consistently. Run 100 times more iterations,
beginning from the initial W0 and H0 identified above:
opt = statset('Maxiter',1000,'Display','final');
[W,H] = nnmf(X,2,'w0',W0,'h0',H0,...
'options',opt,...
'algorithm','als');
rep iteration rms resid |delta x|
1 3 77.5315 3.52673e-005
Final root mean square residual = 77.5315
The two columns of W are the transformed predictors. The two rows of H give
the relative contributions of each of the five predictors in X to the predictors
in W:
H
H =
0.0835 0.0190 0.1782 0.0072 0.9802
0.0558 0.0250 0.9969 0.0085 0.0497
The fifth predictor in X (weight 0.9802) strongly influences the first predictor
in W. The third predictor in X (weight 0.9969) strongly influences the second
predictor in W.
Visualize the relative contributions of the predictors in X with a biplot,
showing the data and original variables in the column space of W:
10-30
Feature Transformation
biplot(H','scores',W,'varlabels',{'','','v3','','v5'});
axis([0 1.1 0 1.1])
xlabel('Column 1')
ylabel('Column 2')
Principal Component Analysis 使variance最大化，不須知道資料的scale
• “Introduction” on page 10-31
• “Example: Principal Component Analysis” on page 10-33
Introduction 內在的 多重
One of the difficulties inherent in multivariate statistics is the problem of
可見的 visualizing data that has many variables. The MATLAB function plot
displays a graph of the relationship between two variables. The plot3
and surf commands display different three-dimensional views. But when
10-31
10 Multivariate Methods
超過三個變數就很難從肉眼看出他們的關係
there are more than three variables, it is more difficult to visualize their
relationships.
Fortunately, in data sets with many variables, groups of variables often
move together. One reason for this is that more than one variable might be
measuring the same driving principle governing the behavior of the system.
In many systems there are only a few such driving forces. But an abundance
of instrumentation enables you to measure dozens of system variables. When
this happens, you can take advantage of this redundancy of information.
You can simplify the problem by replacing a group of variables with a single
new variable.
分量上 精確的
Principal component analysis is a quantitatively rigorous method for achieving
this simplification. The method generates a new set of variables, called
principal components. Each principal component is a linear combination of
the original variables. All the principal components are orthogonal to each
other, so there is no redundant information. The principal components as a
whole form an orthogonal basis for the space of the data.
There are an infinite number of ways to construct an orthogonal basis for
several columns of data. What is so special about the principal component
basis?
單軸
The first principal component is a single axis in space. When you project
投影到此單軸會產生新的變數
each observation on that axis, the resulting values form a new variable. And
這些可變變數是第一主成分 the variance of this variable is the maximum among all possible choices of
所有可能的選擇的最大值 the first axis.
The second principal component is another axis in space, perpendicular to
the first. Projecting the observations on this axis generates another new
variable. The variance of this variable is the maximum among all possible
choices of this second axis.
The full set of principal components is as large as the original set of variables.
But it is commonplace for the sum of the variances of the first few principal
超過
components to exceed 80% of the total variance of the original data. By
examining plots of these few new variables, researchers often develop a
deeper understanding of the driving forces that generated the original data.
10-32
Feature Transformation
You can use the function princomp to find the principal components. To use
princomp, you need to have the actual measured data you want to analyze.
However, if you lack the actual data, but have the sample covariance or
correlation matrix for the data, you can still use the function pcacov to
perform a principal components analysis. See the reference page for pcacov
for a description of its inputs and outputs.
Example: Principal Component Analysis
• “Computing Components” on page 10-33
• “Component Coefficients” on page 10-36
• “Component Scores” on page 10-36
• “Component Variances” on page 10-40
• “Hotelling’s T2” on page 10-42
• “Visualizing the Results” on page 10-42
Computing Components. Consider a sample application that uses nine
different indices of the quality of life in 329 U.S. cities. These
消遣
are climate,天氣
housing, health, crime, transportation, education, arts, recreation, and
economics. For each index, higher is better. For example, a higher index
for crime means a lower crime rate.
Start by loading the data in cities.mat.
load cities
whos
Name Size Bytes Class
categories 9x14 252 char array
names 329x43 28294 char array
ratings 329x9 23688 double array
The whos command generates a table of information about all the variables
in the workspace.
The cities data set contains three variables:
• categories, a string matrix containing the names of the indices
10-33
10 Multivariate Methods
• names, a string matrix containing the 329 city names
• ratings, the data matrix with 329 rows and 9 columns
The categories variable has the following values:
categories
categories =
climate
housing
health
crime
transportation
education
arts
recreation
economics
The first five rows of names are
first5 = names(1:5,:)
first5 =
Abilene, TX
Akron, OH
Albany, GA
Albany-Troy, NY
Albuquerque, NM
To get a quick impression of the ratings data, make a box plot.
boxplot(ratings,'orientation','horizontal','labels',categories)
This command generates the plot below. Note that there is substantially
more variability in the ratings of the arts and housing than in the ratings
of crime and climate.
10-34
Feature Transformation
Ordinarily you might also graph pairs of the original variables, but there are
36 two-variable plots. Perhaps principal components analysis can reduce the
number of variables you need to consider.
Sometimes it makes sense to compute principal components for raw data. This
is appropriate when all the variables are in the same units. Standardizing the
data is often preferable when the variables are in different units or when the
牢固的
variance of the different columns is substantial (as in this case).
You can standardize the data by dividing each column by its standard
deviation.
stdr = std(ratings);
sr = ratings./repmat(stdr,329,1);
Now you are ready to find the principal components.
10-35
10 Multivariate Methods
[coefs,scores,variances,t2] = princomp(sr);
The following sections explain the four outputs from princomp.
Component Coefficients. The first output of the princomp function, coefs,
contains the coefficients of the linear combinations of the original variables
that generate the principal components. The coefficients are also known as
loadings.
The first three principal component coefficient vectors are:
c3 = coefs(:,1:3)
c3 =
0.2064 0.2178 -0.6900
0.3565 0.2506 -0.2082
0.4602 -0.2995 -0.0073
0.2813 0.3553 0.1851
0.3512 -0.1796 0.1464
0.2753 -0.4834 0.2297
0.4631 -0.1948 -0.0265
0.3279 0.3845 -0.0509
0.1354 0.4713 0.6073
The largest coefficients in the first column (first principal component) are
the third and seventh elements, corresponding to the variables health and
arts. All the coefficients of the first principal component have the same sign,
making it a weighted average of all the original variables.
The principal components are unit length and orthogonal:
I = c3'*c3
I =
1.0000 -0.0000 -0.0000
-0.0000 1.0000 -0.0000
-0.0000 -0.0000 1.0000
Component Scores. The second output, scores, contains the coordinates
of the original data in the new coordinate system defined by the principal
components. This output is the same size as the input data matrix.
10-36
Feature Transformation
A plot of the first two columns of scores shows the ratings data projected
onto the first two principal components. princomp computes the scores to
have mean zero.
plot(scores(:,1),scores(:,2),'+')
xlabel('1st Principal Component')
ylabel('2nd Principal Component')
Note the outlying points in the right half of the plot.
While it is possible to create a three-dimensional plot using three columns
of scores, the examples in this section create two-dimensional plots, which
are easier to describe.
The function gname is useful for graphically identifying a few points in a plot
like this. You can call gname with a string matrix containing as many case
10-37
10 Multivariate Methods
labels as points in the plot. The string matrix names works for labeling points
with the city names.
gname(names)
Move your cursor over the plot and click once near each point in the right
half. As you click each point, it is labeled with the proper row from the names
string matrix. Here is the plot after a few clicks:
When you are finished labeling points, press the Return key.
The labeled cities are some of the biggest population centers in the United
States. They are definitely different from the remainder of the data, so
perhaps they should be considered separately. To remove the labeled cities
from the data, first identify their corresponding row numbers as follows:
10-38
Feature Transformation
1 Close the plot window.
2 Redraw the plot by entering
plot(scores(:,1),scores(:,2),'+')
xlabel('1st Principal Component');
ylabel('2nd Principal Component');
3 Enter gname without any arguments.
4 Click near the points you labeled in the preceding figure. This labels the
points by their row numbers, as shown in the following figure.
Then you can create an index variable containing the row numbers of all
the metropolitan areas you choose.
10-39
10 Multivariate Methods
metro = [43 65 179 213 234 270 314];
names(metro,:)
ans =
Boston, MA
Chicago, IL
Los Angeles, Long Beach, CA
New York, NY
Philadelphia, PA-NJ
San Francisco, CA
Washington, DC-MD-VA
To remove these rows from the ratings matrix, enter the following.
rsubset = ratings;
nsubset = names;
nsubset(metro,:) = [];
rsubset(metro,:) = [];
size(rsubset)
ans =
322 9
Component Variances. The third output, variances, is a vector containing
the variance explained by the corresponding principal component. Each
column of scores has a sample variance equal to the corresponding element
of variances.
variances
variances =
3.4083
1.2140
1.1415
0.9209
0.7533
0.6306
0.4930
0.3180
0.1204
You can easily calculate the percent of the total variability explained by each
principal component.
10-40
Feature Transformation
percent_explained = 100*variances/sum(variances)
percent_explained =
37.8699
13.4886
12.6831
10.2324
8.3698
7.0062
5.4783
3.5338
1.3378
Use the pareto function to make a scree plot of the percent variability
explained by each principal component.
pareto(percent_explained)
xlabel('Principal Component')
ylabel('Variance Explained (%)')
10-41
10 Multivariate Methods
The preceding figure shows that the only clear break in the amount of
variance accounted for by each component is between the first and second
components. However, that component by itself explains less than 40% of the
variance, so more components are probably needed. You can see that the first
three principal components explain roughly two-thirds of the total variability
in the standardized ratings, so that might be a reasonable way to reduce the
dimensions in order to visualize the data.
Hotelling’s T2. The last output of the princomp function, t2, is Hotelling’s T2,
a statistical measure of the multivariate distance of each observation from
the center of the data set. This is an analytical way to find the most extreme
points in the data.
[st2, index] = sort(t2,'descend'); % Sort in descending order.
extreme = index(1)
extreme =
213
names(extreme,:)
ans =
New York, NY
It is not surprising that the ratings for New York are the furthest from the
average U.S. town.
Visualizing the Results. Use the biplot function to help visualize both
the principal component coefficients for each variable and the principal
component scores for each observation in a single plot. For example, the
following command plots the results from the principal components analysis
on the cities and labels each of the variables.
biplot(coefs(:,1:2), 'scores',scores(:,1:2),...
'varlabels',categories);
axis([-.26 1 -.51 .51]);
10-42
Feature Transformation
對component 1來說，
利用feature select 選擇arts & health最能代表coponent 1
Each of the nine variables is represented in this plot by a vector, and the
direction and length of the vector indicates how each variable contributes to
the two principal components in the plot. For example, you have seen that the
first principal component, represented in this biplot by the horizontal axis,
has positive coefficients for all nine variables. That corresponds to the nine
vectors directed into the right half of the plot. You have also seen that the
second principal component, represented by the vertical axis, has positive
coefficients for the variables education, health, arts, and transportation, and
negative coefficients for the remaining five variables. That corresponds to
vectors directed into the top and bottom halves of the plot, respectively. This
indicates that this component distinguishes between cities that have high
values for the first set of variables and low for the second, and cities that
have the opposite.
10-43
10 Multivariate Methods
The variable labels in this figure are somewhat crowded. You could either
leave out the VarLabels parameter when making the plot, or simply select
and drag some of the labels to better positions using the Edit Plot tool from
the figure window toolbar.
Each of the 329 observations is represented in this plot by a point, and
their locations indicate the score of each observation for the two principal
components in the plot. For example, points near the left edge of this plot
have the lowest scores for the first principal component. The points are
scaled to fit within the unit square, so only their relative locations may be
determined from the plot.
You can use the Data Cursor, in the Tools menu in the figure window, to
identify the items in this plot. By clicking on a variable (vector), you can read
off that variable’s coefficients for each principal component. By clicking on
an observation (point), you can read off that observation’s scores for each
principal component.
You can also make a biplot in three dimensions. This can be useful if the first
two principal coordinates do not explain enough of the variance in your data.
Selecting Rotate 3D in the Tools menu enables you to rotate the figure to
see it from different angles.
biplot(coefs(:,1:3), 'scores',scores(:,1:3),...
'obslabels',names);
axis([-.26 1 -.51 .51 -.61 .81]);
view([30 40]);
10-44
Feature Transformation
Factor Analysis
• “Introduction” on page 10-45
• “Example: Factor Analysis” on page 10-46
Introduction
Multivariate data often includes a large number of measured variables, and
sometimes those variables overlap, in the sense that groups of them might be
dependent. For example, in a decathlon, each athlete competes in 10 events,
but several of them can be thought of as speed events, while others can be
thought of as strength events, etc. Thus, you can think of a competitor’s 10
event scores as largely dependent on a smaller set of three or four types of
athletic ability.
10-45
10 Multivariate Methods
Factor analysis is a way to fit a model to multivariate data to estimate just this
sort of interdependence. In a factor analysis model, the measured variables
depend on a smaller number of unobserved (latent) factors. Because each
factor might affect several variables in common, they are known as common
factors. Each variable is assumed to be dependent on a linear combination
of the common factors, and the coefficients are known as loadings. Each
measured variable also includes a component due to independent random
variability, known as specific variance because it is specific to one variable.
Specifically, factor analysis assumes that the covariance matrix of your data
is of the form
∑ x = ΛΛΤ + Ψ
where Λ is the matrix of loadings, and the elements of the diagonal matrix
Ψ are the specific variances. The function factoran fits the Factor Analysis
model using maximum likelihood.
Example: Factor Analysis
• “Factor Loadings” on page 10-46
• “Factor Rotation” on page 10-48
• “Factor Scores” on page 10-50
• “Visualizing the Results” on page 10-52
Factor Loadings. Over the course of 100 weeks, the percent change in stock
prices for ten companies has been recorded. Of the ten companies, the first
four can be classified as primarily technology, the next three as financial, and
the last three as retail. It seems reasonable that the stock prices for companies
that are in the same sector might vary together as economic conditions
change. Factor Analysis can provide quantitative evidence that companies
within each sector do experience similar week-to-week changes in stock price.
In this example, you first load the data, and then call factoran, specifying a
model fit with three common factors. By default, factoran computes rotated
estimates of the loadings to try and make their interpretation simpler. But in
this example, you specify an unrotated solution.
10-46
Feature Transformation
load stockreturns
[Loadings,specificVar,T,stats] = ...
factoran(stocks,3,'rotate','none');
The first two factoran return arguments are the estimated loadings and the
estimated specific variances. Each row of the loadings matrix represents one
of the ten stocks, and each column corresponds to a common factor. With
unrotated estimates, interpretation of the factors in this fit is difficult because
most of the stocks contain fairly large coefficients for two or more factors.
Loadings
Loadings =
0.8885 0.2367 -0.2354
0.7126 0.3862 0.0034
0.3351 0.2784 -0.0211
0.3088 0.1113 -0.1905
0.6277 -0.6643 0.1478
0.4726 -0.6383 0.0133
0.1133 -0.5416 0.0322
0.6403 0.1669 0.4960
0.2363 0.5293 0.5770
0.1105 0.1680 0.5524
Note “Factor Rotation” on page 10-48 helps to simplify the structure in the
Loadings matrix, to make it easier to assign meaningful interpretations to
the factors.
From the estimated specific variances, you can see that the model indicates
that a particular stock price varies quite a lot beyond the variation due to
the common factors.
specificVar
specificVar =
0.0991
0.3431
0.8097
0.8559
0.1429
10-47
10 Multivariate Methods
0.3691
0.6928
0.3162
0.3311
0.6544
A specific variance of 1 would indicate that there is no common factor
component in that variable, while a specific variance of 0 would indicate that
the variable is entirely determined by common factors. These data seem to
fall somewhere in between.
The p value returned in the stats structure fails to reject the null hypothesis
of three common factors, suggesting that this model provides a satisfactory
explanation of the covariation in these data.
stats.p
ans =
0.8144
To determine whether fewer than three factors can provide an acceptable fit,
you can try a model with two common factors. The p value for this second fit
is highly significant, and rejects the hypothesis of two factors, indicating that
the simpler model is not sufficient to explain the pattern in these data.
[Loadings2,specificVar2,T2,stats2] = ...
factoran(stocks, 2,'rotate','none');
stats2.p
ans =
3.5610e-006
Factor Rotation. As the results illustrate, the estimated loadings from an
unrotated factor analysis fit can have a complicated structure. The goal of
factor rotation is to find a parameterization in which each variable has only a
small number of large loadings. That is, each variable is affected by a small
number of factors, preferably only one. This can often make it easier to
interpret what the factors represent.
If you think of each row of the loadings matrix as coordinates of a point
in M-dimensional space, then each factor corresponds to a coordinate axis.
Factor rotation is equivalent to rotating those axes and computing new
10-48
Feature Transformation
loadings in the rotated coordinate system. There are various ways to do this.
Some methods leave the axes orthogonal, while others are oblique methods
that change the angles between them. For this example, you can rotate the
estimated loadings by using the promax criterion, a common oblique method.
[LoadingsPM,specVarPM] = factoran(stocks,3,'rotate','promax');
LoadingsPM
LoadingsPM =
0.9452 0.1214 -0.0617
0.7064 -0.0178 0.2058
0.3885 -0.0994 0.0975
0.4162 -0.0148 -0.1298
0.1021 0.9019 0.0768
0.0873 0.7709 -0.0821
-0.1616 0.5320 -0.0888
0.2169 0.2844 0.6635
0.0016 -0.1881 0.7849
-0.2289 0.0636 0.6475
Promax rotation creates a simpler structure in the loadings, one in which
most of the stocks have a large loading on only one factor. To see this
structure more clearly, you can use the biplot function to plot each stock
using its factor loadings as coordinates.
biplot(LoadingsPM,'varlabels',num2str((1:10)'));
axis square
view(155,27);
10-49
10 Multivariate Methods
This plot shows that promax has rotated the factor loadings to a simpler
structure. Each stock depends primarily on only one factor, and it is possible
to describe each factor in terms of the stocks that it affects. Based on which
companies are near which axes, you could reasonably conclude that the first
factor axis represents the financial sector, the second retail, and the third
technology. The original conjecture, that stocks vary primarily within sector,
is apparently supported by the data.
Factor Scores. Sometimes, it is useful to be able to classify an observation
based on its factor scores. For example, if you accepted the three-factor model
and the interpretation of the rotated factors, you might want to categorize
each week in terms of how favorable it was for each of the three stock sectors,
based on the data from the 10 observed stocks.
Because the data in this example are the raw stock price changes, and not
just their correlation matrix, you can have factoran return estimates of the
10-50
Feature Transformation
value of each of the three rotated common factors for each week. You can
then plot the estimated scores to see how the different stock sectors were
affected during each week.
[LoadingsPM,specVarPM,TPM,stats,F] = ...
factoran(stocks, 3,'rotate','promax');
plot3(F(:,1),F(:,2),F(:,3),'b.')
line([-4 4 NaN 0 0 NaN 0 0], [0 0 NaN -4 4 NaN 0 0],...
[0 0 NaN 0 0 NaN -4 4], 'Color','black')
xlabel('Financial Sector')
ylabel('Retail Sector')
zlabel('Technology Sector')
grid on
axis square
view(-22.5, 8)
10-51
10 Multivariate Methods
Oblique rotation often creates factors that are correlated. This plot shows
some evidence of correlation between the first and third factors, and you can
investigate further by computing the estimated factor correlation matrix.
inv(TPM'*TPM)
ans =
1.0000 0.1559 0.4082
0.1559 1.0000 -0.0559
0.4082 -0.0559 1.0000
Visualizing the Results. You can use the biplot function to help visualize
both the factor loadings for each variable and the factor scores for each
observation in a single plot. For example, the following command plots the
results from the factor analysis on the stock data and labels each of the 10
stocks.
biplot(LoadingsPM,'scores',F,'varlabels',num2str((1:10)'))
xlabel('Financial Sector')
ylabel('Retail Sector')
zlabel('Technology Sector')
axis square
view(155,27)
10-52
Feature Transformation
In this case, the factor analysis includes three factors, and so the biplot is
three-dimensional. Each of the 10 stocks is represented in this plot by a vector,
and the direction and length of the vector indicates how each stock depends
on the underlying factors. For example, you have seen that after promax
rotation, the first four stocks have positive loadings on the first factor, and
unimportant loadings on the other two factors. That first factor, interpreted
as a financial sector effect, is represented in this biplot as one of the horizontal
axes. The dependence of those four stocks on that factor corresponds to the
four vectors directed approximately along that axis. Similarly, the dependence
of stocks 5, 6, and 7 primarily on the second factor, interpreted as a retail
sector effect, is represented by vectors directed approximately along that axis.
Each of the 100 observations is represented in this plot by a point, and their
locations indicate the score of each observation for the three factors. For
example, points near the top of this plot have the highest scores for the
10-53
10 Multivariate Methods
technology sector factor. The points are scaled to fit within the unit square, so
only their relative locations can be determined from the plot.
You can use the Data Cursor tool from the Tools menu in the figure window
to identify the items in this plot. By clicking a stock (vector), you can read off
that stock’s loadings for each factor. By clicking an observation (point), you
can read off that observation’s scores for each factor.
10-54
11
Cluster Analysis
• “Introduction” on page 11-2
• “Hierarchical Clustering” on page 11-3
• “K-Means Clustering” on page 11-21
• “Gaussian Mixture Models” on page 11-28
11 Cluster Analysis
Introduction
Cluster analysis, also called segmentation analysis or taxonomy analysis,
creates groups, or clusters, of data. Clusters are formed in such a way that
objects in the same cluster are very similar and objects in different clusters
are very distinct. Measures of similarity depend on the application.
“Hierarchical Clustering” on page 11-3 groups data over a variety of scales by
creating a cluster tree or dendrogram. The tree is not a single set of clusters,
but rather a multilevel hierarchy, where clusters at one level are joined
as clusters at the next level. This allows you to decide the level or scale
of clustering that is most appropriate for your application. The Statistics
Toolbox function clusterdata performs all of the necessary steps for you.
It incorporates the pdist, linkage, and cluster functions, which may be
used separately for more detailed analysis. The dendrogram function plots
the cluster tree.
“K-Means Clustering” on page 11-21 is a partitioning method. The function
kmeans partitions data into k mutually exclusive clusters, and returns
the index of the cluster to which it has assigned each observation. Unlike
hierarchical clustering, k-means clustering operates on actual observations
(rather than the larger set of dissimilarity measures), and creates a single
level of clusters. The distinctions mean that k-means clustering is often more
suitable than hierarchical clustering for large amounts of data.
“Gaussian Mixture Models” on page 11-28 form clusters by representing the
probability density function of observed variables as a mixture of multivariate
normal densities. Mixture models of the gmdistribution class use an
expectation maximization (EM) algorithm to fit data, which assigns posterior
probabilities to each component density with respect to each observation.
Clusters are assigned by selecting the component that maximizes the
posterior probability. Clustering using Gaussian mixture models is sometimes
considered a soft clustering method. The posterior probabilities for each
point indicate that each data point has some probability of belonging to
each cluster. Like k-means clustering, Gaussian mixture modeling uses an
iterative algorithm that converges to a local optimum. Gaussian mixture
modeling may be more appropriate than k-means clustering when clusters
have different sizes and correlation within them.
11-2
Hierarchical Clustering
Hierarchical Clustering
In this section...
“Introduction” on page 11-3
“Algorithm Description” on page 11-3
“Similarity Measures” on page 11-4
“Linkages” on page 11-6
“Dendrograms” on page 11-8
“Verifying the Cluster Tree” on page 11-10
“Creating Clusters” on page 11-16
Introduction
Hierarchical clustering groups data over a variety of scales by creating a
cluster tree or dendrogram. The tree is not a single set of clusters, but rather
a multilevel hierarchy, where clusters at one level are joined as clusters at
the next level. This allows you to decide the level or scale of clustering that
is most appropriate for your application. The Statistics Toolbox function
clusterdata supports agglomerative clustering and performs all of the
necessary steps for you. It incorporates the pdist, linkage, and cluster
functions, which you can use separately for more detailed analysis. The
dendrogram function plots the cluster tree.
Algorithm Description
To perform agglomerative hierarchical cluster analysis on a data set using
Statistics Toolbox functions, follow this procedure:
1 Find the similarity or dissimilarity between every pair of objects
in the data set. In this step, you calculate the distance between objects
using the pdist function. The pdist function supports many different
ways to compute this measurement. See “Similarity Measures” on page
11-4 for more information.
2 Group the objects into a binary, hierarchical cluster tree. In this
step, you link pairs of objects that are in close proximity using the linkage
11-3
11 Cluster Analysis
function. The linkage function uses the distance information generated in
step 1 to determine the proximity of objects to each other. As objects are
paired into binary clusters, the newly formed clusters are grouped into
larger clusters until a hierarchical tree is formed. See “Linkages” on page
11-6 for more information.
3 Determine where to cut the hierarchical tree into clusters. In this
step, you use the cluster function to prune branches off the bottom of
the hierarchical tree, and assign all the objects below each cut to a single
cluster. This creates a partition of the data. The cluster function can
create these clusters by detecting natural groupings in the hierarchical tree
or by cutting off the hierarchical tree at an arbitrary point.
The following sections provide more information about each of these steps.
Note The Statistics Toolbox function clusterdata performs all of the
necessary steps for you. You do not need to execute the pdist, linkage, or
cluster functions separately.
Similarity Measures
You use the pdist function to calculate the distance between every pair of
objects in a data set. For a data set made up of m objects, there are m*(m –
1)/2 pairs in the data set. The result of this computation is commonly known
as a distance or dissimilarity matrix.
There are many ways to calculate this distance information. By default, the
pdist function calculates the Euclidean distance between objects; however,
you can specify one of several other options. See pdist for more information.
Note You can optionally normalize the values in the data set before
calculating the distance information. In a real world data set, variables can
be measured against different scales. For example, one variable can measure
Intelligence Quotient (IQ) test scores and another variable can measure head
circumference. These discrepancies can distort the proximity calculations.
Using the zscore function, you can convert all the values in the data set to
use the same proportional scale. See zscore for more information.
11-4
Hierarchical Clustering
For example, consider a data set, X, made up of five objects where each object
is a set of x,y coordinates.
• Object 1: 1, 2
• Object 2: 2.5, 4.5
• Object 3: 2, 2
• Object 4: 4, 1.5
• Object 5: 4, 2.5
You can define this data set as a matrix
X = [1 2;2.5 4.5;2 2;4 1.5;4 2.5]
and pass it to pdist. The pdist function calculates the distance between
object 1 and object 2, object 1 and object 3, and so on until the distances
between all the pairs have been calculated. The following figure plots these
objects in a graph. The Euclidean distance between object 2 and object 3 is
shown to illustrate one interpretation of distance.
Distance Information
The pdist function returns this distance information in a vector, Y, where
each element contains the distance between a pair of objects.
11-5
11 Cluster Analysis
Y = pdist(X)
Y =
Columns 1 through 5
2.9155 1.0000 3.0414 3.0414 2.5495
Columns 6 through 10
3.3541 2.5000 2.0616 2.0616 1.0000
To make it easier to see the relationship between the distance information
generated by pdist and the objects in the original data set, you can reformat
the distance vector into a matrix using the squareform function. In this
matrix, element i,j corresponds to the distance between object i and object j in
the original data set. In the following example, element 1,1 represents the
distance between object 1 and itself (which is zero). Element 1,2 represents
the distance between object 1 and object 2, and so on.
squareform(Y)
ans =
0 2.9155 1.0000 3.0414 3.0414
2.9155 0 2.5495 3.3541 2.5000
1.0000 2.5495 0 2.0616 2.0616
3.0414 3.3541 2.0616 0 1.0000
3.0414 2.5000 2.0616 1.0000 0
Linkages
Once the proximity between objects in the data set has been computed, you
can determine how objects in the data set should be grouped into clusters,
using the linkage function. The linkage function takes the distance
information generated by pdist and links pairs of objects that are close
together into binary clusters (clusters made up of two objects). The linkage
function then links these newly formed clusters to each other and to other
objects to create bigger clusters until all the objects in the original data set
are linked together in a hierarchical tree.
For example, given the distance vector Y generated by pdist from the sample
data set of x- and y-coordinates, the linkage function generates a hierarchical
cluster tree, returning the linkage information in a matrix, Z.
Z = linkage(Y)
Z =
4.0000 5.0000 1.0000
11-6
Hierarchical Clustering
1.0000 3.0000 1.0000
6.0000 7.0000 2.0616
2.0000 8.0000 2.5000
In this output, each row identifies a link between objects or clusters. The first
two columns identify the objects that have been linked. The third column
contains the distance between these objects. For the sample data set of x-
and y-coordinates, the linkage function begins by grouping objects 4 and 5,
which have the closest proximity (distance value = 1.0000). The linkage
function continues by grouping objects 1 and 3, which also have a distance
value of 1.0000.
The third row indicates that the linkage function grouped objects 6 and 7. If
the original sample data set contained only five objects, what are objects 6
and 7? Object 6 is the newly formed binary cluster created by the grouping
of objects 4 and 5. When the linkage function groups two objects into a
new cluster, it must assign the cluster a unique index value, starting with
the value m+1, where m is the number of objects in the original data set.
(Values 1 through m are already used by the original data set.) Similarly,
object 7 is the cluster formed by grouping objects 1 and 3.
linkage uses distances to determine the order in which it clusters objects.
The distance vector Y contains the distances between the original objects 1
through 5. But linkage must also be able to determine distances involving
clusters that it creates, such as objects 6 and 7. By default, linkage uses a
method known as single linkage. However, there are a number of different
methods available. See the linkage reference page for more information.
As the final cluster, the linkage function grouped object 8, the newly formed
cluster made up of objects 6 and 7, with object 2 from the original data set.
The following figure graphically illustrates the way linkage groups the
objects into a hierarchy of clusters.
11-7
11 Cluster Analysis
Dendrograms
The hierarchical, binary cluster tree created by the linkage function is most
easily understood when viewed graphically. The Statistics Toolbox function
dendrogram plots the tree, as follows:
dendrogram(Z)
11-8
Hierarchical Clustering
2.5
1.5
4 5 1 3 2
In the figure, the numbers along the horizontal axis represent the indices of
the objects in the original data set. The links between objects are represented
as upside-down U-shaped lines. The height of the U indicates the distance
between the objects. For example, the link representing the cluster containing
objects 1 and 3 has a height of 1. The link representing the cluster that groups
object 2 together with objects 1, 3, 4, and 5, (which are already clustered as
object 8) has a height of 2.5. The height represents the distance linkage
computes between objects 2 and 8. For more information about creating a
dendrogram diagram, see the dendrogram reference page.
11-9
11 Cluster Analysis
Verifying the Cluster Tree
After linking the objects in a data set into a hierarchical cluster tree, you
might want to verify that the distances (that is, heights) in the tree reflect
the original distances accurately. In addition, you might want to investigate
natural divisions that exist among links between objects. Statistics Toolbox
functions are available for both of these tasks, as described in the following
sections:
• “Verifying Dissimilarity” on page 11-10
• “Verifying Consistency” on page 11-11
Verifying Dissimilarity
In a hierarchical cluster tree, any two objects in the original data set are
eventually linked together at some level. The height of the link represents
the distance between the two clusters that contain those two objects. This
height is known as the cophenetic distance between the two objects. One
way to measure how well the cluster tree generated by the linkage function
reflects your data is to compare the cophenetic distances with the original
distance data generated by the pdist function. If the clustering is valid, the
linking of objects in the cluster tree should have a strong correlation with
the distances between objects in the distance vector. The cophenet function
compares these two sets of values and computes their correlation, returning a
value called the cophenetic correlation coefficient. The closer the value of the
cophenetic correlation coefficient is to 1, the more accurately the clustering
solution reflects your data.
You can use the cophenetic correlation coefficient to compare the results of
clustering the same data set using different distance calculation methods or
clustering algorithms. For example, you can use the cophenet function to
evaluate the clusters created for the sample data set
c = cophenet(Z,Y)
c =
0.8615
where Z is the matrix output by the linkage function and Y is the distance
vector output by the pdist function.
11-10
Hierarchical Clustering
Execute pdist again on the same data set, this time specifying the city block
metric. After running the linkage function on this new pdist output using
the average linkage method, call cophenet to evaluate the clustering solution.
Y = pdist(X,'cityblock');
Z = linkage(Y,'average');
c = cophenet(Z,Y)
c =
0.9047
The cophenetic correlation coefficient shows that using a different distance
and linkage method creates a tree that represents the original distances
slightly better.
Verifying Consistency
One way to determine the natural cluster divisions in a data set is to compare
the height of each link in a cluster tree with the heights of neighboring links
below it in the tree.
A link that is approximately the same height as the links below it indicates
that there are no distinct divisions between the objects joined at this level of
the hierarchy. These links are said to exhibit a high level of consistency,
because the distance between the objects being joined is approximately the
same as the distances between the objects they contain.
On the other hand, a link whose height differs noticeably from the height of
the links below it indicates that the objects joined at this level in the cluster
tree are much farther apart from each other than their components were when
they were joined. This link is said to be inconsistent with the links below it.
In cluster analysis, inconsistent links can indicate the border of a natural
division in a data set. The cluster function uses a quantitative measure of
inconsistency to determine where to partition your data set into clusters.
The following dendrogram illustrates inconsistent links. Note how the objects
in the dendrogram fall into two groups that are connected by links at a much
higher level in the tree. These links are inconsistent when compared with the
links below them in the hierarchy.
11-11
11 Cluster Analysis
These links show inconsistency when compared
to the links below them.
These links show consistency.
The relative consistency of each link in a hierarchical cluster tree can be
quantified and expressed as the inconsistency coefficient. This value compares
the height of a link in a cluster hierarchy with the average height of links
below it. Links that join distinct clusters have a high inconsistency coefficient;
links that join indistinct clusters have a low inconsistency coefficient.
To generate a listing of the inconsistency coefficient for each link in the
cluster tree, use the inconsistent function. By default, the inconsistent
11-12
Hierarchical Clustering
function compares each link in the cluster hierarchy with adjacent links that
are less than two levels below it in the cluster hierarchy. This is called the
depth of the comparison. You can also specify other depths. The objects at
the bottom of the cluster tree, called leaf nodes, that have no further objects
below them, have an inconsistency coefficient of zero. Clusters that join two
leaves also have a zero inconsistency coefficient.
For example, you can use the inconsistent function to calculate the
inconsistency values for the links created by the linkage function in
“Linkages” on page 11-6.
I = inconsistent(Z)
I =
1.0000 0 1.0000 0
1.0000 0 1.0000 0
1.3539 0.6129 3.0000 1.1547
2.2808 0.3100 2.0000 0.7071
The inconsistent function returns data about the links in an (m-1)-by-4
matrix, whose columns are described in the following table.
Column Description
1 Mean of the heights of all the links included in the calculation
2 Standard deviation of all the links included in the calculation
3 Number of links included in the calculation
4 Inconsistency coefficient
In the sample output, the first row represents the link between objects 4
and 5. This cluster is assigned the index 6 by the linkage function. Because
both 4 and 5 are leaf nodes, the inconsistency coefficient for the cluster is zero.
The second row represents the link between objects 1 and 3, both of which are
also leaf nodes. This cluster is assigned the index 7 by the linkage function.
The third row evaluates the link that connects these two clusters, objects 6
and 7. (This new cluster is assigned index 8 in the linkage output). Column 3
indicates that three links are considered in the calculation: the link itself and
the two links directly below it in the hierarchy. Column 1 represents the mean
of the heights of these links. The inconsistent function uses the height
11-13
11 Cluster Analysis
information output by the linkage function to calculate the mean. Column 2
represents the standard deviation between the links. The last column contains
the inconsistency value for these links, 1.1547. It is the difference between
the current link height and the mean, normalized by the standard deviation:
(2.0616 - 1.3539) / .6129
ans =
1.1547
The following figure illustrates the links and heights included in this
calculation.
Links
Heights
11-14
Hierarchical Clustering
Note In the preceding figure, the lower limit on the y-axis is set to 0 to show
the heights of the links. To set the lower limit to 0, select Axes Properties
from the Edit menu, click the Y Axis tab, and enter 0 in the field immediately
to the right of Y Limits.
Row 4 in the output matrix describes the link between object 8 and object 2.
Column 3 indicates that two links are included in this calculation: the link
itself and the link directly below it in the hierarchy. The inconsistency
coefficient for this link is 0.7071.
The following figure illustrates the links and heights included in this
calculation.
11-15
11 Cluster Analysis
Links
Heights
Creating Clusters
After you create the hierarchical tree of binary clusters, you can prune the
tree to partition your data into clusters using the cluster function. The
cluster function lets you create clusters in two ways, as discussed in the
following sections:
• “Finding Natural Divisions in Data” on page 11-17
• “Specifying Arbitrary Clusters” on page 11-18
11-16
Hierarchical Clustering
Finding Natural Divisions in Data
The hierarchical cluster tree may naturally divide the data into distinct,
well-separated clusters. This can be particularly evident in a dendrogram
diagram created from data where groups of objects are densely packed in
certain areas and not in others. The inconsistency coefficient of the links in
the cluster tree can identify these divisions where the similarities between
objects change abruptly. (See “Verifying the Cluster Tree” on page 11-10 for
more information about the inconsistency coefficient.) You can use this value
to determine where the cluster function creates cluster boundaries.
For example, if you use the cluster function to group the sample data set
into clusters, specifying an inconsistency coefficient threshold of 1.2 as the
value of the cutoff argument, the cluster function groups all the objects
in the sample data set into one cluster. In this case, none of the links in the
cluster hierarchy had an inconsistency coefficient greater than 1.2.
T = cluster(Z,'cutoff',1.2)
T =
1
1
1
1
1
The cluster function outputs a vector, T, that is the same size as the original
data set. Each element in this vector contains the number of the cluster into
which the corresponding object from the original data set was placed.
If you lower the inconsistency coefficient threshold to 0.8, the cluster
function divides the sample data set into three separate clusters.
T = cluster(Z,'cutoff',0.8)
T =
3
2
3
1
1
11-17
11 Cluster Analysis
This output indicates that objects 1 and 3 were placed in cluster 1, objects 4
and 5 were placed in cluster 2, and object 2 was placed in cluster 3.
When clusters are formed in this way, the cutoff value is applied to the
inconsistency coefficient. These clusters may, but do not necessarily,
correspond to a horizontal slice across the dendrogram at a certain height.
If you want clusters corresponding to a horizontal slice of the dendrogram,
you can either use the criterion option to specify that the cutoff should be
based on distance rather than inconsistency, or you can specify the number of
clusters directly as described in the following section.
Specifying Arbitrary Clusters
Instead of letting the cluster function create clusters determined by the
natural divisions in the data set, you can specify the number of clusters you
want created.
For example, you can specify that you want the cluster function to partition
the sample data set into two clusters. In this case, the cluster function
creates one cluster containing objects 1, 3, 4, and 5 and another cluster
containing object 2.
T = cluster(Z,'maxclust',2)
T =
2
1
2
2
2
To help you visualize how the cluster function determines these clusters, the
following figure shows the dendrogram of the hierarchical cluster tree. The
horizontal dashed line intersects two lines of the dendrogram, corresponding
to setting 'maxclust' to 2. These two lines partition the objects into two
clusters: the objects below the left-hand line, namely 1, 3, 4, and 5, belong to
one cluster, while the object below the right-hand line, namely 2, belongs to
the other cluster.
11-18
Hierarchical Clustering
maxclust= 2
On the other hand, if you set 'maxclust' to 3, the cluster function groups
objects 4 and 5 in one cluster, objects 1 and 3 in a second cluster, and object 2
in a third cluster. The following command illustrates this.
T = cluster(Z,'maxclust',3)
T =
1
3
1
2
2
11-19
11 Cluster Analysis
This time, the cluster function cuts off the hierarchy at a lower point,
corresponding to the horizontal line that intersects three lines of the
dendrogram in the following figure.
maxclust= 3
11-20
K-Means Clustering
K-Means Clustering
In this section...
“Introduction” on page 11-21
“Creating Clusters and Determining Separation” on page 11-22
“Determining the Correct Number of Clusters” on page 11-23
“Avoiding Local Minima” on page 11-26
Introduction
K-means clustering is a partitioning method. The function kmeans partitions
data into k mutually exclusive clusters, and returns the index of the cluster
to which it has assigned each observation. Unlike hierarchical clustering,
k-means clustering operates on actual observations (rather than the larger
set of dissimilarity measures), and creates a single level of clusters. The
distinctions mean that k-means clustering is often more suitable than
hierarchical clustering for large amounts of data.
kmeans treats each observation in your data as an object having a location in
space. It finds a partition in which objects within each cluster are as close to
each other as possible, and as far from objects in other clusters as possible.
You can choose from five different distance measures, depending on the kind
of data you are clustering.
Each cluster in the partition is defined by its member objects and by its
centroid, or center. The centroid for each cluster is the point to which the sum
of distances from all objects in that cluster is minimized. kmeans computes
cluster centroids differently for each distance measure, to minimize the sum
with respect to the measure that you specify.
kmeans uses an iterative algorithm that minimizes the sum of distances from
each object to its cluster centroid, over all clusters. This algorithm moves
objects between clusters until the sum cannot be decreased further. The
result is a set of clusters that are as compact and well-separated as possible.
You can control the details of the minimization using several optional input
parameters to kmeans, including ones for the initial values of the cluster
centroids, and for the maximum number of iterations.
11-21
11 Cluster Analysis
Creating Clusters and Determining Separation
The following example explores possible clustering in four-dimensional data
by analyzing the results of partitioning the points into three, four, and five
clusters.
Note Because each part of this example generates random numbers
sequentially, i.e., without setting a new state, you must perform all steps
in sequence to duplicate the results shown. If you perform the steps out of
sequence, the answers will be essentially the same, but the intermediate
results, number of iterations, or ordering of the silhouette plots may differ.
First, load some data:
load kmeansdata;
size(X)
ans =
560 4
Even though these data are four-dimensional, and cannot be easily visualized,
kmeans enables you to investigate whether a group structure exists in them.
Call kmeans with k, the desired number of clusters, equal to 3. For this
example, specify the city block distance measure, and use the default starting
method of initializing centroids from randomly selected data points:
idx3 = kmeans(X,3,'distance','city');
To get an idea of how well-separated the resulting clusters are, you can make
a silhouette plot using the cluster indices output from kmeans. The silhouette
plot displays a measure of how close each point in one cluster is to points in
the neighboring clusters. This measure ranges from +1, indicating points that
are very distant from neighboring clusters, through 0, indicating points that
are not distinctly in one cluster or another, to -1, indicating points that are
probably assigned to the wrong cluster. silhouette returns these values in
its first output:
[silh3,h] = silhouette(X,idx3,'city');
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
xlabel('Silhouette Value')
ylabel('Cluster')
11-22
K-Means Clustering
From the silhouette plot, you can see that most points in the third cluster
have a large silhouette value, greater than 0.6, indicating that the cluster is
somewhat separated from neighboring clusters. However, the first cluster
contains many points with low silhouette values, and the second contains a
few points with negative values, indicating that those two clusters are not
well separated.
Determining the Correct Number of Clusters
Increase the number of clusters to see if kmeans can find a better grouping
of the data. This time, use the optional 'display' parameter to print
information about each iteration:
idx4 = kmeans(X,4, 'dist','city', 'display','iter');
iter phase num sum
1 1 560 2897.56
11-23
11 Cluster Analysis
2 1 53 2736.67
3 1 50 2476.78
4 1 102 1779.68
5 1 5 1771.1
6 2 0 1771.1
6 iterations, total sum of distances = 1771.1
Notice that the total sum of distances decreases at each iteration as kmeans
reassigns points between clusters and recomputes cluster centroids. In this
case, the second phase of the algorithm did not make any reassignments,
indicating that the first phase reached a minimum after five iterations. In
some problems, the first phase might not reach a minimum, but the second
phase always will.
A silhouette plot for this solution indicates that these four clusters are better
separated than the three in the previous solution:
[silh4,h] = silhouette(X,idx4,'city');
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
xlabel('Silhouette Value')
ylabel('Cluster')
11-24
K-Means Clustering
A more quantitative way to compare the two solutions is to look at the average
silhouette values for the two cases:
mean(silh3)
ans =
0.52594
mean(silh4)
ans =
0.63997
Finally, try clustering the data using five clusters:
idx5 = kmeans(X,5,'dist','city','replicates',5);
[silh5,h] = silhouette(X,idx5,'city');
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
xlabel('Silhouette Value')
11-25
11 Cluster Analysis
ylabel('Cluster')
mean(silh5)
ans =
0.52657
This silhouette plot indicates that this is probably not the right number of
clusters, since two of the clusters contain points with mostly low silhouette
values. Without some knowledge of how many clusters are really in the data,
it is a good idea to experiment with a range of values for k.
Avoiding Local Minima
Like many other types of numerical minimizations, the solution that kmeans
reaches often depends on the starting points. It is possible for kmeans to
reach a local minimum, where reassigning any one point to a new cluster
would increase the total sum of point-to-centroid distances, but where a
11-26
K-Means Clustering
better solution does exist. However, you can use the optional 'replicates'
parameter to overcome that problem.
For four clusters, specify five replicates, and use the 'display' parameter to
print out the final sum of distances for each of the solutions.
[idx4,cent4,sumdist] = kmeans(X,4,'dist','city',...
'display','final','replicates',5);
17 iterations, total sum of distances = 2303.36
5 iterations, total sum of distances = 1771.1
6 iterations, total sum of distances = 1771.1
5 iterations, total sum of distances = 1771.1
8 iterations, total sum of distances = 2303.36
The output shows that, even for this relatively simple problem, non-global
minima do exist. Each of these five replicates began from a different randomly
selected set of initial centroids, and kmeans found two different local minima.
However, the final solution that kmeans returns is the one with the lowest
total sum of distances, over all replicates.
sum(sumdist)
ans =
1771.1
11-27
11 Cluster Analysis
Gaussian Mixture Models
In this section...
“Introduction” on page 11-28
“Clustering with Gaussian Mixtures” on page 11-28
Introduction
Gaussian mixture models are formed by combining multivariate normal
density components. For information on individual multivariate normal
densities, see “Multivariate Normal Distribution” on page B-58 and related
distribution functions listed under “Multivariate Distributions” on page 5-8.
In Statistics Toolbox software, use the gmdistribution class to fit data
using an expectation maximization (EM) algorithm, which assigns posterior
probabilities to each component density with respect to each observation.
Gaussian mixture models are often used for data clustering. Clusters are
assigned by selecting the component that maximizes the posterior probability.
Like k-means clustering, Gaussian mixture modeling uses an iterative
algorithm that converges to a local optimum. Gaussian mixture modeling may
be more appropriate than k-means clustering when clusters have different
sizes and correlation within them. Clustering using Gaussian mixture models
is sometimes considered a soft clustering method. The posterior probabilities
for each point indicate that each data point has some probability of belonging
to each cluster.
Creation of Gaussian mixture models is described in the “Gaussian Mixture
Models” on page 5-99 section of Chapter 5, “Probability Distributions”. This
section describes their application in cluster analysis.
Clustering with Gaussian Mixtures
Gaussian mixture distributions can be used for clustering data, by realizing
that the multivariate normal components of the fitted model can represent
clusters.
11-28
Gaussian Mixture Models
1 To demonstrate the process, first generate some simulated data from a
mixture of two bivariate Gaussian distributions using the mvnrnd function:
mu1 = [1 2];
sigma1 = [3 .2; .2 2];
mu2 = [-1 -2];
sigma2 = [2 0; 0 1];
X = [mvnrnd(mu1,sigma1,200);mvnrnd(mu2,sigma2,100)];
scatter(X(:,1),X(:,2),10,'ko')
2 Fit a two-component Gaussian mixture distribution. Here, you know
the correct number of components to use. In practice, with real data,
this decision would require comparing models with different numbers of
components.
11-29
11 Cluster Analysis
options = statset('Display','final');
gm = gmdistribution.fit(X,2,'Options',options);
This displays
49 iterations, log-likelihood = -1207.91
3 Plot the estimated probability density contours for the two-component
mixture distribution. The two bivariate normal components overlap, but
their peaks are distinct. This suggests that the data could reasonably be
divided into two clusters:
hold on
ezcontour(@(x,y)pdf(gm,[x y]),[-8 6],[-8 6]);
hold off
11-30
Gaussian Mixture Models
4 Partition the data into clusters using the cluster method for the fitted
mixture distribution. The cluster method assigns each point to one of the
two components in the mixture distribution.
idx = cluster(gm,X);
cluster1 = (idx == 1);
cluster2 = (idx == 2);
scatter(X(cluster1,1),X(cluster1,2),10,'r+');
hold on
scatter(X(cluster2,1),X(cluster2,2),10,'bo');
hold off
legend('Cluster 1','Cluster 2','Location','NW')
11-31
11 Cluster Analysis
Each cluster corresponds to one of the bivariate normal components in
the mixture distribution. cluster assigns points to clusters based on the
estimated posterior probability that a point came from a component; each
point is assigned to the cluster corresponding to the highest posterior
probability. The posterior method returns those posterior probabilities.
For example, plot the posterior probability of the first component for each
point:
P = posterior(gm,X);
scatter(X(cluster1,1),X(cluster1,2),10,P(cluster1,1),'+')
hold on
scatter(X(cluster2,1),X(cluster2,2),10,P(cluster2,1),'o')
hold off
legend('Cluster 1','Cluster 2','Location','NW')
clrmap = jet(80); colormap(clrmap(9:72,:))
11-32
Gaussian Mixture Models
ylabel(colorbar,'Component 1 Posterior Probability')
Soft Clustering Using Gaussian Mixture Distributions
An alternative to the previous example is to use the posterior probabilities for
"soft clustering". Each point is assigned a membership score to each cluster.
Membership scores are simply the posterior probabilities, and describe
how similar each point is to each cluster’s archetype, i.e., the mean of the
corresponding component. The points can be ranked by their membership
score in a given cluster:
[~,order] = sort(P(:,1));
plot(1:size(X,1),P(order,1),'r-',1:size(X,1),P(order,2),'b-');
legend({'Cluster 1 Score' 'Cluster 2 Score'},'location','NW');
ylabel('Cluster Membership Score');
xlabel('Point Ranking');
11-33
11 Cluster Analysis
Although a clear separation of the data is hard to see in a scatter plot of the
data, plotting the membership scores indicates that the fitted distribution
does a good job of separating the data into groups. Very few points have
scores close to 0.5.
Soft clustering using a Gaussian mixture distribution is similar to fuzzy
K-means clustering, which also assigns each point to each cluster with a
membership score. The fuzzy K-means algorithm assumes that clusters are
roughly spherical in shape, and all of roughly equal size. This is comparable
to a Gaussian mixture distribution with a single covariance matrix that is
shared across all components, and is a multiple of the identity matrix. In
contrast, gmdistribution allows you to specify different covariance options.
The default is to estimate a separate, unconstrained covariance matrix for
11-34
Gaussian Mixture Models
each component. A more restricted option, closer to K-means, would be to
estimate a shared, diagonal covariance matrix:
gm2 = gmdistribution.fit(X,2,'CovType','Diagonal',...
'SharedCov',true);
This covariance option is similar to fuzzy K-means clustering, but provides
more flexibility by allowing unequal variances for different variables.
You can compute the soft cluster membership scores without computing hard
cluster assignments, using posterior, or as part of hard clustering, as the
second output from cluster:
P2 = posterior(gm2,X); % equivalently [idx,P2] = cluster(gm2,X)
[~,order] = sort(P2(:,1));
plot(1:size(X,1),P2(order,1),'r-',1:size(X,1),P2(order,2),'b-');
legend({'Cluster 1 Score' 'Cluster 2 Score'},'location','NW');
ylabel('Cluster Membership Score');
xlabel('Point Ranking');
11-35
11 Cluster Analysis
Assigning New Data to Clusters
In the previous example, fitting the mixture distribution to data using fit,
and clustering those data using cluster, are separate steps. However, the
same data are used in both steps. You can also use the cluster method to
assign new data points to the clusters (mixture components) found in the
original data.
1 Given a data set X, first fit a Gaussian mixture distribution. The previous
code has already done that.
gm
gm =
Gaussian mixture distribution with 2 components in 2 dimensions
11-36
Gaussian Mixture Models
Component 1:
Mixing proportion: 0.312592
Mean: -0.9082 -2.1109
Component 2:
Mixing proportion: 0.687408
Mean: 0.9532 1.8940
2 You can then use cluster to assign each point in a new data set, Y, to one
of the clusters defined for the original data:
Y = [mvnrnd(mu1,sigma1,50);mvnrnd(mu2,sigma2,25)];
idx = cluster(gm,Y);
cluster1 = (idx == 1);
cluster2 = (idx == 2);
scatter(Y(cluster1,1),Y(cluster1,2),10,'r+');
hold on
scatter(Y(cluster2,1),Y(cluster2,2),10,'bo');
hold off
legend('Class 1','Class 2','Location','NW')
11-37
11 Cluster Analysis
As with the previous example, the posterior probabilities for each point can
be treated as membership scores rather than determining "hard" cluster
assignments.
For cluster to provide meaningful results with new data, Y should come
from the same population as X, the original data used to create the mixture
distribution. In particular, the estimated mixing probabilities for the
Gaussian mixture distribution fitted to X are used when computing the
posterior probabilities for Y.
11-38
12
LDA不同類別的差異最大化
Classification
• “Introduction” on page 12-2
• “Discriminant Analysis” on page 12-3
• “Naive Bayes Classification” on page 12-6
• “Classification Trees” on page 12-9
• “Classification Using Nearest Neighbors” on page 12-14
• “Regression and Classification by Bagging Decision Trees” on page 12-30
• “Performance Curves” on page 12-53
12 Classification
Introduction
Models of data with a categorical response are called classifiers. A classifier is
built from training data, for which classifications are known. The classifier
assigns new test data to one of the categorical levels of the response.
Parametric methods, like “Discriminant Analysis” on page 12-3, fit a
parametric model to the training data and interpolate to classify test data.
Nonparametric methods, like “Classification Trees” on page 12-9, use other
means to determine classifications. In this sense, classification methods are
analogous to the methods discussed in “Nonlinear Regression” on page 9-58.
12-2
Discriminant Analysis
Discriminant Analysis
In this section...
“Introduction” on page 12-3
“Example: Discriminant Analysis” on page 12-3
Introduction
Discriminant analysis uses training data to estimate the parameters of
discriminant functions of the predictor variables. Discriminant functions
determine boundaries in predictor space between various classes. The
resulting classifier discriminates among the classes (the categorical levels of
the response) based on the predictor data.
The Statistics Toolbox function classify performs discriminant analysis.
Example: Discriminant Analysis
1 For training data, use Fisher’s sepal measurements for iris versicolor and
virginica:
load fisheriris
SL = meas(51:end,1);
SW = meas(51:end,2);
group = species(51:end);
h1 = gscatter(SL,SW,group,'rb','v^',[],'off');
set(h1,'LineWidth',2)
legend('Fisher versicolor','Fisher virginica',...
'Location','NW')
12-3
12 Classification
2 Classify a grid of measurements on the same scale, using classify:
[X,Y] = meshgrid(linspace(4.5,8),linspace(2,4));
X = X(:); Y = Y(:);
[C,err,P,logp,coeff] = classify([X Y],[SL SW],...
group,'quadratic');
3 Visualize the classification:
hold on;
gscatter(X,Y,C,'rb','.',1,'off');
K = coeff(1,2).const;
L = coeff(1,2).linear;
Q = coeff(1,2).quadratic;
% Plot the curve K + [x,y]*L + [x,y]*Q*[x,y]' = 0:
f = @(x,y) K + L(1)*x + L(2)*y + Q(1,1)*x.^2 + ...
12-4
Discriminant Analysis
(Q(1,2)+Q(2,1))*x.*y + Q(2,2)*y.^2
h2 = ezplot(f,[4.5 8 2 4]);
set(h2,'Color','m','LineWidth',2)
axis([4.5 8 2 4])
xlabel('Sepal Length')
ylabel('Sepal Width')
title('{\bf Classification with Fisher Training Data}')
12-5
12 Classification
Naive Bayes Classification
The Naive Bayes classifier is designed for use when features are independent
of one another within each class, but it appears to work well in practice
even when that independence assumption is not valid. It classifies data in
two steps:
1 Training step: Using the training samples, the method estimates
the parameters of a probability distribution, assuming features are
conditionally independent given the class.
2 Prediction step: For any unseen test sample, the method computes the
posterior probability of that sample belonging to each class. The method
then classifies the test sample according the largest posterior probability.
The class-conditional independence assumption greatly simplifies the training
step since you can estimate the one-dimensional class-conditional density
for each feature individually. While the class-conditional independence
between features is not true in general, research shows that this optimistic
assumption works well in practice. This assumption of class independence
allows the Naive Bayes classifier to better estimate the parameters required
for accurate classification while using less training data than many other
classifiers. This makes it particularly effective for datasets containing many
predictors or features.
Supported Distributions
Naive Bayes classification is based on estimating P(X|Y), the probability or
probability density of features X given class Y. The Naive Bayes classification
object NaiveBayes provides support for normal (Gaussian), kernel,
multinomial, and multivariate multinomial distributions. It is possible to use
different distributions for different features.
Normal (Gaussian) Distribution
The 'normal' distribution is appropriate for features that have normal
distributions in each class. For each feature you model with a normal
distribution, the Naive Bayes classifier estimates a separate normal
12-6
Naive Bayes Classification
distribution for each class by computing the mean and standard deviation of
the training data in that class. For more information on normal distributions,
see “Normal Distribution” on page B-83.
Kernel Distribution
The 'kernel' distribution is appropriate for features that have a continuous
distribution. It does not require a strong assumption such as a normal
distribution and you can use it in cases where the distribution of a feature may
be skewed or have multiple peaks or modes. It requires more computing time
and more memory than the normal distribution. For each feature you model
with a kernel distribution, the Naive Bayes classifier computes a separate
kernel density estimate for each class based on the training data for that class.
By default the kernel is the normal kernel, and the classifier selects a width
automatically for each class and feature. It is possible to specify different
kernels for each feature, and different widths for each feature or class.
Multinomial Distribution
The multinomial distribution (specify with the 'mn' keyword) is appropriate
when all features represent counts of a set of words or tokens. This is
sometimes called the "bag of words" model. For example, an e-mail spam
classifier might be based on features that count the number of occurrences
of various tokens in an e-mail. One feature might count the number of
exclamation points, another might count the number of times the word
"money" appears, and another might count the number of times the recipient’s
name appears. This is a Naive Bayes model under the further assumption
that the total number of tokens (or the total document length) is independent
of response class.
For the multinomial option, each feature represents the count of one token.
The classifier counts the set of relative token probabilities separately for
each class. The classifier defines the multinomial distribution for each row
by the vector of probabilities for the corresponding class, and by N, the total
token count for that row.
Classification is based on the relative frequencies of the tokens. For a row in
which no token appears, N is 0 and no classification is possible. This classifier
is not appropriate when the total number of tokens provides information
about the response class.
12-7
12 Classification
Multivariate Multinomial Distribution
The multivariate multinomial distribution (specify with the 'mvmn' keyword)
is appropriate for categorical features. For example, you could fit a feature
describing the weather in categories such as rain/sun/snow/clouds using the
multivariate multinomial model. The feature categories are sometimes called
the feature levels, and differ from the class levels for the response variable.
For each feature you model with a multivariate multinomial distribution, the
Naive Bayes classifier computes a separate set of probabilities for the set of
feature levels for each class.
12-8
Classification Trees
Classification Trees
In this section...
“Introduction” on page 12-9
“Example: Classification Trees” on page 12-9
“References” on page 12-13
Introduction
Parametric models specify the form of the relationship between predictors
and a response, as in the Hougen-Watson model described in “Parametric
Models” on page 9-59. In many cases, however, the form of the relationship is
unknown, and a parametric model requires assumptions and simplifications.
Regression Trees offer a nonparametric alternative. When response data are
categorical, classification trees are a natural modification.
Note This section demonstrates methods for objects of the classregtree
class. These methods supersede the functions treefit, treedisp, treeval,
treeprune, and treetest, which are maintained in Statistics Toolbox
software only for backwards compatibility.
Example: Classification Trees
This example uses Fisher’s iris data in fisheriris.mat to create a
classification tree for predicting species using measurements of sepal length,
sepal width, petal length, and petal width as predictors. Note that, in this
case, the predictors are continuous and the response is categorical.
1 Load the data and use the classregtree constructor of the classregtree
class to create the classification tree:
load fisheriris
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
12-9
12 Classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
t is a classregtree object and can be operated on with any of the methods
of the class.
2 Use the type method of the classregtree class to show the type of the tree:
treetype = type(t)
treetype =
classification
classregtree creates a classification tree because species is a cell array
of strings, and the response is assumed to be categorical.
3 To view the tree, use the view method of the classregtree class:
view(t)
12-10
Classification Trees
The tree predicts the response values at the circular leaf nodes based on a
series of questions about the iris at the triangular branching nodes. A true
answer to any question follows the branch to the left; a false follows the
branch to the right.
4 The tree does not use sepal measurements for predicting species. These
can go unmeasured in new data, and you can enter them as NaN values for
predictions. For example, to use the tree to predict the species of an iris
with petal length 4.8 and petal width 1.6, type:
predicted = t([NaN NaN 4.8 1.6])
predicted =
'versicolor'
12-11
12 Classification
Note that the object allows for functional evaluation, of the form t(X).
This is a shorthand way of calling the eval method of the classregtree
class. The predicted species is the left leaf node at the bottom of the tree
in the previous view.
5 You can use a variety of other methods of the classregtree class, such as
cutvar and cuttype to get more information about the split at node 6 that
makes the final distinction between versicolor and virginica:
var6 = cutvar(t,6) % What variable determines the split?
var6 =
'PW'
type6 = cuttype(t,6) % What type of split is it?
type6 =
'continuous'
6 Classification trees fit the original (training) data well, but may do a poor
job of classifying new values. Lower branches, especially, may be strongly
affected by outliers. A simpler tree often avoids overfitting. You can use
the prune method of the classregtree class to find the next largest tree
from an optimal pruning sequence:
pruned = prune(t,'level',1)
pruned =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 class = versicolor
7 class = virginica
view(pruned)
12-12
Classification Trees
To find the best classification tree, employing the techniques of resubstitution
and cross-validation, use the test method of the classregtree class.
References
[1] Breiman, L., et al., Classification and Regression Trees, Chapman & Hall,
Boca Raton, 1993.
12-13
12 Classification
Classification Using Nearest Neighbors
In this section...
“Pairwise Distance” on page 12-14
“k-Nearest Neighbor Search” on page 12-17
Pairwise Distance
Categorizing query points based on their distance to points in a training data
set can be a simple yet effective way of classifying new points. You can use a
variety of metrics to determine the distance, described in the following section.
Use pdist2 to find the distance between a sets of data and query points.
Distance Metrics
Given an mx-by-n data matrix X, which is treated as mx (1-by-n) row vectors
x1, x2, ..., xmx, and my-by-n data matrix Y, which is treated as my (1-by-n)
row vectors y1, y2, ...,ymy, the various distances between the vector xs and yt
are defined as follows:
• Euclidean distance
2
dst = ( xs − yt )( xs − yt )′
Notice that the Euclidean distance is a special case of the Minkowski
metric, where p = 2.
• Standardized Euclidean distance
2
dst = ( xs − yt )V −1 ( xs − yt )′
where V is the n-by-n diagonal matrix whose jth diagonal element is S(j)2,
where S is the vector containing the inverse weights.
• Mahalanobis distance
2
dst = ( xs − yt )C −1 ( xs − yt )′
where C is the covariance matrix.
12-14
Classification Using Nearest Neighbors
• City block metric
n
dst = ∑ xsj − ytj
j =1
Notice that the city block distance is a special case of the Minkowski
metric, where p = 1.
• Minkowski metric
n p
dst = p ∑ xsj − ytj
j =1
Notice that for the special case of p = 1, the Minkowski metric gives the
city block metric, for the special case of p = 2, the Minkowski metric gives
the Euclidean distance, and for the special case of p = ∞, the Minkowski
metric gives the Chebychev distance.
• Chebychev distance
{
dst = max j xsj − ytj }
Notice that the Chebychev distance is a special case of the Minkowski
metric, where p = ∞.
• Cosine distance
⎛ xs yt′ ⎞
dst = ⎜ 1 − ⎟
⎜
⎝ ( xs xs′ ) ( yt yt′ ) ⎟⎠
• Correlation distance
( xs − xs ) ( yt − yt )′
dst = 1 −
( xs − xs ) ( xs − xs )′ ( yt − yt ) ( yt − yt )′
where
12-15
12 Classification
1
xs = ∑ xsj
n j
and
1
yt = ∑ ytj
n j
• Hamming distance
dst = (#( xsj ≠ ytj ) / n)
• Jaccard distance
( ) ((
# ⎡ xsj ≠ ytj ∩ xsj ≠ 0 ∪ ytj ≠ 0 ⎤
⎣ ⎦) ( ))
dst =
⎡ ( ) (
# xsj ≠ 0 ∪ ytj ≠ 0
⎣
⎤
⎦ )
• Spearman distance
( rs − rs ) ( rt − rt )′
dst = 1 −
( rs − rs ) ( rs − rs )′ ( rt − rt ) ( rt − rt )′
where
- rsj is the rank of xsj taken over x1j, x2j, ...xmx,j, as computed by tiedrank
- rtj is the rank of ytj taken over y1j, y2j, ...ymy,j, as computed by tiedrank
- rs and rt are the coordinate-wise rank vectors of xs and yt, i.e.,
rs = (rs1, rs2, ... rsn) and rt = (rt1, rt2, ... rtn)
1 ( n + 1)
- rs = ∑
n j
rsj =
2
1 ( n + 1)
- rt = ∑
n j
rtj =
2
12-16
Classification Using Nearest Neighbors
k-Nearest Neighbor Search
Given a set X of n points and a distance function D, k-nearest neighbor (kNN)
search allows you to find the k closest points in X to a query point or set of
points. The kNN search technique and kNN-based algorithms are widely used
as benchmark learning rules — the relative simplicity of the kNN search
technique makes it easy to compare the results from other classification
techniques to kNN results. They have been used in various areas such as
bioinformatics, image processing and data compression, document retrieval,
computer vision, multimedia database, and marketing data analysis. You
can use kNN search for other machine learning algorithms, such as kNN
classification, local weighted regression, missing data imputation and
interpolation, and density estimation. You can also use kNN search with
many distance-based learning functions, such as K-means clustering.
k-Nearest Neighbor Search Using Exhaustive Search
When your input data meets any of the following criteria, knnsearch uses the
exhaustive search method by default to find the k-nearest neighbors:
• The number of columns of X is more than 10
• X is sparse
• The distance measure is one of the following:
- 'seuclidean'
- 'mahalanobis'
- 'cosine'
- 'correlation'
- 'spearman'
- 'hamming'
- 'jaccard'
- A custom distance function
knnsearch will also use the exhaustive search method if your search object
is an ExhaustiveSearcher object. The exhaustive search method finds the
distance from each query point to every point in X, ranks them in ascending
12-17
12 Classification
order, and returns the k points with the smallest distances. For example, this
diagram shows the k=3 nearest neighbors.
k-Nearest Neighbor Search Using a kd-Tree
When your input data meets all of the following criteria, knnsearch creates a
kd-tree by default to find the k-nearest neighbors.
• The number of columns of X is less than 10
• X is not sparse.
• The distance measure is one of the following:
- 'euclidean' (default)
- 'cityblock'
12-18
Classification Using Nearest Neighbors
- 'minkowski'
- 'chebychev'
knnsearch also uses a kd-tree if your search object is a KDTreeSearcher object.
kd-trees divide your data into nodes with at most BucketSize (default is
50) points per node, based on coordinates (as opposed to categories). The
following diagrams illustrate this concept using patch objects to color code
the different “buckets.”
When you want to find the k-nearest neighbors to a given query point,
knnsearch performs the following steps:
1 Determine the node to which the query point belongs. In the following
example, the query point (73,21.5) belongs to Node 12.
2 Find the closest k points within that node and its distance to the query
point. In the following example, the points in red squares are equidistant
12-19
12 Classification
from the query point, and are the closest points to the query point within
Node 12.
3 Choose all other nodes having any area that is within the same distance,
in any direction, from the query point to the kth closest point. In this
example, only Node 13 overlaps the solid black circle centered at the query
point with radius equal to the distance to the closest points within Node 12.
4 Search nodes within that range for any points closer to the query point. In
the following example, the point circled in red is clearly closer to the query
point than those within Node 12.
Using a kd-tree for large datasets with fewer than 10 dimensions (columns)
can be much more efficient than using the exhaustive search method, as
knnsearch needs to calculate only a subset of the distances. To maximize the
efficiency of kd-trees, use a KDTreeSearcher object.
What Are Search Objects?
Objects are, in short, a convenient way of storing information. Classes of
related objects (for example, all search objects) have the same properties
12-20
Classification Using Nearest Neighbors
with values and types relevant to a specified search method. In addition to
storing information within objects, you can perform certain actions (called
methods) on objects.
All search objects have a knnsearch method specific to that class. This allows
you to perform a k-nearest neighbors search on your object in the most efficient
way for that specific object type. In addition, there is a generic knnsearch
function which performs the search without creating or using an object.
To determine which type of object and search method is best for your data,
consider the following:
• Does your data have many columns, say more than 10? The
ExhaustiveSearcher object may perform better.
• Is your data sparse? Use the ExhaustiveSearcher object.
• Do you want to use one of the following distance measures to find the
nearest neighbors? Use the ExhaustiveSearcher object.
- 'seuclidean'
- 'mahalanobis'
- 'cosine'
- 'correlation'
- 'spearman'
- 'hamming'
- 'jaccard'
- A custom distance function
• Is your dataset very large (but with fewer than 10 columns)? Use the
KDTreeSearcher object.
• Are you searching for the nearest neighbors for a large number of query
points? Use the KDTreeSearcher object.
For more detailed information on object-oriented programming in MATLAB,
see Object-Oriented Programming.
12-21
12 Classification
Example: Classifying Query Data Using knnsearch
Classify a new point based on the last two columns of the Fisher iris data.
Using only the last two columns makes it easier to plot:
load fisheriris
x = meas(:,3:4);
gscatter(x(:,1),x(:,2),species)
set(legend,'location','best')
Plot the new point:
newpoint = [5 1.45];
line(newpoint(1),newpoint(2),'marker','x','color','k',...
'markersize',10,'linewidth',2)
12-22
Classification Using Nearest Neighbors
Now find the 10 sample points closest to the new point:
[n,d] = knnsearch(x,newpoint,'k',10)
line(x(n,1),x(n,2),'color',[.5 .5 .5],'marker','o',...
'linestyle','none','markersize',10)
12-23
12 Classification
It appears that knnsearch has found only the nearest eight neighbors. In fact,
this particular dataset contains duplicate values:
x(n,:)
ans =
5.0000 1.5000
4.9000 1.5000
4.9000 1.5000
5.1000 1.5000
5.1000 1.6000
4.8000 1.4000
5.0000 1.7000
4.7000 1.4000
4.7000 1.4000
4.7000 1.5000
12-24
Classification Using Nearest Neighbors
To make duplicate values visible on the plot, use the following code:
% jitter to make repeated points visible
xj = x + .05*(rand(150,2)-.5);
gscatter(xj(:,1),xj(:,2),species)
The jittered points do not affect any analysis of the data, only the
visualization. This example does not jitter the points.
Make the axes equal so the calculated distances correspond to the apparent
distances on the plot axis equal and zoom in to see the neighbors better:
set(gca,'xlim',[4.5 5.5],'ylim',[1 2]); axis square
Now find the species of the 10 neighbors:
tabulate(species(n))
12-25
12 Classification
Value Count Percent
virginica 2 20.00%
versicolor 8 80.00%
Using a rule based on the majority vote of the 10 nearest neighbors, you can
classify this new point as a versicolor.
You can also visually identify the neighbors by drawing a circle around the
group of them:
% Define the center and diameter of a circle, based on the
% location of the new point:
ctr = newpoint - d(end);
diameter = 2*d(end);
% Draw a circle around the 10 nearest neighbors:
h = rectangle('position',[ctr,diameter,diameter],...
'curvature',[1 1]);
set(h,'linestyle',':')
12-26
Classification Using Nearest Neighbors
Using the same dataset, find the 10 nearest neighbors to three new points:
figure
newpoint2 = [5 1.45;6 2;2.75 .75];
gscatter(x(:,1),x(:,2),species)
legend('location','best')
[n2,d2] = knnsearch(x,newpoint2,'k',10);
line(x(n2,1),x(n2,2),'color',[.5 .5 .5],'marker','o',...
'linestyle','none','markersize',10)
line(newpoint2(:,1),newpoint2(:,2),'marker','x','color','k',...
'markersize',10,'linewidth',2,'linestyle','none')
12-27
12 Classification
Find the species of the 10 nearest neighbors for each new point:
tabulate(species(n(1,:)))
Value Count Percent
virginica 2 20.00%
versicolor 8 80.00%
tabulate(species(n(2,:)))
Value Count Percent
virginica 10 100.00%
tabulate(species(n(3,:)))
Value Count Percent
versicolor 7 70.00%
setosa 3 30.00%
12-28
Classification Using Nearest Neighbors
For further examples using the knnsearch methods and function, see the
individual reference pages.
12-29
12 Classification
Regression and Classification by Bagging Decision Trees
In this section...
“Introduction” on page 12-30
“Examples” on page 12-31
“Regression of Insurance Risk Rating for Car Imports” on page 12-31
“Classifying Radar Returns for Ionosphere Data” on page 12-40
“Plotting a Performance Curve” on page 12-49
Introduction
Bagging, which stands for “bootstrap aggregation”, is a type of ensemble
learning. To bag a weak learner such as a decision tree on a dataset, generate
many bootstrap replicas of this dataset and grow decision trees on these
replicas. Obtain each bootstrap replica by randomly selecting N observations
out of N with replacement, where N is the dataset size. To find the predicted
response of a trained ensemble, take an average over predictions from
individual trees.
Bagging works by reducing variance of an unbiased base learner such as
a decision tree. The base learner must be unstable; its configuration must
vary significantly from one bootstrap replica to another. A decision tree with
fine leaves satisfies this requirement.
In the process of training decision trees on bootstrap replicas of input data,
you can also randomly select input variables, or features, for each decision
split. Instead of attempting splits on all variables at each node of a decision
tree, the algorithm only looks at feature subsets of fixed size randomly
selected from input variables. For example, if data has 10 input features,
you can select 3 features at random for each decision split. If the best split
uses one of the omitted seven variables, it is suboptimal. Surprisingly, this
technique tends to improve the predictive power of the ensemble, as random
selection of features reduces correlation between trees in the ensemble and
increases the overall predictive power.
Several features of bagged decision trees make them a unique algorithm.
Drawing N out of N observations with replacement omits on average 37% of
12-30
Regression and Classification by Bagging Decision Trees
observations for each decision tree. These are "out-of-bag" observations. You
can use them to estimate the predictive power and feature importance. For
each observation, you can estimate the out-of-bag prediction by averaging over
predictions from all trees in the ensemble for which this observation is out of
bag. You can then compare the computed prediction against the true response
for this observation. By comparing the out-of-bag predicted responses against
the true responses for all observations used for training, you can estimate the
average out-of-bag error. This out-of-bag average is an unbiased estimator of
the true ensemble error. You can also obtain out-of-bag estimates of feature
importance by randomly permuting out-of-bag data across one variable or
column at a time and estimating the increase in the out-of-bag error due to
this permutation. The larger the increase, the more important the feature.
Thus, you do not need to supply test data for bagged ensembles because you
obtain reliable estimates of the predictive power and feature importance in
the process of training, which is an attractive feature of bagging.
Another attractive feature of bagged decision trees is the proximity matrix.
Every time two observations land on the same leaf of a tree, their proximity
increases by one. For normalization, sum these proximities over all trees
in the ensemble and divided by the number of trees. The resulting matrix
is symmetric with diagonal elements equal to 1 and off-diagonal elements
ranging from 0 to 1. You can use this matrix for finding outlier observations
and discovering clusters in the data through multidimensional scaling.
The following examples showcase the Statistics Toolbox TreeBagger class
and CompactTreeBagger class functionalities.
Examples
The following examples show how to use ensembles of decision trees for
regression and classification.
Regression of Insurance Risk Rating for Car Imports
In this example, use a database of 1985 car imports with 205 observations,
25 input variables, and one response variable, insurance risk rating, or
"symboling". The first 15 variables are numeric and the last 10 are categorical.
The symboling index takes integer values from -3 to 3.
First, load the dataset and split it into predictor and response arrays:
12-31
12 Classification
load imports-85;
Y = X(:,1);
X = X(:,2:end);
Because bagging uses randomized data drawings, its exact outcome depends
on the initial random seed. To reproduce the exact results in this example,
use the random stream settings
s = RandStream('mt19937ar','seed',1945);
RandStream.setDefaultStream(s);
Finding the Optimal Leaf Size
For regression, the general rule is to set leaf size to 5 and select one third of
input features for decision splits at random. In the following step, verify the
optimal leaf size by comparing mean squared errors obtained by regression for
various leaf sizes. oobError computes MSE versus the number of grown trees.
It is necessary to set oobpred to 'on' to obtain out-of-bag predictions later.
leaf = [1 5 10 20 50 100];
col = 'rgbcmy';
figure(1);
for i=1:length(leaf)
b = TreeBagger(50,X,Y,'method','r','oobpred','on',...
'cat',16:25,'minleaf',leaf(i));
plot(oobError(b),col(i));
hold on;
end
xlabel('Number of Grown Trees');
ylabel('Mean Squared Error');
legend({'1' '5' '10' '20' '50' '100'},'Location','NorthEast');
hold off;
12-32
Regression and Classification by Bagging Decision Trees
The red (leaf size 1) curve gives the lowest MSE values.
Estimating Feature Importance
In practical applications, you typically grow ensembles with hundreds of
trees. Only 50 trees were used in the previous exercise for faster processing.
Now that you have estimated the optimal leaf size, grow a larger ensemble
with 100 trees and use it for estimation of feature importance:
b = TreeBagger(100,X,Y,'method','r','oobvarimp','on',...
'cat',16:25,'minleaf',1);
Inspect the error curve again to make sure nothing went wrong during
training:
figure(2);
plot(oobError(b));
xlabel('Number of Grown Trees');
ylabel('Out-of-Bag Mean Squared Error');
12-33
12 Classification
Prediction ability should depend more on important features and less on
unimportant features. You can use this idea to measure feature importance.
For each feature, you can permute the values of this feature across all of
the observations in the data set and measure how much worse the mean
squared error (MSE) becomes after the permutation. You can repeat this
for each feature.
Using the following code, plot the increase in MSE due to permuting out-of-bag
observations across each input variable. The OOBPermutedVarDeltaError
array stores the increase in MSE averaged over all trees in the ensemble and
divided by the standard deviation taken over the trees, for each variable. The
larger this value, the more important the variable. Imposing an arbitrary
cutoff at 0.65, you can select the five most important features.
figure(3);
bar(b.OOBPermutedVarDeltaError);
xlabel('Feature Number');
ylabel('Out-Of-Bag Feature Importance');
idxvar = find(b.OOBPermutedVarDeltaError>0.65)
idxvar =
1 2 4 16 19
12-34
Regression and Classification by Bagging Decision Trees
The OOBIndices property of TreeBagger keeps track of which observations
are out of bag for what trees. Using this property, you can monitor the
fraction of observations in the training data that are in bag for all trees. The
curve starts at approximately 2/3, the fraction of unique observations selected
by one bootstrap replica, and goes down to zero at approximately 10 trees.
finbag = zeros(1,b.NTrees);
for t=1:b.NTrees
finbag(t) = sum(all(~b.OOBIndices(:,1:t),2));
end
finbag = finbag / size(X,1);
figure(4);
plot(finbag);
xlabel('Number of Grown Trees');
12-35
12 Classification
ylabel('Fraction of in-Bag Observations');
Growing Trees on a Reduced Set of Features
Using just the five most powerful features selected in “Estimating Feature
Importance” on page 12-33, find out if it is possible to obtain a similar
predictive power. To begin, grow 100 trees on these features only. The first
three of the five selected features are numeric and the last two are categorical.
b5v = TreeBagger(100,X(:,idxvar),Y,'method','r',...
'oobvarimp','on','cat',4:5,'minleaf',1);
figure(5);
plot(oobError(b5v));
12-36
Regression and Classification by Bagging Decision Trees
xlabel('Number of Grown Trees');
ylabel('Out-of-Bag Mean Squared Error');
figure(6);
bar(b5v.OOBPermutedVarDeltaError);
xlabel('Feature Index');
ylabel('Out-of-Bag Feature Importance');
12-37
12 Classification
These five most powerful features give the same MSE as the full set, and
the ensemble trained on the reduced set ranks these features similarly to
each other. Features 1 and 2 from the reduced set perhaps could be removed
without a significant loss in the predictive power.
Finding Outliers
To find outliers in the training data, compute the proximity matrix using
fillProximities:
b5v = fillProximities(b5v);
The method normalizes this measure by subtracting the mean outlier measure
for the entire sample, taking the magnitude of this difference and dividing the
result by the median absolute deviation for the entire sample.
figure(7);
hist(b5v.OutlierMeasure);
xlabel('Outlier Measure');
ylabel('Number of Observations');
12-38
Regression and Classification by Bagging Decision Trees
Discovering Clusters in the Data
By applying multidimensional scaling to the computed matrix of proximities,
you can inspect the structure of the input data and look for possible clusters of
observations. The mdsProx method returns scaled coordinates and eigenvalues
for the computed proximity matrix. If run with the colors option, this method
makes a scatter plot of two scaled coordinates, first and second by default.
figure(8);
[~,e] = mdsProx(b5v,'colors','k');
xlabel('1st Scaled Coordinate');
ylabel('2nd Scaled Coordinate');
Assess the relative importance of the scaled axes by plotting the first 20
eigenvalues:
figure(9);
bar(e(1:20));
xlabel('Scaled Coordinate Index');
ylabel('Eigenvalue');
12-39
12 Classification
Saving the Ensemble Configuration for Future Use
To use the trained ensemble for predicting the response on unseen data, store
the ensemble to disk and retrieve it later. If you do not want to compute
predictions for out-of-bag data or reuse training data in any other way, there
is no need to store the ensemble object itself. Saving the compact version of
the ensemble would be enough in this case. Extract the compact object from
the ensemble:
c = compact(b5v)
c =
Ensemble with 100 decision trees:
Method: regression
Nvars: 5
This object can be now saved into a *.mat file as usual.
Classifying Radar Returns for Ionosphere Data
You can also use ensembles of decision trees for classification. For this
example, use ionosphere data with 351 observations and 34 real-valued
predictors. The response variable is categorical with two levels:
12-40
Regression and Classification by Bagging Decision Trees
• 'g' for good radar returns
• 'b' for bad radar returns
The goal is to predict good or bad using a set of 34 measurements.
The workflow is similar to the one for “Regression of Insurance Risk Rating
for Car Imports” on page 12-31. Again, fix the initial random seed, grow 50
trees, inspect how the ensemble error changes with accumulation of trees, and
estimate feature importance. For classification, it is best to set the minimal
leaf size to 1 and select the square root of the total number of features for
each decision split at random. These are the default settings for a TreeBagger
used for classification.
load ionosphere;
s = RandStream('mt19937ar','seed',1945);
RandStream.setDefaultStream(s);
b = TreeBagger(50,X,Y,'oobvarimp','on');
figure(10);
plot(oobError(b));
xlabel('Number of Grown Trees');
ylabel('Out-of-Bag Classification Error');
The method trains ensembles with few trees on observations that are in bag
for all trees. For such observations, it is not possible to compute the true
12-41
12 Classification
out-of-bag prediction and TreeBagger returns the most probable class for
classification and the sample mean for regression. You can change the default
value returned for in-bag observations using the DefaultYfit property. If
you set the default value to an empty string for classification, the method
excludes in-bag observations from computation of the out-of-bag error. In this
case, the curve is more variable when the number of trees is small, either
because some observations are never out of bag (and are therefore excluded)
or because their predictions are based on few trees.
b.DefaultYfit = '';
figure(11);
plot(oobError(b));
xlabel('Number of Grown Trees');
ylabel('Out-of-Bag Error Excluding in-Bag Observations');
The OOBIndices property of TreeBagger keeps track of which observations
are out of bag for what trees. Using this property, you can monitor the
fraction of observations in the training data that are in bag for all trees. The
curve starts at approximately 2/3, the fraction of unique observations selected
by one bootstrap replica, and goes down to zero at approximately 10 trees.
finbag = zeros(1,b.NTrees);
for t=1:b.NTrees
finbag(t) = sum(all(~b.OOBIndices(:,1:t),2));
end
12-42
Regression and Classification by Bagging Decision Trees
finbag = finbag / size(X,1);
figure(12);
plot(finbag);
xlabel('Number of Grown Trees');
ylabel('Fraction of in-Bag Observations');
Now estimate feature importance:
figure(13);
bar(b.OOBPermutedVarDeltaError);
xlabel('Feature Index');
ylabel('Out-of-Bag Feature Importance');
idxvar = find(b.OOBPermutedVarDeltaError>0.8)
idxvar =
3 4 5 7 8
12-43
12 Classification
Having selected the five most important features, grow a larger ensemble on
the reduced feature set. Save time by not permuting out-of-bag observations
to obtain new estimates of feature importance for the reduced feature set (set
oobvarimp to 'off'). You would still be interested in obtaining out-of-bag
estimates of classification error (set oobpred to 'on').
b5v = TreeBagger(100,X(:,idxvar),Y,'oobpred','on');
figure(14);
plot(oobError(b5v));
xlabel('Number of Grown Trees');
ylabel('Out-of-Bag Classification Error');
12-44
Regression and Classification by Bagging Decision Trees
For classification ensembles, in addition to classification error (fraction of
misclassified observations), you can also monitor the average classification
margin. For each observation, the margin is defined as the difference between
the score for the true class and the maximal score for other classes predicted
by this tree. The cumulative classification margin uses the scores averaged
over all trees and the mean cumulative classification margin is the cumulative
margin averaged over all observations. The oobMeanMargin method with
the 'mode' argument set to 'cumulative' (default) shows how the mean
cumulative margin changes as the ensemble grows: every new element in the
returned array represents the cumulative margin obtained by including a
new tree in the ensemble. If training is successful, you would expect to see a
gradual increase in the mean classification margin.
For decision trees, a classification score is the probability of observing an
instance of this class in this tree leaf. For example, if the leaf of a grown
decision tree has five 'good' and three 'bad' training observations in it, the
scores returned by this decision tree for any observation fallen on this leaf are
5/8 for the 'good' class and 3/8 for the 'bad' class. These probabilities are
referred to as 'scores' for consistency with other classifiers that may not
have an obvious interpretation for numeric values of returned predictions.
figure(15);
plot(oobMeanMargin(b5v));
12-45
12 Classification
xlabel('Number of Grown Trees');
ylabel('Out-of-Bag Mean Classification Margin');
Again, compute the matrix of proximities and look at the distribution of
outlier measures. Unlike regression, outlier measures for classification
ensembles are computed within each class separately.
b5v = fillProximities(b5v);
figure(16);
hist(b5v.OutlierMeasure);
xlabel('Outlier Measure');
ylabel('Number of Observations');
12-46
Regression and Classification by Bagging Decision Trees
All extreme outliers for this dataset come from the 'good' class:
b5v.Y(b5v.OutlierMeasure>40)
ans =
'g'
'g'
'g'
'g'
'g''
Just like for regression, you can plot scaled coordinates, displaying the two
classes in different colors using the colors argument of mdsProx. This
argument takes a string in which every character represents a color. To find
out the order of classes used by the ensemble, look at the ClassNames property:
b5v.ClassNames
ans =
'g'
'b'
12-47
12 Classification
The 'good' class is first and the 'bad' class is second. Display scaled
coordinates using red for 'good' and blue for 'bad' observations.
figure(17);
[s,e] = mdsProx(b5v,'colors','rb');
xlabel('1st Scaled Coordinate');
ylabel('2nd Scaled Coordinate');
Again, plot the first 20 eigenvalues obtained by scaling. The first eigenvalue in
this case clearly dominates and the first scaled coordinate is most important.
figure(18);
bar(e(1:20));
xlabel('Scaled Coordinate Index');
ylabel('Eigenvalue');
12-48
Regression and Classification by Bagging Decision Trees
Plotting a Performance Curve
Another way of exploring the performance of a classification ensemble
is to plot its Receiver Operating Characteristic (ROC) curve or another
performance curve suitable for the current problem. First, you need to obtain
predictions for out-of-bag observations. For a classification ensemble, the
oobPredict method returns a cell array of classification labels ('g' or 'b'
for ionosphere data) as the first output argument and a numeric array of
scores as the second output argument. The returned array of scores has two
columns, one for each class. In this case, the first column is for the 'good'
class and the second column is for the 'bad' class. One of the columns in the
score matrix is redundant because the scores represent class probabilities in
tree leaves and by definition add up to 1.
[Yfit,Sfit] = oobPredict(b5v);
Use the perfcurve utility (described in more detail in “Performance Curves”
on page 12-53) to compute a performance curve. By default, perfcurve
returns the standard ROC curve, which is the true positive rate versus false
positive rate. perfcurve requires true class labels, scores, and the positive
class label for input. In this case, choose the 'good' class as positive. The
scores for this class are in the first column of Sfit.
[fpr,tpr] = perfcurve(b5v.Y,Sfit(:,1),'g');
12-49
12 Classification
figure(19);
plot(fpr,tpr);
xlabel('False Positive Rate');
ylabel('True Positive Rate');
Instead of the standard ROC curve, you might want to plot, for example,
ensemble accuracy versus threshold on the score for the 'good' class. The
ycrit input argument of perfcurve lets you specify the criterion for the
y-axis, and the third output argument of perfcurve returns an array of
thresholds for the positive class score. Accuracy is the fraction of correctly
classified observations, or equivalently, one minus classification error.
[fpr,accu,thre] = perfcurve(b5v.Y,Sfit(:,1),'g','ycrit','accu');
figure(20);
plot(thre,accu);
xlabel('Threshold for ''good'' Returns');
ylabel('Classification Accuracy');
12-50
Regression and Classification by Bagging Decision Trees
The curve shows a flat region indicating that any threshold from 0.2 to 0.6
is a reasonable choice. By default, the function assigns classification labels
using 0.5 as the boundary between the two classes. You can find exactly
what accuracy this corresponds to:
i50 = find(accu>=0.50,1,'first')
accu(abs(thre-0.5)<eps)
returns
i50 =
ans =
0.9430
The maximal accuracy is a little higher than the default one.
[maxaccu,iaccu] = max(accu)
returns
12-51
12 Classification
maxaccu =
0.9459
iaccu =
91
The optimal threshold is therefore:
thre(iaccu)
ans =
0.5056
12-52
Performance Curves
Performance Curves
In this section...
“Introduction” on page 12-53
“What are ROC Curves?” on page 12-53
“Evaluating Classifier Performance Using perfcurve” on page 12-53
Introduction
After a classification algorithm such as NaiveBayes or TreeBagger has
trained on data, you may want to examine the performance of this algorithm
on a specific test dataset. One common way of doing this would be to compute
a gross measure of performance such as quadratic loss, accuracy, such as
quadratic loss or accuracy, averaged over the entire test dataset.
What are ROC Curves?
You may want to inspect the classifier performance more closely, for
example, by plotting a Receiver Operating Characteristic (ROC) curve. By
definition, a ROC curve [1,2] shows true positive rate versus false positive
rate (equivalently, sensitivity versus 1–specificity) for different thresholds of
the classifier output. You can use it, for example, to find the threshold that
maximizes the classification accuracy or to assess, in more broad terms, how
the classifier performs in the regions of high sensitivity and high specificity.
Evaluating Classifier Performance Using perfcurve
perfcurve computes measures for a plot of classifier performance. You can
use this utility to evaluate classifier performance on test data after you train
the classifier. Various measures such as mean squared error, classification
error, or exponential loss can summarize the predictive power of a classifier
in a single number. However, a performance curve offers more information
as it lets you explore the classifier performance across a range of thresholds
on its output.
You can use perfcurve with any classifier or, more broadly, with any method
that returns a numeric score for an instance of input data. By convention
adopted here,
12-53
12 Classification
• A high score returned by a classifier for any given instance signifies that
the instance is likely from the positive class.
• A low score signifies that the instance is likely from the negative classes.
For some classifiers, you can interpret the score as the posterior probability
of observing an instance of the positive class at point X. An example of such
a score is the fraction of positive observations in a leaf of a decision tree. In
this case, scores fall into the range from 0 to 1 and scores from positive and
negative classes add up to unity. Other methods can return scores ranging
between minus and plus infinity, without any obvious mapping from the
score to the posterior class probability.
perfcurve does not impose any requirements on the input score range.
Because of this lack of normalization, you can use perfcurve to process scores
returned by any classification, regression, or fit method. perfcurve does
not make any assumptions about the nature of input scores or relationships
between the scores for different classes. As an example, consider a problem
with three classes, A, B, and C, and assume that the scores returned by some
classifier for two instances are as follows:
A B C
instance 1 0.4 0.5 0.1
instance 2 0.4 0.1 0.5
If you want to compute a performance curve for separation of classes A and B,
with C ignored, you need to address the ambiguity in selecting A over B. You
could opt to use the score ratio, s(A)/s(B), or score difference, s(A)-s(B);
this choice could depend on the nature of these scores and their normalization.
perfcurve always takes one score per instance. If you only supply scores for
class A, perfcurve does not distinguish between observations 1 and 2. The
performance curve in this case may not be optimal.
perfcurve is intended for use with classifiers that return scores, not those
that return only predicted classes. As a counter-example, consider a decision
tree that returns only hard classification labels, 0 or 1, for data with two
classes. In this case, the performance curve reduces to a single point because
classified instances can be split into positive and negative categories in one
way only.
12-54
Performance Curves
For input, perfcurve takes true class labels for some data and scores assigned
by a classifier to these data. By default, this utility computes a Receiver
Operating Characteristic (ROC) curve and returns values of 1–specificity,
or false positive rate, for X and sensitivity, or true positive rate, for Y. You
can choose other criteria for X and Y by selecting one out of several provided
criteria or specifying an arbitrary criterion through an anonymous function.
You can display the computed performance curve using plot(X,Y).
perfcurve can compute values for various criteria to plot either on the x- or
the y-axis. All such criteria are described by a 2-by-2 confusion matrix, a
2-by-2 cost matrix, and a 2-by-1 vector of scales applied to class counts.
The confusion matrix, C, is defined as
⎛ TP FN ⎞
⎜ ⎟
⎝ FP
where TN ⎠
• P stands for "positive".
• N stands for "negative".
• T stands for "true".
• F stands for "false".
For example, the first row of the confusion matrix defines how the classifier
identifies instances of the positive class: C(1,1) is the count of correctly
identified positive instances and C(1,2) is the count of positive instances
misidentified as negative.
The cost matrix defines the cost of misclassification for each category:
⎛ Cost( P | P) Cost( N | P) ⎞
⎜ ⎟
⎝ Cost
where ( P | N ) Cost
Cost(I|J) ( N cost
is the | N ) of
⎠ assigning an instance of class J to class I.
Usually Cost(I|J)=0 for I=J. For flexibility, perfcurve allows you to specify
nonzero costs for correct classification as well.
12-55
12 Classification
The two scales include prior information about class probabilities.
perfcurve computes these scales by taking scale(P)=prior(P)*N and
scale(N)=prior(N)*P and normalizing the sum scale(P)+scale(N)
to 1. P=TP+FN and N=TN+FP are the total instance counts in the positive
and negative class, respectively. The function then applies the scales as
multiplicative factors to the counts from the corresponding class: perfcurve
multiplies counts from the positive class by scale(P) and counts from the
negative class by scale(N). Consider, for example, computation of positive
predictive value, PPV = TP/(TP+FP). TP counts come from the positive class
and FP counts come from the negative class. Therefore, you need to scale TP
by scale(P) and FP by scale(N), and the modified formula for PPV with prior
probabilities taken into account is now:
scale( P) * TP
PPV =
If all scoresscale( P)data
in the + scale
* TP are ( N )a* certain
above FP threshold, perfcurve classifies all
instances as 'positive'. This means that TP is the total number of instances
in the positive class and FP is the total number of instances in the negative
class. In this case, PPV is simply given by the prior:
prior( P)
PPV =
The perfcurve P) + prior
prior(function (N)
returns two vectors, X and Y, of performance
measures. Each measure is some function of confusion, cost, and scale
values. You can request specific measures by name or provide a function
handle to compute a custom measure. The function you provide should take
confusion, cost, and scale as its three inputs and return a vector of output
values.
The criterion for X must be a monotone function of the positive classification
count, or equivalently, threshold for the supplied scores. If perfcurve cannot
perform a one-to-one mapping between values of the X criterion and score
thresholds, it exits with an error message.
By default, perfcurve computes values of the X and Y criteria for all possible
score thresholds. Alternatively, it can compute a reduced number of specific X
values supplied as an input argument. In either case, for M requested values,
perfcurve computes M+1 values for X and Y. The first value out of these M+1
values is special. perfcurve computes it by setting the TP instance count
12-56
Performance Curves
to zero and setting TN to the total count in the negative class. This value
corresponds to the 'reject all' threshold. On a standard ROC curve, this
translates into an extra point placed at (0,0).
If there are NaN values among input scores, perfcurve can process them
in either of two ways:
• It can discard rows with NaN scores.
• It can add them to false classification counts in the respective class.
That is, for any threshold, instances with NaN scores from the positive class
are counted as false negative (FN), and instances with NaN scores from the
negative class are counted as false positive (FP). In this case, the first value
of X or Y is computed by setting TP to zero and setting TN to the total count
minus the NaN count in the negative class. For illustration, consider an
example with two rows in the positive and two rows in the negative class,
each pair having a NaN score:
Class Score
Negative 0.2
Negative NaN
Positive 0.7
Positive NaN
If you discard rows with NaN scores, then as the score cutoff varies, perfcurve
computes performance measures as in the following table. For example, a
cutoff of 0.5 corresponds to the middle row where rows 1 and 3 are classified
correctly, and rows 2 and 4 are omitted.
TP FN FP TN
0 1 0 1
1 0 0 1
1 0 1 0
If you add rows with NaN scores to the false category in their respective
classes, perfcurve computes performance measures as in the following table.
For example, a cutoff of 0.5 corresponds to the middle row where now rows
12-57
12 Classification
2 and 4 are counted as incorrectly classified. Notice that only the FN and FP
columns differ between these two tables.
TP FN FP TN
0 2 1 1
1 1 1 1
1 1 2 0
For data with three or more classes, perfcurve takes one positive class and a
list of negative classes for input. The function computes the X and Y values
using counts in the positive class to estimate TP and FN, and using counts in
all negative classes to estimate TN and FP. perfcurve can optionally compute
Y values for each negative class separately and, in addition to Y, return a
matrix of size M-by-C, where M is the number of elements in X or Y and C is
the number of negative classes. You can use this functionality to monitor
components of the negative class contribution. For example, you can plot TP
counts on the X-axis and FP counts on the Y-axis. In this case, the returned
matrix shows how the FP component is split across negative classes.
You can also use perfcurve to estimate confidence intervals. perfcurve
computes confidence bounds using either cross-validation or bootstrap. If you
supply cell arrays for labels and scores, perfcurve uses cross-validation
and treats elements in the cell arrays as cross-validation folds. If you set
input parameter NBoot to a positive integer, perfcurve generates nboot
bootstrap replicas to compute pointwise confidence bounds.
perfcurve estimates the confidence bounds using one of two methods:
• Vertical averaging (VA) — estimate confidence bounds on Y and T at
fixed values of X. Use the XVals input parameter to use this method for
computing confidence bounds.
• Threshold averaging (TA) — estimate confidence bounds for X and Y at
fixed thresholds for the positive class score. Use the TVals input parameter
to use this method for computing confidence bounds.
To use observation weights instead of observation counts, you can use
the 'Weights' parameter in your call to perfcurve. When you use this
parameter, to compute X, Y and T or to compute confidence bounds by
cross-validation, perfcurve uses your supplied observation weights instead of
12-58
Performance Curves
observation counts. To compute confidence bounds by bootstrap, perfcurve
samples N out of N with replacement using your weights as multinomial
sampling probabilities.
12-59
12 Classification
12-60
13
Markov Models
• “Introduction” on page 13-2
• “Markov Chains” on page 13-3
• “Hidden Markov Models” on page 13-5
13 Markov Models
Introduction
Markov processes are examples of stochastic processes—processes that
generate random sequences of outcomes or states according to certain
probabilities. Markov processes are distinguished by being memoryless—their
next state depends only on their current state, not on the history that led them
there. Models of Markov processes are used in a wide variety of applications,
from daily stock prices to the positions of genes in a chromosome.
13-2
Markov Chains
Markov Chains
A Markov model is given visual representation with a state diagram, such
as the one below.
State Diagram for a Markov Model
The rectangles in the diagram represent the possible states of the process you
are trying to model, and the arrows represent transitions between states.
The label on each arrow represents the probability of that transition. At
each step of the process, the model may generate an output, or emission,
depending on which state it is in, and then make a transition to another
state. An important characteristic of Markov models is that the next state
depends only on the current state, and not on the history of transitions that
lead to the current state.
For example, for a sequence of coin tosses the two states are heads and tails.
The most recent coin toss determines the current state of the model and each
subsequent toss determines the transition to the next state. If the coin is fair,
the transition probabilities are all 1/2. The emission might simply be the
current state. In more complicated models, random processes at each state
will generate emissions. You could, for example, roll a die to determine the
emission at any step.
13-3
13 Markov Models
Markov chains are mathematical descriptions of Markov models with a
discrete set of states. Markov chains are characterized by:
• A set of states {1, 2, ..., M}
• An M-by-M transition matrix T whose i,j entry is the probability of a
transition from state i to state j. The sum of the entries in each row of
T must be 1, because this is the sum of the probabilities of making a
transition from a given state to each of the other states.
• A set of possible outputs, or emissions, {s1, s2, ... , sN}. By default, the set of
emissions is {1, 2, ... , N}, where N is the number of possible emissions, but
you can choose a different set of numbers or symbols.
• An M-by-N emission matrix E whose i,k entry gives the probability of
emitting symbol sk given that the model is in state i.
Markov chains begin in an initial state i0 at step 0. The chain then transitions
to state i1 with probability T1i1 , and emits an output sk1 with probability
Ei1k1 . Consequently, the probability of observing the sequence of states
i1i2 ...ir and the sequence of emissions sk1 sk2 ...skr in the first r steps, is
T1i1 Ei1k1 Ti1i2 Ei2 k2 ...Tir −1ir Eir k
13-4
Hidden Markov Models
Hidden Markov Models
In this section...
“Introduction” on page 13-5
“Analyzing Hidden Markov Models” on page 13-7
Introduction
A hidden Markov model is one in which you observe a sequence of emissions,
but do not know the sequence of states the model went through to generate
the emissions. Analyses of hidden Markov models seek to recover the
sequence of states from the observed data.
As an example, consider a Markov model with two states and six possible
emissions. The model uses:
• A red die, having six sides, labeled 1 through 6.
• A green die, having twelve sides, five of which are labeled 2 through 6,
while the remaining seven sides are labeled 1.
• A weighted red coin, for which the probability of heads is .9 and the
probability of tails is .1.
• A weighted green coin, for which the probability of heads is .95 and the
probability of tails is .05.
The model creates a sequence of numbers from the set {1, 2, 3, 4, 5, 6} with the
following rules:
• Begin by rolling the red die and writing down the number that comes up,
which is the emission.
• Toss the red coin and do one of the following:
- If the result is heads, roll the red die and write down the result.
- If the result is tails, roll the green die and write down the result.
• At each subsequent step, you flip the coin that has the same color as the die
you rolled in the previous step. If the coin comes up heads, roll the same die
as in the previous step. If the coin comes up tails, switch to the other die.
13-5
13 Markov Models
The state diagram for this model has two states, red and green, as shown in
the following figure.
You determine the emission from a state by rolling the die with the same color
as the state. You determine the transition to the next state by flipping the
coin with the same color as the state.
The transition matrix is:
⎡ 0 .9 0 .1 ⎤
T=⎢ ⎥
⎣0.05 0.95⎦
The emissions matrix is:
⎡1 1 1 1 1 1⎤
⎢ 6⎥
E=⎢6 6 6 6 6
⎥
⎢7 1 1 1 1 1⎥
⎢⎣ 12 12 12 12 12 12 ⎥⎦
The model is not hidden because you know the sequence of states from the
colors of the coins and dice. Suppose, however, that someone else is generating
13-6
Hidden Markov Models
the emissions without showing you the dice or the coins. All you see is the
sequence of emissions. If you start seeing more 1s than other numbers, you
might suspect that the model is in the green state, but you cannot be sure
because you cannot see the color of the die being rolled.
Hidden Markov models raise the following questions:
• Given a sequence of emissions, what is the most likely state path?
• Given a sequence of emissions, how can you estimate transition and
emission probabilities of the model?
• What is the forward probability that the model generates a given sequence?
• What is the posterior probability that the model is in a particular state at
any point in the sequence?
Analyzing Hidden Markov Models
• “Generating a Test Sequence” on page 13-8
• “Estimating the State Sequence” on page 13-8
• “Estimating Transition and Emission Matrices” on page 13-9
• “Estimating Posterior State Probabilities” on page 13-11
• “Changing the Initial State Distribution” on page 13-12
Statistics Toolbox functions related to hidden Markov models are:
• hmmgenerate — Generates a sequence of states and emissions from a
Markov model
• hmmestimate — Calculates maximum likelihood estimates of transition
and emission probabilities from a sequence of emissions and a known
sequence of states
• hmmtrain — Calculates maximum likelihood estimates of transition and
emission probabilities from a sequence of emissions
• hmmviterbi — Calculates the most probable state path for a hidden
Markov model
13-7
13 Markov Models
• hmmdecode — Calculates the posterior state probabilities of a sequence
of emissions
This section shows how to use these functions to analyze hidden Markov
models.
Generating a Test Sequence
The following commands create the transition and emission matrices for the
model described in the “Introduction” on page 13-5:
TRANS = [.9 .1; .05 .95;];
EMIS = [1/6, 1/6, 1/6, 1/6, 1/6, 1/6;...
7/12, 1/12, 1/12, 1/12, 1/12, 1/12];
To generate a random sequence of states and emissions from the model, use
hmmgenerate:
[seq,states] = hmmgenerate(1000,TRANS,EMIS);
The output seq is the sequence of emissions and the output states is the
sequence of states.
hmmgenerate begins in state 1 at step 0, makes the transition to state i1 at
step 1, and returns i1 as the first entry in states. To change the initial state,
see “Changing the Initial State Distribution” on page 13-12.
Estimating the State Sequence
Given the transition and emission matrices TRANS and EMIS, the function
hmmviterbi uses the Viterbi algorithm to compute the most likely sequence
of states the model would go through to generate a given sequence seq of
emissions:
likelystates = hmmviterbi(seq, TRANS, EMIS);
likelystates is a sequence the same length as seq.
To test the accuracy of hmmviterbi, compute the percentage of the actual
sequence states that agrees with the sequence likelystates.
13-8
Hidden Markov Models
sum(states==likelystates)/1000
ans =
0.8200
In this case, the most likely sequence of states agrees with the random
sequence 82% of the time.
Estimating Transition and Emission Matrices
• “Using hmmestimate” on page 13-9
• “Using hmmtrain” on page 13-10
The functions hmmestimate and hmmtrain estimate the transition and
emission matrices TRANS and EMIS given a sequence seq of emissions.
Using hmmestimate. The function hmmestimate requires that you know
the sequence of states states that the model went through to generate seq.
The following takes the emission and state sequences and returns estimates
of the transition and emission matrices:
[TRANS_EST, EMIS_EST] = hmmestimate(seq, states)
TRANS_EST =
0.8989 0.1011
0.0585 0.9415
EMIS_EST =
0.1721 0.1721 0.1749 0.1612 0.1803 0.1393
0.5836 0.0741 0.0804 0.0789 0.0726 0.1104
You can compare the outputs with the original transition and emission
matrices, TRANS and EMIS:
TRANS
TRANS =
0.9000 0.1000
0.0500 0.9500
EMIS
13-9
13 Markov Models
EMIS =
0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
0.5833 0.0833 0.0833 0.0833 0.0833 0.0833
Using hmmtrain. If you do not know the sequence of states states, but you
have initial guesses for TRANS and EMIS, you can still estimate TRANS and
EMIS using hmmtrain.
Suppose you have the following initial guesses for TRANS and EMIS.
TRANS_GUESS = [.85 .15; .1 .9];
EMIS_GUESS = [.17 .16 .17 .16 .17 .17;.6 .08 .08 .08 .08 08];
You estimate TRANS and EMIS as follows:
[TRANS_EST2, EMIS_EST2] = hmmtrain(seq, TRANS_GUESS, EMIS_GUESS)
TRANS_EST2 =
0.2286 0.7714
0.0032 0.9968
EMIS_EST2 =
0.1436 0.2348 0.1837 0.1963 0.2350 0.0066
0.4355 0.1089 0.1144 0.1082 0.1109 0.1220
hmmtrain uses an iterative algorithm that alters the matrices TRANS_GUESS
and EMIS_GUESS so that at each step the adjusted matrices are more likely to
generate the observed sequence, seq. The algorithm halts when the matrices
in two successive iterations are within a small tolerance of each other.
If the algorithm fails to reach this tolerance within a maximum number of
iterations, whose default value is 100, the algorithm halts. In this case,
hmmtrain returns the last values of TRANS_EST and EMIS_EST and issues a
warning that the tolerance was not reached.
If the algorithm fails to reach the desired tolerance, increase the default value
of the maximum number of iterations with the command:
hmmtrain(seq,TRANS_GUESS,EMIS_GUESS,'maxiterations',maxiter)
where maxiter is the maximum number of steps the algorithm executes.
13-10
Hidden Markov Models
Change the default value of the tolerance with the command:
hmmtrain(seq, TRANS_GUESS, EMIS_GUESS, 'tolerance', tol)
where tol is the desired value of the tolerance. Increasing the value of tol
makes the algorithm halt sooner, but the results are less accurate.
Two factors reduce the reliability of the output matrices of hmmtrain:
• The algorithm converges to a local maximum that does not represent the
true transition and emission matrices. If you suspect this, use different
initial guesses for the matrices TRANS_EST and EMIS_EST.
• The sequence seq may be too short to properly train the matrices. If you
suspect this, use a longer sequence for seq.
Estimating Posterior State Probabilities
The posterior state probabilities of an emission sequence seq are the
conditional probabilities that the model is in a particular state when it
generates a symbol in seq, given that seq is emitted. You compute the
posterior state probabilities with hmmdecode:
PSTATES = hmmdecode(seq,TRANS,EMIS)
The output PSTATES is an M-by-L matrix, where M is the number of states
and L is the length of seq. PSTATES(i,j) is the conditional probability that
the model is in state i when it generates the jth symbol of seq, given that
seq is emitted.
hmmdecode begins with the model in state 1 at step 0, prior to the first
emission. PSTATES(i,1) is the probability that the model is in state i at the
following step 1. To change the initial state, see “Changing the Initial State
Distribution” on page 13-12.
To return the logarithm of the probability of the sequence seq, use the second
output argument of hmmdecode:
[PSTATES,logpseq] = hmmdecode(seq,TRANS,EMIS)
The probability of a sequence tends to 0 as the length of the sequence
increases, and the probability of a sufficiently long sequence becomes less
13-11
13 Markov Models
than the smallest positive number your computer can represent. hmmdecode
returns the logarithm of the probability to avoid this problem.
Changing the Initial State Distribution
By default, Statistics Toolbox hidden Markov model functions begin in state 1.
In other words, the distribution of initial states has all of its probability mass
concentrated at state 1. To assign a different distribution of probabilities, p =
[p1, p2, ..., pM], to the M initial states, do the following:
1 Create an M+1-by-M+1 augmented transition matrix, T̂ of the following
form:
⎡0 p ⎤
T̂ = ⎢ ⎥
⎣0 T ⎦
where T is the true transition matrix. The first column of T̂ contains M+1
zeros. p must sum to 1.
2 Create an M+1-by-N augmented emission matrix, Ê , that has the
following form:
⎡0⎤
Ê = ⎢ ⎥
⎣ E⎦
If the transition and emission matrices are TRANS and EMIS, respectively, you
create the augmented matrices with the following commands:
TRANS_HAT = [0 p; zeros(size(TRANS,1),1) TRANS];
EMIS_HAT = [zeros(1,size(EMIS,2)); EMIS];
13-12
14
Design of Experiments
• “Introduction” on page 14-2
• “Full Factorial Designs” on page 14-3
• “Fractional Factorial Designs” on page 14-5
• “Response Surface Designs” on page 14-9
• “D-Optimal Designs” on page 14-15
14 Design of Experiments
Introduction
Passive data collection leads to a number of problems in statistical modeling.
Observed changes in a response variable may be correlated with, but
not caused by, observed changes in individual factors (process variables).
Simultaneous changes in multiple factors may produce interactions that are
difficult to separate into individual effects. Observations may be dependent,
while a model of the data considers them to be independent.
Designed experiments address these problems. In a designed experiment,
the data-producing process is actively manipulated to improve the quality
of information and to eliminate redundant data. A common goal of all
experimental designs is to collect data as parsimoniously as possible while
providing sufficient information to accurately estimate model parameters.
For example, a simple model of a response y in an experiment with two
controlled factors x1 and x2 might look like this:
y = 0 + 1 x1 + 2 x2 + 3 x1 x2 +
Here ε includes both experimental error and the effects of any uncontrolled
factors in the experiment. The terms β1x1 and β2x2 are main effects and the
term β3x1x2 is a two-way interaction effect. A designed experiment would
systematically manipulate x1 and x2 while measuring y, with the objective of
accurately estimating β0, β1, β2, and β3.
14-2
Full Factorial Designs
Full Factorial Designs
In this section...
“Multilevel Designs” on page 14-3
“Two-Level Designs” on page 14-4
Multilevel Designs
To systematically vary experimental factors, assign each factor a discrete
set of levels. Full factorial designs measure response variables using every
treatment (combination of the factor levels). A full factorial design for n
factors with N1, ..., Nn levels requires N1 × ... × Nn experimental runs—one for
each treatment. While advantageous for separating individual effects, full
factorial designs can make large demands on data collection.
As an example, suppose a machine shop has three machines and four
operators. If the same operator always uses the same machine, it is
impossible to determine if a machine or an operator is the cause of variation
in production. By allowing every operator to use every machine, effects are
separated. A full factorial list of treatments is generated by the Statistics
Toolbox function fullfact:
dFF = fullfact([3,4])
dFF =
1 1
2 1
3 1
1 2
2 2
3 2
1 3
2 3
3 3
1 4
2 4
3 4
Each of the 3×4 = 12 rows of dFF represent one machine/operator combination.
14-3
14 Design of Experiments
Two-Level Designs
Many experiments can be conducted with two-level factors, using two-level
designs. For example, suppose the machine shop in the previous example
always keeps the same operator on the same machine, but wants to measure
production effects that depend on the composition of the day and night
shifts. The Statistics Toolbox function ff2n generates a full factorial list of
treatments:
dFF2 = ff2n(4)
dFF2 =
0 0 0 0
0 0 0 1
0 0 1 0
0 0 1 1
0 1 0 0
0 1 0 1
0 1 1 0
0 1 1 1
1 0 0 0
1 0 0 1
1 0 1 0
1 0 1 1
1 1 0 0
1 1 0 1
1 1 1 0
1 1 1 1
Each of the 24 = 16 rows of dFF2 represent one schedule of operators for the
day (0) and night (1) shifts.
14-4
Fractional Factorial Designs
Fractional Factorial Designs
In this section...
“Introduction” on page 14-5
“Plackett-Burman Designs” on page 14-5
“General Fractional Designs” on page 14-6
Introduction
Two-level designs are sufficient for evaluating many production processes.
Factor levels of ±1 can indicate categorical factors, normalized factor extremes,
or simply “up” and “down” from current factor settings. Experimenters
evaluating process changes are interested primarily in the factor directions
that lead to process improvement.
For experiments with many factors, two-level full factorial designs can lead to
large amounts of data. For example, a two-level full factorial design with 10
factors requires 210 = 1024 runs. Often, however, individual factors or their
interactions have no distinguishable effects on a response. This is especially
true of higher order interactions. As a result, a well-designed experiment can
use fewer runs for estimating model parameters.
Fractional factorial designs use a fraction of the runs required by full
factorial designs. A subset of experimental treatments is selected based on
an evaluation (or assumption) of which factors and interactions have the
most significant effects. Once this selection is made, the experimental design
must separate these effects. In particular, significant effects should not
be confounded, that is, the measurement of one should not depend on the
measurement of another.
Plackett-Burman Designs
Plackett-Burman designs are used when only main effects are considered
significant. Two-level Plackett-Burman designs require a number of
experimental runs that are a multiple of 4 rather than a power of 2. The
MATLAB function hadamard generates these designs:
dPB = hadamard(8)
14-5
14 Design of Experiments
dPB =
1 1 1 1 1 1 1 1
1 -1 1 -1 1 -1 1 -1
1 1 -1 -1 1 1 -1 -1
1 -1 -1 1 1 -1 -1 1
1 1 1 1 -1 -1 -1 -1
1 -1 1 -1 -1 1 -1 1
1 1 -1 -1 -1 -1 1 1
1 -1 -1 1 -1 1 1 -1
Binary factor levels are indicated by ±1. The design is for eight runs (the rows
of dPB) manipulating seven two-level factors (the last seven columns of dPB).
The number of runs is a fraction 8/27 = 0.0625 of the runs required by a full
factorial design. Economy is achieved at the expense of confounding main
effects with any two-way interactions.
General Fractional Designs
At the cost of a larger fractional design, you can specify which interactions
you wish to consider significant. A design of resolution R is one in which no
n-factor interaction is confounded with any other effect containing less than
R – n factors. Thus, a resolution III design does not confound main effects
with one another but may confound them with two-way interactions (as in
“Plackett-Burman Designs” on page 14-5), while a resolution IV design does
not confound either main effects or two-way interactions but may confound
two-way interactions with each other.
Specify general fractional factorial designs using a full factorial design for
a selected subset of basic factors and generators for the remaining factors.
Generators are products of the basic factors, giving the levels for the
remaining factors. Use the Statistics Toolbox function fracfact to generate
these designs:
dfF = fracfact('a b c d bcd acd')
dfF =
-1 -1 -1 -1 -1 -1
-1 -1 -1 1 1 1
-1 -1 1 -1 1 1
-1 -1 1 1 -1 -1
-1 1 -1 -1 1 -1
14-6
Fractional Factorial Designs
-1 1 -1 1 -1 1
-1 1 1 -1 -1 1
-1 1 1 1 1 -1
1 -1 -1 -1 -1 1
1 -1 -1 1 1 -1
1 -1 1 -1 1 -1
1 -1 1 1 -1 1
1 1 -1 -1 1 1
1 1 -1 1 -1 -1
1 1 1 -1 -1 -1
1 1 1 1 1 1
This is a six-factor design in which four two-level basic factors (a, b, c, and
d in the first four columns of dfF) are measured in every combination of
levels, while the two remaining factors (in the last three columns of dfF) are
measured only at levels defined by the generators bcd and acd, respectively.
Levels in the generated columns are products of corresponding levels in the
columns that make up the generator.
The challenge of creating a fractional factorial design is to choose basic factors
and generators so that the design achieves a specified resolution in a specified
number of runs. Use the Statistics Toolbox function fracfactgen to find
appropriate generators:
generators = fracfactgen('a b c d e f',4,4)
generators =
'a'
'b'
'c'
'd'
'bcd'
'acd'
These are generators for a six-factor design with factors a through f, using 24
= 16 runs to achieve resolution IV. The fracfactgen function uses an efficient
search algorithm to find generators that meet the requirements.
An optional output from fracfact displays the confounding pattern of the
design:
[dfF,confounding] = fracfact(generators);
14-7
14 Design of Experiments
confounding
confounding =
'Term' 'Generator' 'Confounding'
'X1' 'a' 'X1'
'X2' 'b' 'X2'
'X3' 'c' 'X3'
'X4' 'd' 'X4'
'X5' 'bcd' 'X5'
'X6' 'acd' 'X6'
'X1*X2' 'ab' 'X1*X2 + X5*X6'
'X1*X3' 'ac' 'X1*X3 + X4*X6'
'X1*X4' 'ad' 'X1*X4 + X3*X6'
'X1*X5' 'abcd' 'X1*X5 + X2*X6'
'X1*X6' 'cd' 'X1*X6 + X2*X5 + X3*X4'
'X2*X3' 'bc' 'X2*X3 + X4*X5'
'X2*X4' 'bd' 'X2*X4 + X3*X5'
'X2*X5' 'cd' 'X1*X6 + X2*X5 + X3*X4'
'X2*X6' 'abcd' 'X1*X5 + X2*X6'
'X3*X4' 'cd' 'X1*X6 + X2*X5 + X3*X4'
'X3*X5' 'bd' 'X2*X4 + X3*X5'
'X3*X6' 'ad' 'X1*X4 + X3*X6'
'X4*X5' 'bc' 'X2*X3 + X4*X5'
'X4*X6' 'ac' 'X1*X3 + X4*X6'
'X5*X6' 'ab' 'X1*X2 + X5*X6'
The confounding pattern shows that main effects are effectively separated
by the design, but two-way interactions are confounded with various other
two-way interactions.
14-8
Response Surface Designs
Response Surface Designs
In this section...
“Introduction” on page 14-9
“Central Composite Designs” on page 14-9
“Box-Behnken Designs” on page 14-13
Introduction
As discussed in “Response Surface Models” on page 9-45, quadratic response
surfaces are simple models that provide a maximum or minimum without
making additional assumptions about the form of the response. Quadratic
models can be calibrated using full factorial designs with three or more levels
for each factor, but these designs generally require more runs than necessary
to accurately estimate model parameters. This section discusses designs for
calibrating quadratic models that are much more efficient, using three or five
levels for each factor, but not using all combinations of levels.
Central Composite Designs
Central composite designs (CCDs), also known as Box-Wilson designs, are
appropriate for calibrating the full quadratic models described in “Response
Surface Models” on page 9-45. There are three types of CCDs—circumscribed,
inscribed, and faced—pictured below:
14-9
14 Design of Experiments
14-10
Response Surface Designs
Each design consists of a factorial design (the corners of a cube) together with
center and star points that allow for estimation of second-order effects. For
a full quadratic model with n factors, CCDs have enough design points to
estimate the (n+2)(n+1)/2 coefficients in a full quadratic model with n factors.
The type of CCD used (the position of the factorial and star points) is
determined by the number of factors and by the desired properties of the
design. The following table summarizes some important properties. A design
is rotatable if the prediction variance depends only on the distance of the
design point from the center of the design.
14-11
14 Design of Experiments
Design Rotatable Factor Uses Points Accuracy of
Levels Outside ±1 Estimates
Circumscribed Yes 5 Yes Good over entire
(CCC) design space
Inscribed Yes 5 No Good over central
(CCI) subset of design space
Faced (CCF) No 3 No Fair over entire
design space; poor
for pure quadratic
coefficients
Generate CCDs with the Statistics Toolbox function ccdesign:
dCC = ccdesign(3,'type','circumscribed')
dCC =
-1.0000 -1.0000 -1.0000
-1.0000 -1.0000 1.0000
-1.0000 1.0000 -1.0000
-1.0000 1.0000 1.0000
1.0000 -1.0000 -1.0000
1.0000 -1.0000 1.0000
1.0000 1.0000 -1.0000
1.0000 1.0000 1.0000
-1.6818 0 0
1.6818 0 0
0 -1.6818 0
0 1.6818 0
0 0 -1.6818
0 0 1.6818
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
14-12
Response Surface Designs
0 0 0
The repeated center point runs allow for a more uniform estimate of the
prediction variance over the entire design space.
Box-Behnken Designs
Like the designs described in “Central Composite Designs” on page
14-9, Box-Behnken designs are used to calibrate full quadratic models.
Box-Behnken designs are rotatable and, for a small number of factors (four or
less), require fewer runs than CCDs. By avoiding the corners of the design
space, they allow experimenters to work around extreme factor combinations.
Like an inscribed CCD, however, extremes are then poorly estimated.
The geometry of a Box-Behnken design is pictured in the following figure.
Design points are at the midpoints of edges of the design space and at the
center, and do not contain an embedded factorial design.
14-13
14 Design of Experiments
Generate Box-Behnken designs with the Statistics Toolbox function bbdesign:
dBB = bbdesign(3)
dBB =
-1 -1 0
-1 1 0
1 -1 0
1 1 0
-1 0 -1
-1 0 1
1 0 -1
1 0 1
0 -1 -1
0 -1 1
0 1 -1
0 1 1
0 0 0
0 0 0
0 0 0
Again, the repeated center point runs allow for a more uniform estimate of
the prediction variance over the entire design space.
14-14
D-Optimal Designs
D-Optimal Designs
In this section...
“Introduction” on page 14-15
“Generating D-Optimal Designs” on page 14-16
“Augmenting D-Optimal Designs” on page 14-19
“Specifying Fixed Covariate Factors” on page 14-20
“Specifying Categorical Factors” on page 14-21
“Specifying Candidate Sets” on page 14-21
Introduction
Traditional experimental designs (“Full Factorial Designs” on page 14-3,
“Fractional Factorial Designs” on page 14-5, and “Response Surface Designs”
on page 14-9) are appropriate for calibrating linear models in experimental
settings where factors are relatively unconstrained in the region of interest.
In some cases, however, models are necessarily nonlinear. In other cases,
certain treatments (combinations of factor levels) may be expensive or
infeasible to measure. D-optimal designs are model-specific designs that
address these limitations of traditional designs.
A D-optimal design is generated by an iterative search algorithm and seeks
to minimize the covariance of the parameter estimates for a specified model.
This is equivalent to maximizing the determinant D = |XTX|, where X is the
design matrix of model terms (the columns) evaluated at specific treatments
in the design space (the rows). Unlike traditional designs, D-optimal designs
do not require orthogonal design matrices, and as a result, parameter
estimates may be correlated. Parameter estimates may also be locally, but
not globally, D-optimal.
There are several Statistics Toolbox functions for generating D-optimal
designs:
14-15
14 Design of Experiments
Function Description
candexch Uses a row-exchange algorithm to generate a D-optimal design
with a specified number of runs for a specified model and a
specified candidate set. This is the second component of the
algorithm used by rowexch.
candgen Generates a candidate set for a specified model. This is the
first component of the algorithm used by rowexch.
cordexch Uses a coordinate-exchange algorithm to generate a D-optimal
design with a specified number of runs for a specified model.
daugment Uses a coordinate-exchange algorithm to augment an existing
D-optimal design with additional runs to estimate additional
model terms.
dcovary Uses a coordinate-exchange algorithm to generate a D-optimal
design with fixed covariate factors.
rowexch Uses a row-exchange algorithm to generate a D-optimal design
with a specified number of runs for a specified model. The
algorithm calls candgen and then candexch. (Call candexch
separately to specify a candidate set.)
The following sections explain how to use these functions to generate
D-optimal designs.
Note The Statistics Toolbox function rsmdemo generates simulated data for
experimental settings specified by either the user or by a D-optimal design
generated by cordexch. It uses the rstool interface to visualize response
surface models fit to the data, and it uses the nlintool interface to visualize
a nonlinear model fit to the data.
Generating D-Optimal Designs
Two Statistics Toolbox algorithms generate D-optimal designs:
• The cordexch function uses a coordinate-exchange algorithm
• The rowexch function uses a row-exchange algorithm
14-16
D-Optimal Designs
Both cordexch and rowexch use iterative search algorithms. They operate by
incrementally changing an initial design matrix X to increase D = |XTX| at
each step. In both algorithms, there is randomness built into the selection of
the initial design and into the choice of the incremental changes. As a result,
both algorithms may return locally, but not globally, D-optimal designs. Run
each algorithm multiple times and select the best result for your final design.
Both functions have a 'tries' parameter that automates this repetition
and comparison.
At each step, the row-exchange algorithm exchanges an entire row of X with a
row from a design matrix C evaluated at a candidate set of feasible treatments.
The rowexch function automatically generates a C appropriate for a specified
model, operating in two steps by calling the candgen and candexch functions
in sequence. Provide your own C by calling candexch directly. In either case,
if C is large, its static presence in memory can affect computation.
The coordinate-exchange algorithm, by contrast, does not use a candidate
set. (Or rather, the candidate set is the entire design space.) At each step,
the coordinate-exchange algorithm exchanges a single element of X with a
new element evaluated at a neighboring point in design space. The absence
of a candidate set reduces demands on memory, but the smaller scale of the
search means that the coordinate-exchange algorithm is more likely to become
trapped in a local minimum than the row-exchange algorithm.
For example, suppose you want a design to estimate the parameters in the
following three-factor, seven-term interaction model:
y = 0 + 1 x 1 + 2 x 2 + 3 x 3 + 12 x 1 x 2 + 13 x 1 x 3 + 23 x 2 x 3 +
Use cordexch to generate a D-optimal design with seven runs:
nfactors = 3;
nruns = 7;
[dCE,X] = cordexch(nfactors,nruns,'interaction','tries',10)
dCE =
-1 1 1
-1 -1 -1
1 1 1
-1 1 -1
1 -1 1
14-17
14 Design of Experiments
1 -1 -1
-1 -1 1
X =
1 -1 1 1 -1 -1 1
1 -1 -1 -1 1 1 1
1 1 1 1 1 1 1
1 -1 1 -1 -1 1 -1
1 1 -1 1 -1 1 -1
1 1 -1 -1 -1 -1 1
1 -1 -1 1 1 -1 -1
Columns of the design matrix X are the model terms evaluated at each row of
the design dCE. The terms appear in order from left to right:
1 Constant term
2 Linear terms (1, 2, 3)
3 Interaction terms (12, 13, 23)
Use X to fit the model, as described in “Linear Regression” on page 9-3, to
response data measured at the design points in dCE.
Use rowexch in a similar fashion to generate an equivalent design:
[dRE,X] = rowexch(nfactors,nruns,'interaction','tries',10)
dRE =
-1 -1 1
1 -1 1
1 -1 -1
1 1 1
-1 -1 -1
-1 1 -1
-1 1 1
X =
1 -1 -1 1 1 -1 -1
1 1 -1 1 -1 1 -1
1 1 -1 -1 -1 -1 1
1 1 1 1 1 1 1
1 -1 -1 -1 1 1 1
1 -1 1 -1 -1 1 -1
14-18
D-Optimal Designs
1 -1 1 1 -1 -1 1
Augmenting D-Optimal Designs
In practice, you may want to add runs to a completed experiment to learn
more about a process and estimate additional model coefficients. The
daugment function uses a coordinate-exchange algorithm to augment an
existing D-optimal design.
For example, the following eight-run design is adequate for estimating main
effects in a four-factor model:
dCEmain = cordexch(4,8)
dCEmain =
1 -1 -1 1
-1 -1 1 1
-1 1 -1 1
1 1 1 -1
1 1 1 1
-1 1 -1 -1
1 -1 -1 -1
-1 -1 1 -1
To estimate the six interaction terms in the model, augment the design with
eight additional runs:
dCEinteraction = daugment(dCEmain,8,'interaction')
dCEinteraction =
1 -1 -1 1
-1 -1 1 1
-1 1 -1 1
1 1 1 -1
1 1 1 1
-1 1 -1 -1
1 -1 -1 -1
-1 -1 1 -1
-1 1 1 1
-1 -1 -1 -1
1 -1 1 -1
1 1 -1 1
-1 1 1 -1
14-19
14 Design of Experiments
1 1 -1 -1
1 -1 1 1
1 1 1 -1
The augmented design is full factorial, with the original eight runs in the
first eight rows.
The 'start' parameter of the candexch function provides the same
functionality as daugment, but uses a row exchange algorithm rather than a
coordinate-exchange algorithm.
Specifying Fixed Covariate Factors
In many experimental settings, certain factors and their covariates are
constrained to a fixed set of levels or combinations of levels. These cannot be
varied when searching for an optimal design. The dcovary function allows
you to specify fixed covariate factors in the coordinate exchange algorithm.
For example, suppose you want a design to estimate the parameters in a
three-factor linear additive model, with eight runs that necessarily occur at
different times. If the process experiences temporal linear drift, you may
want to include the run time as a variable in the model. Produce the design as
follows:
time = linspace(-1,1,8)';
[dCV,X] = dcovary(3,time,'linear')
dCV =
-1.0000 1.0000 1.0000 -1.0000
1.0000 -1.0000 -1.0000 -0.7143
-1.0000 -1.0000 -1.0000 -0.4286
1.0000 -1.0000 1.0000 -0.1429
1.0000 1.0000 -1.0000 0.1429
-1.0000 1.0000 -1.0000 0.4286
1.0000 1.0000 1.0000 0.7143
-1.0000 -1.0000 1.0000 1.0000
X =
1.0000 -1.0000 1.0000 1.0000 -1.0000
1.0000 1.0000 -1.0000 -1.0000 -0.7143
1.0000 -1.0000 -1.0000 -1.0000 -0.4286
1.0000 1.0000 -1.0000 1.0000 -0.1429
14-20
D-Optimal Designs
1.0000 1.0000 1.0000 -1.0000 0.1429
1.0000 -1.0000 1.0000 -1.0000 0.4286
1.0000 1.0000 1.0000 1.0000 0.7143
1.0000 -1.0000 -1.0000 1.0000 1.0000
The column vector time is a fixed factor, normalized to values between ±1.
The number of rows in the fixed factor specifies the number of runs in the
design. The resulting design dCV gives factor settings for the three controlled
model factors at each time.
Specifying Categorical Factors
Categorical factors take values in a discrete set of levels. Both cordexch and
rowexch have a 'categorical' parameter that allows you to specify the
indices of categorical factors and a 'levels' parameter that allows you to
specify a number of levels for each factor.
For example, the following eight-run design is for a linear additive model with
five factors in which the final factor is categorical with three levels:
dCEcat = cordexch(5,8,'linear','categorical',5,'levels',3)
dCEcat =
-1 -1 1 1 2
-1 -1 -1 -1 3
1 1 1 1 3
1 1 -1 -1 2
1 -1 -1 1 3
-1 1 -1 1 1
-1 1 1 -1 3
1 -1 1 -1 1
Specifying Candidate Sets
The row-exchange algorithm exchanges rows of an initial design matrix X
with rows from a design matrix C evaluated at a candidate set of feasible
treatments. The rowexch function automatically generates a C appropriate for
a specified model, operating in two steps by calling the candgen and candexch
functions in sequence. Provide your own C by calling candexch directly.
14-21
14 Design of Experiments
For example, the following uses rowexch to generate a five-run design for
a two-factor pure quadratic model using a candidate set that is produced
internally:
dRE1 = rowexch(2,5,'purequadratic','tries',10)
dRE1 =
-1 1
0 0
1 -1
1 0
1 1
The same thing can be done using candgen and candexch in sequence:
[dC,C] = candgen(2,'purequadratic') % Candidate set, C
dC =
-1 -1
0 -1
1 -1
-1 0
0 0
1 0
-1 1
0 1
1 1
C =
1 -1 -1 1 1
1 0 -1 0 1
1 1 -1 1 1
1 -1 0 1 0
1 0 0 0 0
1 1 0 1 0
1 -1 1 1 1
1 0 1 0 1
1 1 1 1 1
treatments = candexch(C,5,'tries',10) % D-opt subset
treatments =
2
1
7
3
14-22
D-Optimal Designs
4
dRE2 = dC(treatments,:) % Display design
dRE2 =
0 -1
-1 -1
-1 1
1 -1
-1 0
You can replace C in this example with a design matrix evaluated at your own
candidate set. For example, suppose your experiment is constrained so that
the two factors cannot have extreme settings simultaneously. The following
produces a restricted candidate set:
constraint = sum(abs(dC),2) < 2; % Feasible treatments
my_dC = dC(constraint,:)
my_dC =
0 -1
-1 0
0 0
1 0
0 1
Use the x2fx function to convert the candidate set to a design matrix:
my_C = x2fx(my_dC,'purequadratic')
my_C =
1 0 -1 0 1
1 -1 0 1 0
1 0 0 0 0
1 1 0 1 0
1 0 1 0 1
Find the required design in the same manner:
my_treatments = candexch(my_C,5,'tries',10) % D-opt subset
my_treatments =
2
4
5
1
14-23
14 Design of Experiments
3
my_dRE = my_dC(my_treatments,:) % Display design
my_dRE =
-1 0
1 0
0 1
0 -1
0 0
14-24
15
Statistical Process Control
• “Introduction” on page 15-2
• “Control Charts” on page 15-3
• “Capability Studies” on page 15-6
15 Statistical Process Control
Introduction
Statistical process control (SPC) refers to a number of different methods for
monitoring and assessing the quality of manufactured goods. Combined
with methods from the Chapter 14, “Design of Experiments”, SPC is used in
programs that define, measure, analyze, improve, and control development
and production processes. These programs are often implemented using
“Design for Six Sigma” methodologies.
15-2
Control Charts
Control Charts
A control chart displays measurements of process samples over time. The
measurements are plotted together with user-defined specification limits and
process-defined control limits. The process can then be compared with its
specifications—to see if it is in control or out of control.
The chart is just a monitoring tool. Control activity might occur if the chart
indicates an undesirable, systematic change in the process. The control
chart is used to discover the variation, so that the process can be adjusted
to reduce it.
Control charts are created with the controlchart function. Any of the
following chart types may be specified:
• Xbar or mean
• Standard deviation
• Range
• Exponentially weighted moving average
• Individual observation
• Moving range of individual observations
• Moving average of individual observations
• Proportion defective
• Number of defectives
• Defects per unit
• Count of defects
Control rules are specified with the controlrules function.
For example, the following commands create an xbar chart, using the
“Western Electric 2” rule (2 of 3 points at least 2 standard errors above the
center line) to mark out of control measurements:
load parts;
st = controlchart(runout,'rules','we2');
15-3
15 Statistical Process Control
x = st.mean;
cl = st.mu;
se = st.sigma./sqrt(st.n);
hold on
plot(cl+2*se,'m')
Measurements that violate the control rule can then be identified:
R = controlrules('we2',x,cl,se);
I = find(R)
15-4
Control Charts
I =
21
23
24
25
26
27
15-5
15 Statistical Process Control
Capability Studies
Before going into production, many manufacturers run a capability study to
determine if their process will run within specifications enough of the time.
Capability indices produced by such a study are used to estimate expected
percentages of defective parts.
Capability studies are conducted with the capability function. The following
capability indices are produced:
• mu — Sample mean
• sigma — Sample standard deviation
• P — Estimated probability of being within the lower (L) and upper (U)
specification limits
• Pl — Estimated probability of being below L
• Pu — Estimated probability of being above U
• Cp — (U-L)/(6*sigma)
• Cpl — (mu-L)./(3.*sigma)
• Cpu — (U-mu)./(3.*sigma)
• Cpk — min(Cpl,Cpu)
As an example, simulate a sample from a process with a mean of 3 and a
standard deviation of 0.005:
data = normrnd(3,0.005,100,1);
Compute capability indices if the process has an upper specification limit of
3.01 and a lower specification limit of 2.99:
S = capability(data,[2.99 3.01])
S =
mu: 3.0006
sigma: 0.0047
P: 0.9669
Pl: 0.0116
Pu: 0.0215
15-6
Capability Studies
Cp: 0.7156
Cpl: 0.7567
Cpu: 0.6744
Cpk: 0.6744
Visualize the specification and process widths:
capaplot(data,[2.99 3.01]);
grid on
15-7
15 Statistical Process Control
15-8
16
Function Reference
File I/O (p. 16-2) Data file input/output
Data Organization (p. 16-3) Data arrays and groups
Descriptive Statistics (p. 16-8) Data summaries
Statistical Visualization (p. 16-11) Data patterns and trends
Probability Distributions (p. 16-15) Modeling data frequency
Hypothesis Tests (p. 16-31) Inferences from data
Analysis of Variance (p. 16-32) Modeling data variance
Regression Analysis (p. 16-33) Continuous data models
Multivariate Methods (p. 16-38) Visualization and reduction
Cluster Analysis (p. 16-40) Identifying data categories
Model Assessment (p. 16-41) Identifying data categories
Classification (p. 16-42) Categorical data models
Hidden Markov Models (p. 16-46) Stochastic data models
Design of Experiments (p. 16-47) Systematic data collection
Statistical Process Control (p. 16-51) Production monitoring
GUIs (p. 16-52) Interactive tools
Utilities (p. 16-53) General purpose
16 Function Reference
File I/O
caseread Read case names from file
casewrite Write case names to file
tblread Read tabular data from file
tblwrite Write tabular data to file
tdfread Read tab-delimited file
xptread Create dataset array from data
stored in SAS XPORT format file
16-2
Data Organization
Data Organization
Categorical Arrays (p. 16-3)
Dataset Arrays (p. 16-6)
Grouped Data (p. 16-7)
Categorical Arrays
addlevels (categorical) Add levels to categorical array
cat (categorical) Concatenate categorical arrays
categorical Create categorical array
cellstr (categorical) Convert categorical array to cell
array of strings
char (categorical) Convert categorical array to
character array
circshift (categorical) Shift categorical array circularly
ctranspose (categorical) Transpose categorical matrix
double (categorical) Convert categorical array to double
array
droplevels (categorical) Drop levels
end (categorical) Last index in indexing expression for
categorical array
flipdim (categorical) Flip categorical array along specified
dimension
fliplr (categorical) Flip categorical matrix in left/right
direction
flipud (categorical) Flip categorical matrix in up/down
direction
getlabels (categorical) Access categorical array labels
getlevels (categorical) Get categorical array levels
16-3
16 Function Reference
hist (categorical) Plot histogram of categorical data
horzcat (categorical) Horizontal concatenation for
categorical arrays
int16 (categorical) Convert categorical array to signed
16-bit integer array
int32 (categorical) Convert categorical array to signed
32-bit integer array
int64 (categorical) Convert categorical array to signed
64-bit integer array
int8 (categorical) Convert categorical array to signed
8-bit integer array
intersect (categorical) Set intersection for categorical
arrays
ipermute (categorical) Inverse permute dimensions of
categorical array
isempty (categorical) True for empty categorical array
isequal (categorical) True if categorical arrays are equal
islevel (categorical) Test for levels
ismember (categorical) True for elements of categorical
array in set
ismember (ordinal) Test for membership
isscalar (categorical) True if categorical array is scalar
isundefined (categorical) Test for undefined elements
isvector (categorical) True if categorical array is vector
length (categorical) Length of categorical array
levelcounts (categorical) Element counts by level
mergelevels (ordinal) Merge levels
ndims (categorical) Number of dimensions of categorical
array
nominal Construct nominal categorical array
16-4
Data Organization
numel (categorical) Number of elements in categorical
array
ordinal Construct ordinal categorical array
permute (categorical) Permute dimensions of categorical
array
reorderlevels (categorical) Reorder levels
repmat (categorical) Replicate and tile categorical array
reshape (categorical) Resize categorical array
rot90 (categorical) Rotate categorical matrix 90 degrees
setdiff (categorical) Set difference for categorical arrays
setlabels (categorical) Label levels
setxor (categorical) Set exclusive-or for categorical
arrays
shiftdim (categorical) Shift dimensions of categorical array
single (categorical) Convert categorical array to single
array
size (categorical) Size of categorical array
sort (ordinal) Sort elements of ordinal array
sortrows (ordinal) Sort rows
squeeze (categorical) Squeeze singleton dimensions from
categorical array
summary (categorical) Summary statistics for categorical
array
times (categorical) Product of categorical arrays
transpose (categorical) Transpose categorical matrix
uint16 (categorical) Convert categorical array to
unsigned 16-bit integers
uint32 (categorical) Convert categorical array to
unsigned 32-bit integers
16-5
16 Function Reference
uint64 (categorical) Convert categorical array to
unsigned 64-bit integers
uint8 (categorical) Convert categorical array to
unsigned 8-bit integers
union (categorical) Set union for categorical arrays
unique (categorical) Unique values in categorical array
vertcat (categorical) Vertical concatenation for categorical
arrays
Dataset Arrays
cat (dataset) Concatenate dataset arrays
cellstr (dataset) Create cell array of strings from
dataset array
dataset Construct dataset array
datasetfun (dataset) Apply function to dataset array
variables
double (dataset) Convert dataset variables to double
array
end (dataset) Last index in indexing expression for
dataset array
export (dataset) Write dataset array to file
get (dataset) Access dataset array properties
grpstats (dataset) Summary statistics by group for
dataset arrays
horzcat (dataset) Horizontal concatenation for dataset
arrays
isempty (dataset) True for empty dataset array
join (dataset) Merge observations
length (dataset) Length of dataset array
16-6
Data Organization
ndims (dataset) Number of dimensions of dataset
array
numel (dataset) Number of elements in dataset array
replacedata (dataset) Replace dataset variables
set (dataset) Set and display properties
single (dataset) Convert dataset variables to single
array
size (dataset) Size of dataset array
sortrows (dataset) Sort rows of dataset array
stack (dataset) Stack data from multiple variables
into single variable
summary (dataset) Print summary of dataset array
unique (dataset) Unique observations in dataset
array
unstack (dataset) Unstack data from single variable
into multiple variables
vertcat (dataset) Vertical concatenation for dataset
arrays
Grouped Data
gplotmatrix Matrix of scatter plots by group
grp2idx Create index vector from grouping
variable
grpstats Summary statistics by group
gscatter Scatter plot by group
16-7
16 Function Reference
Descriptive Statistics
Summaries (p. 16-8)
Measures of Central Tendency
(p. 16-8)
Measures of Dispersion (p. 16-8)
Measures of Shape (p. 16-9)
Statistics Resampling (p. 16-9)
Data with Missing Values (p. 16-9)
Data Correlation (p. 16-10)
Summaries
crosstab Cross-tabulation
grpstats Summary statistics by group
summary (categorical) Summary statistics for categorical
array
tabulate Frequency table
Measures of Central Tendency
geomean Geometric mean
harmmean Harmonic mean
trimmean Mean excluding outliers
Measures of Dispersion
iqr Interquartile range
mad Mean or median absolute deviation
16-8
Descriptive Statistics
moment Central moments
range Range of values
Measures of Shape
kurtosis Kurtosis
moment Central moments
prctile Calculate percentile values
quantile Quantiles
skewness Skewness
zscore Standardized z-scores
Statistics Resampling
bootci Bootstrap confidence interval
bootstrp Bootstrap sampling
jackknife Jackknife sampling
Data with Missing Values
nancov Covariance ignoring NaN values
nanmax Maximum ignoring NaN values
nanmean Mean ignoring NaN values
nanmedian Median ignoring NaN values
nanmin Minimum ignoring NaN values
nanstd Standard deviation ignoring NaN
values
nansum Sum ignoring NaN values
nanvar Variance, ignoring NaN values
16-9
16 Function Reference
Data Correlation
canoncorr Canonical correlation
cholcov Cholesky-like covariance
decomposition
cophenet Cophenetic correlation coefficient
corr Linear or rank correlation
corrcov Convert covariance matrix to
correlation matrix
partialcorr Linear or rank partial correlation
coefficients
tiedrank Rank adjusted for ties
16-10
Statistical Visualization
Statistical Visualization
Distribution Plots (p. 16-11)
Scatter Plots (p. 16-12)
ANOVA Plots (p. 16-12)
Regression Plots (p. 16-13)
Multivariate Plots (p. 16-13)
Cluster Plots (p. 16-13)
Classification Plots (p. 16-14)
DOE Plots (p. 16-14)
SPC Plots (p. 16-14)
Distribution Plots
boxplot Box plot
cdfplot Empirical cumulative distribution
function plot
dfittool Interactive distribution fitting
disttool Interactive density and distribution
plots
ecdfhist Empirical cumulative distribution
function histogram
fsurfht Interactive contour plot
hist3 Bivariate histogram
histfit Histogram with normal fit
normplot Normal probability plot
normspec Normal density plot between
specifications
pareto Pareto chart
probplot Probability plots
16-11
16 Function Reference
qqplot Quantile-quantile plot
randtool Interactive random number
generation
scatterhist Scatter plot with marginal
histograms
surfht Interactive contour plot
wblplot Weibull probability plot
Scatter Plots
gline Interactively add line to plot
gname Add case names to plot
gplotmatrix Matrix of scatter plots by group
gscatter Scatter plot by group
lsline Add least-squares line to scatter plot
refcurve Add reference curve to plot
refline Add reference line to plot
scatterhist Scatter plot with marginal
histograms
ANOVA Plots
anova1 One-way analysis of variance
aoctool Interactive analysis of covariance
manovacluster Dendrogram of group mean clusters
following MANOVA
multcompare Multiple comparison test
16-12
Statistical Visualization
Regression Plots
addedvarplot Added-variable plot
gline Interactively add line to plot
lsline Add least-squares line to scatter plot
polytool Interactive polynomial fitting
rcoplot Residual case order plot
refcurve Add reference curve to plot
refline Add reference line to plot
robustdemo Interactive robust regression
rsmdemo Interactive response surface
demonstration
rstool Interactive response surface
modeling
view (classregtree) Plot tree
Multivariate Plots
andrewsplot Andrews plot
biplot Biplot
glyphplot Glyph plot
parallelcoords Parallel coordinates plot
Cluster Plots
dendrogram Dendrogram plot
manovacluster Dendrogram of group mean clusters
following MANOVA
silhouette Silhouette plot
16-13
16 Function Reference
Classification Plots
perfcurve Compute Receiver Operating
Characteristic (ROC) curve or other
performance curve for classifier
output
view (classregtree) Plot tree
DOE Plots
interactionplot Interaction plot for grouped data
maineffectsplot Main effects plot for grouped data
multivarichart Multivari chart for grouped data
rsmdemo Interactive response surface
demonstration
rstool Interactive response surface
modeling
SPC Plots
capaplot Process capability plot
controlchart Shewhart control charts
histfit Histogram with normal fit
normspec Normal density plot between
specifications
16-14
Probability Distributions
Probability Distributions
Distribution Objects (p. 16-15)
Distribution Plots (p. 16-16)
Probability Density (p. 16-17)
Cumulative Distribution (p. 16-19)
Inverse Cumulative Distribution
(p. 16-21)
Distribution Statistics (p. 16-23)
Distribution Fitting (p. 16-24)
Negative Log-Likelihood (p. 16-26)
Random Number Generators
(p. 16-26)
Quasi-Random Numbers (p. 16-28)
Piecewise Distributions (p. 16-29)
Distribution Objects
cdf (ProbDist) Return cumulative distribution
function (CDF) for ProbDist object
fitdist Fit probability distribution to data
icdf (ProbDistUnivKernel) Return inverse cumulative
distribution function (ICDF) for
ProbDistUnivKernel object
icdf (ProbDistUnivParam) Return inverse cumulative
distribution function (ICDF) for
ProbDistUnivParam object
iqr (ProbDistUnivKernel) Return interquartile range (IQR) for
ProbDistUnivKernel object
iqr (ProbDistUnivParam) Return interquartile range (IQR) for
ProbDistUnivParam object
16-15
16 Function Reference
mean (ProbDistUnivParam) Return mean of ProbDistUnivParam
object
median (ProbDistUnivKernel) Return median of
ProbDistUnivKernel object
median (ProbDistUnivParam) Return median of
ProbDistUnivParam object
paramci (ProbDistUnivParam) Return parameter confidence
intervals of ProbDistUnivParam
object
pdf (ProbDist) Return probability density function
(PDF) for ProbDist object
ProbDistUnivKernel Construct ProbDistUnivKernel
object
ProbDistUnivParam Construct ProbDistUnivParam
object
random (ProbDist) Generate random number drawn
from ProbDist object
std (ProbDistUnivParam) Return standard deviation of
ProbDistUnivParam object
var (ProbDistUnivParam) Return variance of
ProbDistUnivParam object
Distribution Plots
boxplot Box plot
cdfplot Empirical cumulative distribution
function plot
dfittool Interactive distribution fitting
disttool Interactive density and distribution
plots
ecdfhist Empirical cumulative distribution
function histogram
16-16
Probability Distributions
fsurfht Interactive contour plot
hist3 Bivariate histogram
histfit Histogram with normal fit
normplot Normal probability plot
normspec Normal density plot between
specifications
pareto Pareto chart
probplot Probability plots
qqplot Quantile-quantile plot
randtool Interactive random number
generation
scatterhist Scatter plot with marginal
histograms
surfht Interactive contour plot
wblplot Weibull probability plot
Probability Density
betapdf Beta probability density function
binopdf Binomial probability density
function
chi2pdf Chi-square probability density
function
copulapdf Copula probability density function
disttool Interactive density and distribution
plots
evpdf Extreme value probability density
function
exppdf Exponential probability density
function
16-17
16 Function Reference
fpdf F probability density function
gampdf Gamma probability density function
geopdf Geometric probability density
function
gevpdf Generalized extreme value
probability density function
gppdf Generalized Pareto probability
density function
hygepdf Hypergeometric probability density
function
ksdensity Kernel smoothing density estimate
lognpdf Lognormal probability density
function
mnpdf Multinomial probability density
function
mvnpdf Multivariate normal probability
density function
mvtpdf Multivariate t probability density
function
nbinpdf Negative binomial probability
density function
ncfpdf Noncentral F probability density
function
nctpdf Noncentral t probability density
function
ncx2pdf Noncentral chi-square probability
density function
normpdf Normal probability density function
pdf Probability density functions
pdf (gmdistribution) Probability density function for
Gaussian mixture distribution
16-18
Probability Distributions
pdf (piecewisedistribution) Probability density function for
piecewise distribution
poisspdf Poisson probability density function
random (piecewisedistribution) Random numbers from piecewise
distribution
raylpdf Rayleigh probability density function
tpdf Student’s t probability density
function
unidpdf Discrete uniform probability density
function
unifpdf Continuous uniform probability
density function
wblpdf Weibull probability density function
Cumulative Distribution
betacdf Beta cumulative distribution
function
binocdf Binomial cumulative distribution
function
cdf Cumulative distribution functions
cdf (gmdistribution) Cumulative distribution function for
Gaussian mixture distribution
cdf (piecewisedistribution) Cumulative distribution function for
piecewise distribution
cdfplot Empirical cumulative distribution
function plot
chi2cdf Chi-square cumulative distribution
function
copulacdf Copula cumulative distribution
function
16-19
16 Function Reference
disttool Interactive density and distribution
plots
ecdf Empirical cumulative distribution
function
ecdfhist Empirical cumulative distribution
function histogram
evcdf Extreme value cumulative
distribution function
expcdf Exponential cumulative distribution
function
fcdf F cumulative distribution function
gamcdf Gamma cumulative distribution
function
geocdf Geometric cumulative distribution
function
gevcdf Generalized extreme value
cumulative distribution function
gpcdf Generalized Pareto cumulative
distribution function
hygecdf Hypergeometric cumulative
distribution function
logncdf Lognormal cumulative distribution
function
mvncdf Multivariate normal cumulative
distribution function
mvtcdf Multivariate t cumulative
distribution function
ncfcdf Noncentral F cumulative
distribution function
nctcdf Noncentral t cumulative distribution
function
16-20
Probability Distributions
ncx2cdf Noncentral chi-square cumulative
distribution function
normcdf Normal cumulative distribution
function
poisscdf Poisson cumulative distribution
function
raylcdf Rayleigh cumulative distribution
function
tcdf Student’s t cumulative distribution
function
unidcdf Discrete uniform cumulative
distribution function
unifcdf Continuous uniform cumulative
distribution function
wblcdf Weibull cumulative distribution
function
Inverse Cumulative Distribution
betainv Beta inverse cumulative distribution
function
binoinv Binomial inverse cumulative
distribution function
chi2inv Chi-square inverse cumulative
distribution function
evinv Extreme value inverse cumulative
distribution function
expinv Exponential inverse cumulative
distribution function
finv F inverse cumulative distribution
function
16-21
16 Function Reference
gaminv Gamma inverse cumulative
distribution function
geoinv Geometric inverse cumulative
distribution function
gevinv Generalized extreme value inverse
cumulative distribution function
gpinv Generalized Pareto inverse
cumulative distribution function
hygeinv Hypergeometric inverse cumulative
distribution function
icdf Inverse cumulative distribution
functions
icdf (piecewisedistribution) Inverse cumulative distribution
function for piecewise distribution
logninv Lognormal inverse cumulative
distribution function
nbininv Negative binomial inverse
cumulative distribution function
ncfinv Noncentral F inverse cumulative
distribution function
nctinv Noncentral t inverse cumulative
distribution function
ncx2inv Noncentral chi-square inverse
cumulative distribution function
norminv Normal inverse cumulative
distribution function
poissinv Poisson inverse cumulative
distribution function
raylinv Rayleigh inverse cumulative
distribution function
tinv Student’s t inverse cumulative
distribution function
16-22
Probability Distributions
unidinv Discrete uniform inverse cumulative
distribution function
unifinv Continuous uniform inverse
cumulative distribution function
wblinv Weibull inverse cumulative
distribution function
Distribution Statistics
betastat Beta mean and variance
binostat Binomial mean and variance
chi2stat Chi-square mean and variance
copulastat Copula rank correlation
evstat Extreme value mean and variance
expstat Exponential mean and variance
fstat F mean and variance
gamstat Gamma mean and variance
geostat Geometric mean and variance
gevstat Generalized extreme value mean
and variance
gpstat Generalized Pareto mean and
variance
hygestat Hypergeometric mean and variance
lognstat Lognormal mean and variance
nbinstat Negative binomial mean and
variance
ncfstat Noncentral F mean and variance
nctstat Noncentral t mean and variance
ncx2stat Noncentral chi-square mean and
variance
16-23
16 Function Reference
normstat Normal mean and variance
poisstat Poisson mean and variance
raylstat Rayleigh mean and variance
tstat Student’s t mean and variance
unidstat Discrete uniform mean of and
variance
unifstat Continuous uniform mean and
variance
wblstat Weibull mean and variance
Distribution Fitting
Supported Distributions (p. 16-24)
Piecewise Distributions (p. 16-25)
Supported Distributions
betafit Beta parameter estimates
binofit Binomial parameter estimates
copulafit Fit copula to data
copulaparam Copula parameters as function of
rank correlation
dfittool Interactive distribution fitting
evfit Extreme value parameter estimates
expfit Exponential parameter estimates
fit (gmdistribution) Gaussian mixture parameter
estimates
gamfit Gamma parameter estimates
gevfit Generalized extreme value
parameter estimates
16-24
Probability Distributions
gpfit Generalized Pareto parameter
estimates
histfit Histogram with normal fit
johnsrnd Johnson system random numbers
lognfit Lognormal parameter estimates
mle Maximum likelihood estimates
mlecov Asymptotic covariance of maximum
likelihood estimators
nbinfit Negative binomial parameter
estimates
normfit Normal parameter estimates
normplot Normal probability plot
pearsrnd Pearson system random numbers
poissfit Poisson parameter estimates
raylfit Rayleigh parameter estimates
unifit Continuous uniform parameter
estimates
wblfit Weibull parameter estimates
wblplot Weibull probability plot
Piecewise Distributions
boundary (piecewisedistribution) Piecewise distribution boundaries
lowerparams (paretotails) Lower Pareto tails parameters
nsegments (piecewisedistribution) Number of segments
paretotails Construct Pareto tails object
piecewisedistribution Create piecewise distribution object
segment (piecewisedistribution) Segments containing values
upperparams (paretotails) Upper Pareto tails parameters
16-25
16 Function Reference
Negative Log-Likelihood
betalike Beta negative log-likelihood
evlike Extreme value negative
log-likelihood
explike Exponential negative log-likelihood
gamlike Gamma negative log-likelihood
gevlike Generalized extreme value negative
log-likelihood
gplike Generalized Pareto negative
log-likelihood
lognlike Lognormal negative log-likelihood
mvregresslike Negative log-likelihood for
multivariate regression
normlike Normal negative log-likelihood
wbllike Weibull negative log-likelihood
Random Number Generators
betarnd Beta random numbers
binornd Binomial random numbers
chi2rnd Chi-square random numbers
copularnd Copula random numbers
evrnd Extreme value random numbers
exprnd Exponential random numbers
frnd F random numbers
gamrnd Gamma random numbers
geornd Geometric random numbers
gevrnd Generalized extreme value random
numbers
16-26
Probability Distributions
gprnd Generalized Pareto random numbers
hygernd Hypergeometric random numbers
iwishrnd Inverse Wishart random numbers
johnsrnd Johnson system random numbers
lhsdesign Latin hypercube sample
lhsnorm Latin hypercube sample from normal
distribution
lognrnd Lognormal random numbers
mhsample Metropolis-Hastings sample
mnrnd Multinomial random numbers
mvnrnd Multivariate normal random
numbers
mvtrnd Multivariate t random numbers
nbinrnd Negative binomial random numbers
ncfrnd Noncentral F random numbers
nctrnd Noncentral t random numbers
ncx2rnd Noncentral chi-square random
numbers
normrnd Normal random numbers
pearsrnd Pearson system random numbers
poissrnd Poisson random numbers
randg Gamma random numbers
random Random numbers
random (gmdistribution) Random numbers from Gaussian
mixture distribution
random (piecewisedistribution) Random numbers from piecewise
distribution
randsample Random sample
16-27
16 Function Reference
randtool Interactive random number
generation
raylrnd Rayleigh random numbers
slicesample Slice sampler
trnd Student’s t random numbers
unidrnd Discrete uniform random numbers
unifrnd Continuous uniform random
numbers
wblrnd Weibull random numbers
wishrnd Wishart random numbers
Quasi-Random Numbers
addlistener (qrandstream) Add listener for event
delete (qrandstream) Delete handle object
end (qrandset) Last index in indexing expression for
point set
eq (qrandstream) Test handle equality
findobj (qrandstream) Find objects matching specified
conditions
findprop (qrandstream) Find property of MATLAB handle
object
ge (qrandstream) Greater than or equal relation for
handles
gt (qrandstream) Greater than relation for handles
haltonset Construct Halton quasi-random
point set
isvalid (qrandstream) Test handle validity
le (qrandstream) Less than or equal relation for
handles
16-28
Probability Distributions
length (qrandset) Length of point set
lt (qrandstream) Less than relation for handles
ndims (qrandset) Number of dimensions in matrix
ne (qrandstream) Not equal relation for handles
net (qrandset) Generate quasi-random point set
notify (qrandstream) Notify listeners of event
qrand (qrandstream) Generate quasi-random points from
stream
qrandset Abstract quasi-random point set
class
qrandstream Construct quasi-random number
stream
rand (qrandstream) Generate quasi-random points from
stream
reset (qrandstream) Reset state
scramble (qrandset) Scramble quasi-random point set
size (qrandset) Number of dimensions in matrix
sobolset Construct Sobol quasi-random point
set
Piecewise Distributions
boundary (piecewisedistribution) Piecewise distribution boundaries
cdf (piecewisedistribution) Cumulative distribution function for
piecewise distribution
icdf (piecewisedistribution) Inverse cumulative distribution
function for piecewise distribution
lowerparams (paretotails) Lower Pareto tails parameters
nsegments (piecewisedistribution) Number of segments
paretotails Construct Pareto tails object
16-29
16 Function Reference
pdf (piecewisedistribution) Probability density function for
piecewise distribution
piecewisedistribution Create piecewise distribution object
random (piecewisedistribution) Random numbers from piecewise
distribution
segment (piecewisedistribution) Segments containing values
upperparams (paretotails) Upper Pareto tails parameters
16-30
Hypothesis Tests
Hypothesis Tests
ansaribradley Ansari-Bradley test
barttest Bartlett’s test
canoncorr Canonical correlation
chi2gof Chi-square goodness-of-fit test
dwtest Durbin-Watson test
friedman Friedman’s test
jbtest Jarque-Bera test
kruskalwallis Kruskal-Wallis test
kstest One-sample Kolmogorov-Smirnov
test
kstest2 Two-sample Kolmogorov-Smirnov
test
lillietest Lilliefors test
linhyptest Linear hypothesis test
ranksum Wilcoxon rank sum test
runstest Run test for randomness
sampsizepwr Sample size and power of test
signrank Wilcoxon signed rank test
signtest Sign test
ttest One-sample and paired-sample t-test
ttest2 Two-sample t-test
vartest Chi-square variance test
vartest2 Two-sample F-test for equal
variances
vartestn Bartlett multiple-sample test for
equal variances
16-31
16 Function Reference
zscore Standardized z-scores
ztest z-test
Analysis of Variance
ANOVA Plots (p. 16-32)
ANOVA Operations (p. 16-32)
ANOVA Plots
anova1 One-way analysis of variance
aoctool Interactive analysis of covariance
manovacluster Dendrogram of group mean clusters
following MANOVA
multcompare Multiple comparison test
ANOVA Operations
anova1 One-way analysis of variance
anova2 Two-way analysis of variance
anovan N-way analysis of variance
aoctool Interactive analysis of covariance
dummyvar Create dummy variables
friedman Friedman’s test
kruskalwallis Kruskal-Wallis test
manova1 One-way multivariate analysis of
variance
16-32
Regression Analysis
manovacluster Dendrogram of group mean clusters
following MANOVA
multcompare Multiple comparison test
Regression Analysis
Regression Plots (p. 16-33)
Linear Regression (p. 16-34)
Nonlinear Regression (p. 16-35)
Regression Trees (p. 16-35)
Ensemble Methods (p. 16-36)
Regression Plots
addedvarplot Added-variable plot
gline Interactively add line to plot
lsline Add least-squares line to scatter plot
polytool Interactive polynomial fitting
rcoplot Residual case order plot
refcurve Add reference curve to plot
refline Add reference line to plot
robustdemo Interactive robust regression
rsmdemo Interactive response surface
demonstration
rstool Interactive response surface
modeling
view (classregtree) Plot tree
16-33
16 Function Reference
Linear Regression
coxphfit Cox proportional hazards regression
dummyvar Create dummy variables
glmfit Generalized linear model regression
glmval Generalized linear model values
invpred Inverse prediction
leverage Leverage
mnrfit Multinomial logistic regression
mnrval Multinomial logistic regression
values
mvregress Multivariate linear regression
mvregresslike Negative log-likelihood for
multivariate regression
plsregress Partial least-squares regression
polyconf Polynomial confidence intervals
polytool Interactive polynomial fitting
regress Multiple linear regression
regstats Regression diagnostics
ridge Ridge regression
robustdemo Interactive robust regression
robustfit Robust regression
rsmdemo Interactive response surface
demonstration
rstool Interactive response surface
modeling
stepwise Interactive stepwise regression
stepwisefit Stepwise regression
x2fx Convert predictor matrix to design
matrix
16-34
Regression Analysis
Nonlinear Regression
dummyvar Create dummy variables
hougen Hougen-Watson model
nlinfit Nonlinear regression
nlintool Interactive nonlinear regression
nlmefit Nonlinear mixed-effects estimation
nlparci Nonlinear regression parameter
confidence intervals
nlpredci Nonlinear regression prediction
confidence intervals
Regression Trees
catsplit (classregtree) Categorical splits used for branches
in decision tree
children (classregtree) Child nodes
classcount (classregtree) Class counts
classprob (classregtree) Class probabilities
classregtree Construct classification and
regression trees
cutcategories (classregtree) Cut categories
cutpoint (classregtree) Returns decision tree cut point
values
cuttype (classregtree) Cut types
cutvar (classregtree) Cut variable names
eval (classregtree) Predicted responses
isbranch (classregtree) Test node for branch
nodeerr (classregtree) Return vector of node errors
nodeprob (classregtree) Node probabilities
16-35
16 Function Reference
nodesize (classregtree) Return node size
numnodes (classregtree) Number of nodes
parent (classregtree) Parent node
prune (classregtree) Prune tree
risk (classregtree) Node risks
test (classregtree) Error rate
type (classregtree) Tree type
varimportance (classregtree) Compute embedded estimates of
input feature importance
view (classregtree) Plot tree
Ensemble Methods
append (TreeBagger) Append new trees to ensemble
combine (CompactTreeBagger) Combine two ensembles
compact (TreeBagger) Compact ensemble of decision trees
CompactTreeBagger Create CompactTreeBagger object
error (CompactTreeBagger) Error (misclassification probability
or MSE)
error (TreeBagger) Error (misclassification probability
or MSE)
fillProximities (TreeBagger) Proximity matrix for training data
growTrees (TreeBagger) Train additional trees and add to
ensemble
margin (CompactTreeBagger) Classification margin
margin (TreeBagger) Classification margin
mdsProx (CompactTreeBagger) Multidimensional scaling of
proximity matrix
mdsProx (TreeBagger) Multidimensional scaling of
proximity matrix
16-36
Regression Analysis
meanMargin (CompactTreeBagger) Mean classification margin
meanMargin (TreeBagger) Mean classification margin
oobError (TreeBagger) Out-of-bag error
oobMargin (TreeBagger) Out-of-bag margins
oobMeanMargin (TreeBagger) Out-of-bag mean margins
oobPredict (TreeBagger) Ensemble predictions for out-of-bag
observations
outlierMeasure Outlier measure for data
(CompactTreeBagger)
predict (CompactTreeBagger) Predict response
predict (TreeBagger) Predict response
proximity (CompactTreeBagger) Proximity matrix for data
SetDefaultYfit Set default value for predict
(CompactTreeBagger)
TreeBagger Create ensemble of bagged decision
trees
16-37
16 Function Reference
Multivariate Methods
Multivariate Plots (p. 16-38)
Multidimensional Scaling (p. 16-38)
Procrustes Analysis (p. 16-38)
Feature Selection (p. 16-39)
Feature Transformation (p. 16-39)
Multivariate Plots
andrewsplot Andrews plot
biplot Biplot
glyphplot Glyph plot
parallelcoords Parallel coordinates plot
Multidimensional Scaling
cmdscale Classical multidimensional scaling
mahal Mahalanobis distance
mdscale Nonclassical multidimensional
scaling
pdist Pairwise distance between pairs of
objects
pdist2 Pairwise distance between two sets
of observations
squareform Format distance matrix
Procrustes Analysis
procrustes Procrustes analysis
16-38
Multivariate Methods
Feature Selection
sequentialfs Sequential feature selection
Feature Transformation
Nonnegative Matrix Factorization
(p. 16-39)
Principal Component Analysis
(p. 16-39)
Factor Analysis (p. 16-39)
Nonnegative Matrix Factorization
nnmf Nonnegative matrix factorization
Principal Component Analysis
barttest Bartlett’s test
pareto Pareto chart
pcacov Principal component analysis on
covariance matrix
pcares Residuals from principal component
analysis
princomp Principal component analysis on
data
Factor Analysis
factoran Factor analysis
16-39
16 Function Reference
Cluster Analysis
Cluster Plots (p. 16-40)
Hierarchical Clustering (p. 16-40)
K-Means Clustering (p. 16-41)
Gaussian Mixture Models (p. 16-41)
Cluster Plots
dendrogram Dendrogram plot
manovacluster Dendrogram of group mean clusters
following MANOVA
silhouette Silhouette plot
Hierarchical Clustering
cluster Construct agglomerative clusters
from linkages
clusterdata Construct agglomerative clusters
from data
cophenet Cophenetic correlation coefficient
inconsistent Inconsistency coefficient
linkage Create agglomerative hierarchical
cluster tree
pdist Pairwise distance between pairs of
objects
pdist2 Pairwise distance between two sets
of observations
squareform Format distance matrix
16-40
Model Assessment
K-Means Clustering
kmeans K-means clustering
mahal Mahalanobis distance
Gaussian Mixture Models
cdf (gmdistribution) Cumulative distribution function for
Gaussian mixture distribution
cluster (gmdistribution) Construct clusters from Gaussian
mixture distribution
gmdistribution Construct Gaussian mixture
distribution
mahal (gmdistribution) Mahalanobis distance to component
means
pdf (gmdistribution) Probability density function for
Gaussian mixture distribution
posterior (gmdistribution) Posterior probabilities of components
random (gmdistribution) Random numbers from Gaussian
mixture distribution
Model Assessment
confusionmat Confusion matrix
crossval Loss estimate using cross-validation
cvpartition Create cross-validation partition for
data
repartition (cvpartition) Repartition data for cross-validation
test (cvpartition) Test indices for cross-validation
training (cvpartition) Training indices for cross-validation
16-41
16 Function Reference
Classification
Classification Plots (p. 16-42)
Discriminant Analysis (p. 16-42)
Classification Trees (p. 16-42)
Naive Bayes Classification (p. 16-43)
Ensemble Methods (p. 16-44)
Classification Plots
perfcurve Compute Receiver Operating
Characteristic (ROC) curve or other
performance curve for classifier
output
view (classregtree) Plot tree
Discriminant Analysis
classify Discriminant analysis
mahal Mahalanobis distance
Classification Trees
catsplit (classregtree) Categorical splits used for branches
in decision tree
children (classregtree) Child nodes
classcount (classregtree) Class counts
classprob (classregtree) Class probabilities
classregtree Construct classification and
regression trees
cutcategories (classregtree) Cut categories
16-42
Classification
cutpoint (classregtree) Returns decision tree cut point
values
cuttype (classregtree) Cut types
cutvar (classregtree) Cut variable names
eval (classregtree) Predicted responses
isbranch (classregtree) Test node for branch
nodeerr (classregtree) Return vector of node errors
nodeprob (classregtree) Node probabilities
nodesize (classregtree) Return node size
numnodes (classregtree) Number of nodes
parent (classregtree) Parent node
prune (classregtree) Prune tree
risk (classregtree) Node risks
test (classregtree) Error rate
type (classregtree) Tree type
varimportance (classregtree) Compute embedded estimates of
input feature importance
view (classregtree) Plot tree
Naive Bayes Classification
fit (NaiveBayes) Create Naive Bayes classifier object
by fitting training data
NaiveBayes Create NaiveBayes object
posterior (NaiveBayes) Compute posterior probability of
each class for test data
predict (NaiveBayes) Predict class label for test data
16-43
16 Function Reference
Ensemble Methods
append (TreeBagger) Append new trees to ensemble
combine (CompactTreeBagger) Combine two ensembles
compact (TreeBagger) Compact ensemble of decision trees
CompactTreeBagger Create CompactTreeBagger object
error (CompactTreeBagger) Error (misclassification probability
or MSE)
error (TreeBagger) Error (misclassification probability
or MSE)
fillProximities (TreeBagger) Proximity matrix for training data
growTrees (TreeBagger) Train additional trees and add to
ensemble
margin (CompactTreeBagger) Classification margin
margin (TreeBagger) Classification margin
mdsProx (CompactTreeBagger) Multidimensional scaling of
proximity matrix
mdsProx (TreeBagger) Multidimensional scaling of
proximity matrix
meanMargin (CompactTreeBagger) Mean classification margin
meanMargin (TreeBagger) Mean classification margin
oobError (TreeBagger) Out-of-bag error
oobMargin (TreeBagger) Out-of-bag margins
oobMeanMargin (TreeBagger) Out-of-bag mean margins
oobPredict (TreeBagger) Ensemble predictions for out-of-bag
observations
outlierMeasure Outlier measure for data
(CompactTreeBagger)
predict (CompactTreeBagger) Predict response
predict (TreeBagger) Predict response
16-44
Classification
proximity (CompactTreeBagger) Proximity matrix for data
SetDefaultYfit Set default value for predict
(CompactTreeBagger)
TreeBagger Create ensemble of bagged decision
trees
16-45
16 Function Reference
Hidden Markov Models
hmmdecode Hidden Markov model posterior
state probabilities
hmmestimate Hidden Markov model parameter
estimates from emissions and states
hmmgenerate Hidden Markov model states and
emissions
hmmtrain Hidden Markov model parameter
estimates from emissions
hmmviterbi Hidden Markov model most probable
state path
16-46
Design of Experiments
Design of Experiments
DOE Plots (p. 16-47)
Full Factorial Designs (p. 16-47)
Fractional Factorial Designs
(p. 16-48)
Response Surface Designs (p. 16-48)
D-Optimal Designs (p. 16-48)
Latin Hypercube Designs (p. 16-48)
Quasi-Random Designs (p. 16-49)
DOE Plots
interactionplot Interaction plot for grouped data
maineffectsplot Main effects plot for grouped data
multivarichart Multivari chart for grouped data
rsmdemo Interactive response surface
demonstration
rstool Interactive response surface
modeling
Full Factorial Designs
ff2n Two-level full factorial design
fullfact Full factorial design
16-47
16 Function Reference
Fractional Factorial Designs
fracfact Fractional factorial design
fracfactgen Fractional factorial design
generators
Response Surface Designs
bbdesign Box-Behnken design
ccdesign Central composite design
D-Optimal Designs
candexch Candidate set row exchange
candgen Candidate set generation
cordexch Coordinate exchange
daugment D-optimal augmentation
dcovary D-optimal design with fixed
covariates
rowexch Row exchange
rsmdemo Interactive response surface
demonstration
Latin Hypercube Designs
lhsdesign Latin hypercube sample
lhsnorm Latin hypercube sample from normal
distribution
16-48
Design of Experiments
Quasi-Random Designs
addlistener (qrandstream) Add listener for event
delete (qrandstream) Delete handle object
end (qrandset) Last index in indexing expression for
point set
eq (qrandstream) Test handle equality
findobj (qrandstream) Find objects matching specified
conditions
findprop (qrandstream) Find property of MATLAB handle
object
ge (qrandstream) Greater than or equal relation for
handles
gt (qrandstream) Greater than relation for handles
haltonset Construct Halton quasi-random
point set
isvalid (qrandstream) Test handle validity
le (qrandstream) Less than or equal relation for
handles
length (qrandset) Length of point set
lt (qrandstream) Less than relation for handles
ndims (qrandset) Number of dimensions in matrix
ne (qrandstream) Not equal relation for handles
net (qrandset) Generate quasi-random point set
notify (qrandstream) Notify listeners of event
qrand (qrandstream) Generate quasi-random points from
stream
qrandset Abstract quasi-random point set
class
qrandstream Construct quasi-random number
stream
16-49
16 Function Reference
rand (qrandstream) Generate quasi-random points from
stream
reset (qrandstream) Reset state
scramble (qrandset) Scramble quasi-random point set
size (qrandset) Number of dimensions in matrix
sobolset Construct Sobol quasi-random point
set
16-50
Statistical Process Control
Statistical Process Control
SPC Plots (p. 16-51)
SPC Functions (p. 16-51)
SPC Plots
capaplot Process capability plot
controlchart Shewhart control charts
histfit Histogram with normal fit
normspec Normal density plot between
specifications
SPC Functions
capability Process capability indices
controlrules Western Electric and Nelson control
rules
gagerr Gage repeatability and
reproducibility study
16-51
16 Function Reference
GUIs
aoctool Interactive analysis of covariance
dfittool Interactive distribution fitting
disttool Interactive density and distribution
plots
fsurfht Interactive contour plot
polytool Interactive polynomial fitting
randtool Interactive random number
generation
regstats Regression diagnostics
robustdemo Interactive robust regression
rsmdemo Interactive response surface
demonstration
rstool Interactive response surface
modeling
surfht Interactive contour plot
16-52
Utilities
Utilities
combnk Enumeration of combinations
perms Enumeration of permutations
statget Access values in statistics options
structure
statset Create statistics options structure
zscore Standardized z-scores
16-53
16 Function Reference
16-54
17
Class Reference
• “Data Organization” on page 17-2
• “Probability Distributions” on page 17-3
• “Regression Analysis” on page 17-4
• “Gaussian Mixture Models” on page 17-4
• “Model Assessment” on page 17-5
• “Classification” on page 17-5
• “Quasi-Random Design of Experiments” on page 17-6
17 Class Reference
Data Organization
In this section...
“Categorical Arrays” on page 17-2
“Dataset Arrays” on page 17-2
Categorical Arrays
categorical Arrays for categorical data
nominal Arrays for nominal categorical data
ordinal Arrays for ordinal categorical data
Dataset Arrays
dataset Arrays for statistical data
17-2
Probability Distributions
Probability Distributions
In this section...
“Distribution Objects” on page 17-3
“Quasi-Random Numbers” on page 17-3
“Piecewise Distributions” on page 17-4
Distribution Objects
ProbDist Object representing probability
distribution
ProbDistKernel Object representing nonparametric
probability distribution defined by
kernel smoothing
ProbDistParametric Object representing parametric
probability distribution
ProbDistUnivKernel Object representing univariate
kernel probability distribution
ProbDistUnivParam Object representing univariate
parametric probability distribution
Quasi-Random Numbers
haltonset Halton quasi-random point sets
qrandset Quasi-random point sets
qrandstream Quasi-random number streams
sobolset Sobol quasi-random point sets
17-3
17 Class Reference
Piecewise Distributions
paretotails Empirical distributions with Pareto
tails
piecewisedistribution Piecewise-defined distributions
Regression Analysis
In this section...
“Regression Trees” on page 17-4
“Ensemble Method Classes” on page 17-4
Regression Trees
classregtree Classification and regression trees
Ensemble Method Classes
CompactTreeBagger Compact ensemble of decision trees
grown by bootstrap aggregation
TreeBagger Bootstrap aggregation for ensemble
of decision trees
Gaussian Mixture Models
gmdistribution Gaussian mixture models
17-4
Model Assessment
Model Assessment
cvpartition Data partitions for cross-validation
Classification
In this section...
“Classification Trees” on page 17-5
“Naive Bayes Classification ” on page 17-5
“Ensemble Method Classes” on page 17-5
Classification Trees
classregtree Classification and regression trees
Naive Bayes Classification
NaiveBayes Naive Bayes classifier
Ensemble Method Classes
CompactTreeBagger Compact ensemble of decision trees
grown by bootstrap aggregation
TreeBagger Bootstrap aggregation for ensemble
of decision trees
17-5
17 Class Reference
Quasi-Random Design of Experiments
haltonset Halton quasi-random point sets
qrandset Quasi-random point sets
qrandstream Quasi-random number streams
sobolset Sobol quasi-random point sets
17-6
18
Functions — Alphabetical
List
addedvarplot
Purpose Added-variable plot
Syntax addedvarplot(X,y,num,inmodel)
addedvarplot(X,y,num,inmodel,stats)
Description addedvarplot(X,y,num,inmodel) displays an added variable plot
using the predictive terms in X, the response values in y, the added
term in column num of X, and the model with current terms specified by
inmodel. X is an n-by-p matrix of n observations of p predictive terms.
y is vector of n response values. num is a scalar index specifying the
column of X with the term to be added. inmodel is a logical vector of p
elements specifying the columns of X in the current model. By default,
all elements of inmodel are false.
Note addedvarplot automatically includes a constant term in all
models. Do not enter a column of 1s directly into X.
addedvarplot(X,y,num,inmodel,stats) uses the stats output from
the stepwisefit function to improve the efficiency of repeated calls to
addedvarplot. Otherwise, this syntax is equivalent to the previous
syntax.
Added variable plots are used to determine the unique effect of adding
a new term to a multilinear model. The plot shows the relationship
between the part of the response unexplained by terms already in the
model and the part of the new term unexplained by terms already in
the model. The “unexplained” parts are measured by the residuals of
the respective regressions. A scatter of the residuals from the two
regressions forms the added variable plot.
In addition to the scatter of residuals, the plot produced by
addedvarplot shows 95% confidence intervals on predictions from the
fitted line. The fitted line has intercept zero because, under typical
linear model assumptions, both of the plotted variables have mean zero.
The slope of the fitted line is the coefficient that the new term would
have if it were added to the model with terms inmodel.
18-2
addedvarplot
Added variable plots are sometimes known as partial regression
leverage plots.
Examples Load the data in hald.mat, which contains observations of the heat of
reaction of various cement mixtures:
load hald
whos
Name Size Bytes Class Attributes
Description 22x58 2552 char
hald 13x5 520 double
heat 13x1 104 double
ingredients 13x4 416 double
Create an added variable plot to investigate the addition of the third
column of ingredients to a model consisting of the first two columns:
inmodel = [true true false false];
addedvarplot(ingredients,heat,3,inmodel)
18-3
addedvarplot
The wide scatter and the low slope of the fitted line are evidence against
the statistical significance of adding the third column to the model.
See Also stepwisefit | stepwise
18-4
categorical.addlevels
Purpose Add levels to categorical array
Syntax B = addlevels(A,newlevels)
Description B = addlevels(A,newlevels) adds new levels to the categorical array
A. newlevels is a cell array of strings or a 2-D character matrix that
specifies the levels to add. addlevels adds the new levels at the end of
the list of possible categorical levels in A, but does not modify the value
of any element. B does not contain elements at the new levels.
Examples Example 1
Add levels for additional species in Fisher’s iris data:
load fisheriris
species = nominal(species,...
{'Species1','Species2','Species3'},...
{'setosa','versicolor','virginica'});
species = addlevels(species,{'Species4','Species5'});
getlabels(species)
ans =
'Species1' 'Species2' 'Species3' 'Species4' 'Species5'
Example 2
1 Load patient data from the CSV file hospital.dat and store the
information in a dataset array with observation names given by the
first column in the data (patient identification):
patients = dataset('file','hospital.dat',...
'delimiter',',',...
'ReadObsNames',true);
2 Make the {0,1}-valued variable smoke nominal, and change the labels
to 'No' and 'Yes':
patients.smoke = nominal(patients.smoke,{'No','Yes'});
18-5
categorical.addlevels
3 Add new levels to smoke as placeholders for more detailed histories
of smokers:
patients.smoke = addlevels(patients.smoke,...
{'0-5 Years','5-10 Years','LongTerm'});
4 Assuming the nonsmokers have never smoked, relabel the 'No' level:
patients.smoke = setlabels(patients.smoke,'Never','No');
5 Drop the undifferentiated 'Yes' level from smoke:
patients.smoke = droplevels(patients.smoke,'Yes');
Warning: OLDLEVELS contains categorical levels that
were present in A, caused some array elements to have
undefined levels.
Note that smokers now have an undefined level.
6 Set each smoker to one of the new levels, by observation name:
patients.smoke('YPL-320') = '5-10 Years';
See Also droplevels | getlabels | islevel | mergelevels | reorderlevels
18-6
qrandstream.addlistener
Purpose Add listener for event
Syntax el = addlistener(hsource,'eventname',callback)
el = addlistener(hsource,property,'eventname',callback)
Description el = addlistener(hsource,'eventname',callback) creates a
listener for the event named eventname, the source of which is handle
object hsource. If hsource is an array of source handles, the listener
responds to the named event on any handle in the array. callback is a
function handle that is invoked when the event is triggered.
el = addlistener(hsource,property,'eventname',callback)
adds a listener for a property event. eventname must be one of the
strings 'PreGet', 'PostGet', 'PreSet', and 'PostSet'. property
must be either a property name or cell array of property names, or a
meta.property or array of meta.property. The properties must belong
to the class of hsource. If hsource is scalar, property can include
dynamic properties.
For all forms, addlistener returns an event.listener. To remove
a listener, delete the object returned by addlistener. For example,
delete(el) calls the handle class delete method to remove the listener
and delete it from the workspace.
See Also delete | dynamicprops | event.listener | events | meta.property
| notify | qrandstream | reset
18-7
gmdistribution.AIC property
Purpose Akaike Information Criterion
Description The Akaike Information Criterion: 2*NlogL+2*m, where m is the number
of estimated parameters.
Note This property applies only to gmdistribution objects constructed
with fit.
18-8
andrewsplot
Purpose Andrews plot
Syntax andrewsplot(X)
andrewsplot(X,...,'Standardize',standopt)
andrewsplot(X,...,'Quantile',alpha)
andrewsplot(X,...,'Group',group)
andrewsplot(X,...,’PropName’,PropVal,...)
h = andrewsplot(X,...)
Description andrewsplot(X) creates an Andrews plot of the multivariate data in
the matrix X. The rows of X correspond to observations, the columns to
variables. Andrews plots represent each observation by a function f(t) of
a continuous dummy variable t over the interval [0,1]. f(t) is defined for
the i th observation in X as
f (t) = X (i,1) / 2 + X (i, 2)sin(2 t) + X (i, 3) cos(2 t) + …
andrewsplot treats NaN values in X as missing values and ignores the
corresponding rows.
andrewsplot(X,...,'Standardize',standopt) creates an Andrews
plot where standopt is one of the following:
• 'on' — scales each column of X to have mean 0 and standard
deviation 1 before making the plot.
• 'PCA' — creates an Andrews plot from the principal component
scores of X, in order of decreasing eigenvalue. (See princomp.)
• 'PCAStd' — creates an Andrews plot using the standardized
principal component scores. (See princomp.)
andrewsplot(X,...,'Quantile',alpha) plots only the median and
the alpha and (1 – alpha) quantiles of f(t) at each value of t. This is
useful if X contains many observations.
andrewsplot(X,...,'Group',group) plots the data in different groups
with different colors. Groups are defined by group, a numeric array
containing a group index for each observation. group can also be a
18-9
andrewsplot
categorical array, character matrix, or cell array of strings containing a
group name for each observation. (See “Grouped Data” on page 2-34.)
andrewsplot(X,...,’PropName’,PropVal,...) sets optional
lineseries object properties to the specified values for all lineseries
objects created by andrewsplot. (See Lineseries Properties.)
h = andrewsplot(X,...) returns a column vector of handles to the
lineseries objects created by andrewsplot, one handle per row of X. If
you use the 'Quantile' input parameter, h contains one handle for each
of the three lineseries objects created. If you use both the 'Quantile'
and the 'Group' input parameters, h contains three handles for each
group.
Examples Make a grouped plot of the Fisher iris data:
load fisheriris
andrewsplot(meas,'group',species)
18-10
andrewsplot
Plot only the median and quartiles of each group:
andrewsplot(meas,'group',species,'quantile',.25)
18-11
andrewsplot
See Also parallelcoords | glyphplot
How To • “Grouped Data” on page 2-34
18-12
anova1
Purpose One-way analysis of variance
Syntax p = anova1(X)
p = anova1(X,group)
p = anova1(X,group,displayopt)
[p,table] = anova1(...)
[p,table,stats] = anova1(...)
Description p = anova1(X) performs balanced one-way ANOVA for comparing
the means of two or more columns of data in the matrix X, where
each column represents an independent sample containing mutually
independent observations. The function returns the p value under the
null hypothesis that all samples in X are drawn from populations with
the same mean.
If p is near zero, it casts doubt on the null hypothesis and suggests
that at least one sample mean is significantly different than the other
sample means. Common significance levels are 0.05 or 0.01.
The anova1 function displays two figures, the standard ANOVA table
and a box plot of the columns of X.
The standard ANOVA table divides the variability of the data into two
parts:
• Variability due to the differences among the column means
(variability between groups)
• Variability due to the differences between the data in each column
and the column mean (variability within groups)
The standard ANOVA table has six columns:
1 The source of the variability.
2 The sum of squares (SS) due to each source.
3 The degrees of freedom (df) associated with each source.
18-13
anova1
4 The mean squares (MS) for each source, which is the ratio SS/df.
5 The F-statistic, which is the ratio of the mean squares.
6 The p value, which is derived from the cdf of F.
The box plot of the columns of X suggests the size of the F-statistic and
the p value. Large differences in the center lines of the boxes correspond
to large values of F and correspondingly small values of p.
Columns of X with NaN values are disregarded.
p = anova1(X,group) performs ANOVA by group.
If X is a matrix, anova1 treats each column as a separate group, and
evaluates whether the population means of the columns are equal. This
form of anova1 is appropriate when each group has the same number of
elements (balanced ANOVA). group can be a character array or a cell
array of strings, with one row per column of X, containing group names.
Enter an empty array ([]) or omit this argument if you do not want to
specify group names.
If X is a vector, group must be a categorical variable, vector, string
array, or cell array of strings with one name for each element of X. X
values corresponding to the same value of group are placed in the same
group. This form of anova1 is appropriate when groups have different
numbers of elements (unbalanced ANOVA).
If group contains empty or NaN-valued cells or strings, the corresponding
observations in X are disregarded.
p = anova1(X,group,displayopt) enables the ANOVA table and box
plot displays when displayopt is 'on' (default) and suppresses the
displays when displayopt is 'off'. Notches in the boxplot provide a
test of group medians (see boxplot) different from the F test for means
in the ANOVA table.
[p,table] = anova1(...) returns the ANOVA table (including
column and row labels) in the cell array table. Copy a text version of
the ANOVA table to the clipboard using the Copy Text item on the
Edit menu.
18-14
anova1
[p,table,stats] = anova1(...) returns a structure stats used
to perform a follow-up multiple comparison test. anova1 evaluates
the hypothesis that the samples all have the same mean against the
alternative that the means are not all the same. Sometimes it is
preferable to perform a test to determine which pairs of means are
significantly different, and which are not. Use the multcompare function
to perform such tests by supplying the stats structure as input.
Assumptions
The ANOVA test makes the following assumptions about the data in X:
• All sample populations are normally distributed.
• All sample populations have equal variance.
• All observations are mutually independent.
The ANOVA test is known to be robust with respect to modest violations
of the first two assumptions.
Examples Example 1
Create X with columns that are constants plus random normal
disturbances with mean zero and standard deviation one:
X = meshgrid(1:5)
X =
1 2 3 4 5
1 2 3 4 5
1 2 3 4 5
1 2 3 4 5
1 2 3 4 5
X = X + normrnd(0,1,5,5)
X =
1.3550 2.0662 2.4688 5.9447 5.4897
2.0693 1.7611 1.4864 4.8826 6.3222
2.1919 0.7276 3.1905 4.8768 4.6841
2.7620 1.8179 3.9506 4.4678 4.9291
18-15
anova1
-0.3626 1.1685 3.5742 2.1945 5.9465
Perform one-way ANOVA:
p = anova1(X)
p =
7.9370e-006
18-16
anova1
The very small p value indicates that differences between column
means are highly significant. The probability of this outcome under the
null hypothesis (that samples drawn from the same population would
have means differing by the amounts seen in X) is equal to the p value.
Example 2
The following example is from a study of the strength of structural
beams in Hogg. The vector strength measures deflections of beams in
thousandths of an inch under 3,000 pounds of force. The vector alloy
identifies each beam as steel ('st'), alloy 1 ('al1'), or alloy 2 ('al2').
(Although alloy is sorted in this example, grouping variables do not
need to be sorted.) The null hypothesis is that steel beams are equal in
strength to beams made of the two more expensive alloys.
strength = [82 86 79 83 84 85 86 87 74 82 ...
78 75 76 77 79 79 77 78 82 79];
alloy = {'st','st','st','st','st','st','st','st',...
'al1','al1','al1','al1','al1','al1',...
'al2','al2','al2','al2','al2','al2'};
p = anova1(strength,alloy)
p =
1.5264e-004
18-17
anova1
The p value suggests rejection of the null hypothesis. The box plot
shows that steel beams deflect more than beams made of the more
expensive alloys.
References [1] Hogg, R. V., and J. Ledolter. Engineering Statistics. New York:
MacMillan, 1987.
See Also anova2 | anovan | boxplot | manova1 | multcompare
How To • “Grouped Data” on page 2-34
18-18
anova2
Purpose Two-way analysis of variance
Syntax p = anova2(X,reps)
p = anova2(X,reps,displayopt)
[p,table] = anova2(...)
[p,table,stats] = anova2(...)
Description p = anova2(X,reps) performs a balanced two-way ANOVA for
comparing the means of two or more columns and two or more rows of
the observations in X. The data in different columns represent changes
in factor A. The data in different rows represent changes in factor B.
If there is more than one observation for each combination of factors,
input reps indicates the number of replicates in each position, which
must be constant. (For unbalanced designs, use anovan.)
The matrix below shows the format for a set-up where column factor
A has two levels, row factor B has three levels, and there are two
replications (reps = 2). The subscripts indicate row, column, and
replicate, respectively.
A =1 A=2
⎡ x111 x121 ⎤ ⎫
⎢x ⎬B =1
⎢ 112 x122 ⎥⎥ ⎭
⎢ x211 x221 ⎥ ⎫
⎢ ⎥ ⎬B = 2
⎢ x212 x222 ⎥ ⎭
⎢x x321 ⎥ ⎫
⎢ 311 ⎥ ⎬B = 3
⎣⎢ x312 x322 ⎥⎦ ⎭
When reps is 1 (default), anova2 returns two p-values in vector p:
1 The p value for the null hypothesis, H0A, that all samples from factor
A (i.e., all column-samples in X) are drawn from the same population
2 The p value for the null hypothesis, H0B, that all samples from factor
B (i.e., all row-samples in X) are drawn from the same population
18-19
anova2
When reps is greater than 1, anova2 returns a third p value in vector
p:
3 The p value for the null hypothesis, H0AB, that the effects due to
factors A and B are additive (i.e., that there is no interaction between
factors A and B)
If any p value is near zero, this casts doubt on the associated null
hypothesis. A sufficiently small p value for H0A suggests that at least
one column-sample mean is significantly different that the other
column-sample means; i.e., there is a main effect due to factor A. A
sufficiently small p value for H0B suggests that at least one row-sample
mean is significantly different than the other row-sample means; i.e.,
there is a main effect due to factor B. A sufficiently small p value for
H0AB suggests that there is an interaction between factors A and B.
The choice of a limit for the p value to determine whether a result
is “statistically significant” is left to the researcher. It is common to
declare a result significant if the p value is less than 0.05 or 0.01.
anova2 also displays a figure showing the standard ANOVA table,
which divides the variability of the data in X into three or four parts
depending on the value of reps:
• The variability due to the differences among the column means
• The variability due to the differences among the row means
• The variability due to the interaction between rows and columns (if
reps is greater than its default value of one)
• The remaining variability not explained by any systematic source
The ANOVA table has five columns:
• The first shows the source of the variability.
• The second shows the Sum of Squares (SS) due to each source.
• The third shows the degrees of freedom (df) associated with each
source.
18-20
anova2
• The fourth shows the Mean Squares (MS), which is the ratio SS/df.
• The fifth shows the F statistics, which is the ratio of the mean
squares.
p = anova2(X,reps,displayopt) enables the ANOVA table display
when displayopt is 'on' (default) and suppresses the display when
displayopt is 'off'.
[p,table] = anova2(...) returns the ANOVA table (including
column and row labels) in cell array table. (Copy a text version of the
ANOVA table to the clipboard by using the Copy Text item on the Edit
menu.)
[p,table,stats] = anova2(...) returns a stats structure that you
can use to perform a follow-up multiple comparison test.
The anova2 test evaluates the hypothesis that the row, column, and
interaction effects are all the same, against the alternative that they
are not all the same. Sometimes it is preferable to perform a test
to determine which pairs of effects are significantly different, and
which are not. Use the multcompare function to perform such tests by
supplying the stats structure as input.
Examples The data below come from a study of popcorn brands and popper type
(Hogg 1987). The columns of the matrix popcorn are brands (Gourmet,
National, and Generic). The rows are popper type (Oil and Air.) The
study popped a batch of each brand three times with each popper. The
values are the yield in cups of popped popcorn.
load popcorn
popcorn
popcorn =
5.5000 4.5000 3.5000
5.5000 4.5000 4.0000
6.0000 4.0000 3.0000
6.5000 5.0000 4.0000
7.0000 5.5000 5.0000
18-21
anova2
7.0000 5.0000 4.5000
p = anova2(popcorn,3)
p =
0.0000 0.0001 0.7462
The vector p shows the p-values for the three brands of popcorn, 0.0000,
the two popper types, 0.0001, and the interaction between brand and
popper type, 0.7462. These values indicate that both popcorn brand and
popper type affect the yield of popcorn, but there is no evidence of a
synergistic (interaction) effect of the two.
The conclusion is that you can get the greatest yield using the Gourmet
brand and an Air popper (the three values popcorn(4:6,1)).
References [1] Hogg, R. V., and J. Ledolter. Engineering Statistics. New York:
MacMillan, 1987.
See Also anova1 | anovan
18-22
anovan
Purpose N-way analysis of variance
Syntax p = anovan(y,group)
p = anovan(y,group,param,val)
[p,table] = anovan(y,group,param,val)
[p,table,stats] = anovan(y,group,param,val)
[p,table,stats,terms] = anovan(y,group,param,val)
Description p = anovan(y,group) performs multiway (n-way) analysis of variance
(ANOVA) for testing the effects of multiple factors (grouping variables)
on the mean of the vector y. This test compares the variance explained
by factors to the left over variance that cannot be explained. The factors
and factor levels of the observations in y are assigned by the cell array
group. Each of the cells in the cell array group contains a list of factor
levels identifying the observations in y with respect to one of the factors.
The list within each cell can be a categorical array, numeric vector,
character matrix, or single-column cell array of strings, and must have
the same number of elements as y. The fitted ANOVA model includes
the main effects of each grouping variable. All grouping variables are
treated as fixed effects by default. The result p is a vector of p-values,
one per term. For an example, see “Example of Three-Way ANOVA”
on page 18-27.
p = anovan(y,group,param,val) specifies one or more of the
parameter name/value pairs described in the following table.
Parameter Value
'alpha' A number between 0 and 1 requesting 100(1 –
alpha)% confidence bounds (default 0.05 for 95%
confidence)
'continuous' A vector of indices indicating which grouping
variables should be treated as continuous predictors
rather than as categorical predictors.
'display' 'on' displays an ANOVA table (the default)
'off' omits the display
18-23
anovan
Parameter Value
'model' The type of model used. See “Model Type” on page
18-25 for a description of this parameter.
'nested' A matrix M of 0’s and 1’s specifying the nesting
relationships among the grouping variables. M(i,j) is
1 if variable i is nested in variable j.
'random' A vector of indices indicating which grouping
variables are random effects (all are fixed by default).
See “ANOVA with Random Effects” on page 8-19 for
an example of how to use 'random'.
'sstype' 1, 2, 3 (default), or h specifies the type of sum of
squares. See “Sum of Squares” on page 18-26 for a
description of this parameter.
'varnames' A character matrix or a cell array of strings specifying
names of grouping variables, one per grouping
variable. When you do not specify 'varnames', the
default labels 'X1', 'X2', 'X3', ..., 'XN' are used.
See “ANOVA with Random Effects” on page 8-19 for
an example of how to use 'varnames'.
[p,table] = anovan(y,group,param,val) returns the ANOVA table
(including factor labels) in cell array table. (Copy a text version of
the ANOVA table to the clipboard by using the Copy Text item on the
Edit menu.)
[p,table,stats] = anovan(y,group,param,val) returns a stats
structure that you can use to perform a follow-up multiple comparison
test with the multcompare function. See “The Stats Structure” on page
18-30The Stats Structure for more information.
[p,table,stats,terms] = anovan(y,group,param,val) returns the
main and interaction terms used in the ANOVA computations. The
terms are encoded in the output matrix terms using the same format
described above for input 'model'. When you specify 'model' itself in
this matrix format, the matrix returned in terms is identical.
18-24
anovan
Model Type
This section explains how to use the argument 'model' with the syntax:
[...] = anovan(y,group,'model',modeltype)
The argument modeltype, which specifies the type of model the function
uses, can be any one of the following:
• 'linear' — The default 'linear' model computes only the p-values
for the null hypotheses on the N main effects.
• 'interaction' — The 'interaction' model computes the p-values
⎛N⎞
for null hypotheses on the N main effects and the ⎜ ⎟ two-factor
interactions. ⎝2⎠
• 'full' — The 'full' model computes the p-values for null
hypotheses on the N main effects and interactions at all levels.
• An integer — For an integer value of modeltype, k (k ≤ N),
anovan computes all interaction levels through the kth level. For
example, the value 3 means main effects plus two- and three-factor
interactions. The values k = 1 and k = 2 are equivalent to the
'linear' and 'interaction' specifications, respectively, while the
value k = N is equivalent to the 'full' specification.
• A matrix of term definitions having the same form as the input to the
x2fx function. All entries must be 0 or 1 (no higher powers).
For more precise control over the main and interaction terms that
anovan computes, modeltype can specify a matrix containing one row
for each main or interaction term to include in the ANOVA model. Each
row defines one term using a vector of N zeros and ones. The table
below illustrates the coding for a 3-factor ANOVA.
Matrix Row ANOVA Term
[1 0 0] Main term A
[0 1 0] Main term B
18-25
anovan
Matrix Row ANOVA Term
[0 0 1] Main term C
[1 1 0] Interaction term AB
[1 0 1] Interaction term AC
[0 1 1] Interaction term BC
[1 1 1] Interaction term ABC
For example, if modeltype is the matrix [0 1 0;0 0 1;0 1 1], the
output vector p contains the p-values for the null hypotheses on the
main effects B and C and the interaction effect BC, in that order. A
simple way to generate the modeltype matrix is to modify the terms
output, which codes the terms in the current model using the format
described above. If anovan returns [0 1 0;0 0 1;0 1 1] for terms, for
example, and there is no significant result for interaction BC, you can
recompute the ANOVA on just the main effects B and C by specifying
[0 1 0;0 0 1] for modeltype.
Sum of Squares
This section explains how to use the argument 'sstype' with the
syntax:
[...] = anovan(y,group,'sstype',type)
This syntax computes the ANOVA using the type of sum of squares
specified by type, which can be 1, 2, 3, or h. While the numbers 1 – 3
designate Type 1, Type 2, or Type 3 sum of squares, respectively, h
represents a hierarchical model similar to type 2, but with continuous
as well as categorical factors used to determine the hierarchy of
terms. The default value is 3. For a model containing main effects
but no interactions, the value of type only influences computations
on unbalanced data.
The sum of squares for any term is determined by comparing two
models. The Type 1 sum of squares for a term is the reduction in
residual sum of squares obtained by adding that term to a fit that
already includes the terms listed before it. The Type 2 sum of squares is
18-26
anovan
the reduction in residual sum of squares obtained by adding that term
to a model consisting of all other terms that do not contain the term
in question. The Type 3 sum of squares is the reduction in residual
sum of squares obtained by adding that term to a model containing all
other terms, but with their effects constrained to obey the usual “sigma
restrictions” that make models estimable.
Suppose you are fitting a model with two factors and their interaction,
and that the terms appear in the order A, B, AB. Let R(·) represent the
residual sum of squares for a model, so for example R(A, B, AB) is the
residual sum of squares fitting the whole model, R(A) is the residual
sum of squares fitting just the main effect of A, and R(1) is the residual
sum of squares fitting just the mean. The three types of sums of squares
are as follows:
Type 2 Sum of Type 3 Sum of
Term Type 1 Sum of Squares Squares Squares
A R(1) – R(A) R(B) – R(A, B) R(B, AB) – R(A, B, AB)
B R(A) – R(A, B) R(A) – R(A, B) R(A, AB) – R(A, B, AB)
AB R(A, B) – R(A, B, AB) R(A, B) – R(A, B, AB) R(A, B) – R(A, B, AB)
The models for Type 3 sum of squares have sigma restrictions imposed.
This means, for example, that in fitting R(B, AB), the array of AB
effects is constrained to sum to 0 over A for each value of B, and over B
for each value of A.
Example of Three-Way ANOVA
As an example of three-way ANOVA, consider the vector y and group
inputs below.
y = [52.7 57.5 45.9 44.5 53.0 57.0 45.9 44.0]';
g1 = [1 2 1 2 1 2 1 2];
g2 = {'hi';'hi';'lo';'lo';'hi';'hi';'lo';'lo'};
g3 = {'may';'may';'may';'may';'june';'june';'june';'june'};
18-27
anovan
This defines a three-way ANOVA with two levels of each factor. Every
observation in y is identified by a combination of factor levels. If the
factors are A, B, and C, then observation y(1) is associated with
• Level 1 of factor A
• Level 'hi' of factor B
• Level 'may' of factor C
Similarly, observation y(6) is associated with
• Level 2 of factor A
• Level 'hi' of factor B
• Level 'june' of factor C
To compute the ANOVA, enter
p = anovan(y,{g1 g2 g3})
p =
0.4174
0.0028
0.9140
Output vector p contains p-values for the null hypotheses on the N main
effects. Element p(1) contains the p value for the null hypotheses,
H0A, that samples at all levels of factor A are drawn from the same
population; element p(2) contains the p value for the null hypotheses,
H0B, that samples at all levels of factor B are drawn from the same
population; and so on.
If any p value is near zero, this casts doubt on the associated null
hypothesis. For example, a sufficiently small p value for H0A suggests
that at least one A-sample mean is significantly different from the other
A-sample means; that is, there is a main effect due to factor A. You
need to choose a bound for the p value to determine whether a result is
statistically significant. It is common to declare a result significant if
the p value is less than 0.05 or 0.01.
18-28
anovan
anovan also displays a figure showing the standard ANOVA table,
which by default divides the variability of the data in x into
• The variability due to differences between the levels of each factor
accounted for in the model (one row for each factor)
• The remaining variability not explained by any systematic source
The ANOVA table has six columns:
• The first shows the source of the variability.
• The second shows the sum of squares (SS) due to each source.
• The third shows the degrees of freedom (df) associated with each
source.
• The fourth shows the mean squares (MS), which is the ratio SS/df.
• The fifth shows the F statistics, which are the ratios of the mean
squares.
• The sixth shows the p-values for the F statistics.
The table is shown in the following figure:
Two-Factor Interactions
By default, anovan computes p-values just for the three main effects.
To also compute p-values for the two-factor interactions, X1*X2, X1*X3,
18-29
anovan
and X2*X3, add the name/value pair 'model', 'interaction' as input
arguments.
p = anovan(y,{g1 g2 g3},'model','interaction')
p =
0.0347
0.0048
0.2578
0.0158
0.1444
0.5000
The first three entries of p are the p-values for the main effects. The
last three entries are the p-values for the two-factor interactions. You
can determine the order in which the two-factor interactions occur from
the ANOVAN table shown in the following figure.
The Stats Structure
The anovan test evaluates the hypothesis that the different levels of a
factor (or more generally, a term) have the same effect, against the
alternative that they do not all have the same effect. Sometimes it is
preferable to perform a test to determine which pairs of levels are
significantly different, and which are not. Use the multcompare function
to perform such tests by supplying the stats structure as input.
18-30
anovan
The stats structure contains the fields listed below, in addition to a
number of other fields required for doing multiple comparisons using
the multcompare function:
Field Description
coeffs Estimated coefficients
coeffnames Name of term for each coefficient
vars Matrix of grouping variable values for each term
resid Residuals from the fitted model
The stats structure also contains the following fields if there are
random effects:
Field Description
ems Expected mean squares
denom Denominator definition
rtnames Names of random terms
varest Variance component estimates (one per random term)
varci Confidence intervals for variance components
Examples “Example: Two-Way ANOVA” on page 8-10 shows how to use anova2 to
analyze the effects of two factors on a response in a balanced design.
For a design that is not balanced, use anovan instead.
The data in carbig.mat gives measurements on 406 cars. Use anonvan
to study how the mileage depends on where and when the cars were
made:
load carbig
p = anovan(MPG,{org when},'model',2,'sstype',3,...
'varnames',{'Origin';'Mfg date'})
p =
18-31
anovan
0
0
0.3059
The p value for the interaction term is not small, indicating little
evidence that the effect of the year or manufacture (when) depends on
where the car was made (org). The linear effects of those two factors,
however, are significant.
References [1] Hogg, R. V., and J. Ledolter. Engineering Statistics. New York:
MacMillan, 1987.
See Also anova1 | anova2 | multcompare
How To • “Grouped Data” on page 2-34
18-32
ansaribradley
Purpose Ansari-Bradley test
Syntax h = ansaribradley(x,y)
h = ansaribradley(x,y,alpha)
h = ansaribradley(x,y,alpha,tail)
[h,p] = ansaribradley(...)
[h,p,stats] = ansaribradley(...)
[...] = ansaribradley(x,y,alpha,tail,exact)
[...] = ansaribradley(x,y,alpha,tail,exact,dim)
Description h = ansaribradley(x,y) performs an Ansari-Bradley test of the
hypothesis that two independent samples, in the vectors x and y, come
from the same distribution, against the alternative that they come
from distributions that have the same median and shape but different
dispersions (e.g. variances). The result is h = 0 if the null hypothesis of
identical distributions cannot be rejected at the 5% significance level,
or h = 1 if the null hypothesis can be rejected at the 5% level. x and y
can have different lengths.
x and y can also be matrices or N-dimensional arrays. For matrices,
ansaribradley performs separate tests along each column, and returns
a vector of results. x and y must have the same number of columns.
For N-dimensional arrays, ansaribradley works along the first
nonsingleton dimension. x and y must have the same size along all
the remaining dimensions.
h = ansaribradley(x,y,alpha) performs the test at the significance
level (100*alpha), where alpha is a scalar.
h = ansaribradley(x,y,alpha,tail) performs the test against the
alternative hypothesis specified by the string tail. tail is one of:
• 'both' — Two-tailed test (dispersion parameters are not equal)
• 'right' — Right-tailed test (dispersion of X is greater than
dispersion of Y)
• 'left' — Left-tailed test (dispersion of X is less than dispersion of Y)
18-33
ansaribradley
[h,p] = ansaribradley(...) returns the p value, i.e., the probability
of observing the given result, or one more extreme, by chance if the
null hypothesis is true. Small values of p cast doubt on the validity of
the null hypothesis.
[h,p,stats] = ansaribradley(...) returns a structure stats with
the following fields:
• W — Value of the test statistic W, which is the sum of the
Ansari-Bradley ranks for the X sample
• Wstar — Approximate normal statistic W*
[...] = ansaribradley(x,y,alpha,tail,exact) computes p using
an exact calculation of the distribution of W with exact = 'on'. This
can be time-consuming for large samples. exact = 'off' computes p
using a normal approximation for the distribution of W*. The default if
exact is empty is to use the exact calculation if N, the total number of
rows in x and y, is 25 or less, and to use the normal approximation if N
> 25. Pass in [] for alpha and tail to use their default values while
specifying a value for exact. Note that N is computed before any NaN
values (representing missing data) are removed.
[...] = ansaribradley(x,y,alpha,tail,exact,dim) works along
dimension dim of x and y.
The Ansari-Bradley test is a nonparametric alternative to the
two-sample F test of equal variances. It does not require the assumption
that x and y come from normal distributions. The dispersion of a
distribution is generally measured by its variance or standard deviation,
but the Ansari-Bradley test can be used with samples from distributions
that do not have finite variances.
The theory behind the Ansari-Bradley test requires that the groups
have equal medians. Under that assumption and if the distributions
in each group are continuous and identical, the test does not depend
on the distributions in each group. If the groups do not have the
same medians, the results may be misleading. Ansari and Bradley
recommend subtracting the median in that case, but the distribution of
18-34
ansaribradley
the resulting test, under the null hypothesis, is no longer independent
of the common distribution of x and y. If you want to perform the tests
with medians subtracted, you should subtract the medians from x and
y before calling ansaribradley.
Examples Is the dispersion significantly different for two model years?
load carsmall
[h,p,stats] = ansaribradley(MPG(Model_Year==82),MPG(Model_Year==76))
h =
0
p =
0.8426
stats =
W: 526.9000
Wstar: 0.1986
See Also vartest | vartestn | ttest2
18-35
aoctool
Purpose Interactive analysis of covariance
Syntax aoctool(x,y,group)
aoctool(x,y,group,alpha)
aoctool(x,y,group,alpha,xname,yname,gname)
aoctool(x,y,group,alpha,xname,yname,gname,displayopt)
aoctool(x,y,group,alpha,xname,yname,gname,displayopt,model)
h = aoctool(...)
[h,atab,ctab] = aoctool(...)
[h,atab,ctab,stats] = aoctool(...)
Description aoctool(x,y,group) fits a separate line to the column vectors, x and y,
for each group defined by the values in the array group. group may be
a categorical variable, vector, character array, or cell array of strings.
(See “Grouped Data” on page 2-34.) These types of models are known
as one-way analysis of covariance (ANOCOVA) models. The output
consists of three figures:
• An interactive graph of the data and prediction curves
• An ANOVA table
• A table of parameter estimates
You can use the figures to change models and to test different parts
of the model. More information about interactive use of the aoctool
function appears in “Analysis of Covariance Tool” on page 8-27.
aoctool(x,y,group,alpha) determines the confidence levels of the
prediction intervals. The confidence level is 100(1-alpha)%. The
default value of alpha is 0.05.
aoctool(x,y,group,alpha,xname,yname,gname) specifies the name
to use for the x, y, and g variables in the graph and tables. If you
enter simple variable names for the x, y, and g arguments, the aoctool
function uses those names. If you enter an expression for one of these
arguments, you can specify a name to use in place of that expression by
supplying these arguments. For example, if you enter m(:,2) as the x
argument, you might choose to enter 'Col 2' as the xname argument.
18-36
aoctool
aoctool(x,y,group,alpha,xname,yname,gname,displayopt) enables
the graph and table displays when displayopt is 'on' (default) and
suppresses those displays when displayopt is 'off'.
aoctool(x,y,group,alpha,xname,yname,gname,displayopt,model)
specifies the initial model to fit. The value of model can be any of the
following:
• 'same mean' — Fit a single mean, ignoring grouping
• 'separate means' — Fit a separate mean to each group
• 'same line' — Fit a single line, ignoring grouping
• 'parallel lines' — Fit a separate line to each group, but constrain
the lines to be parallel
• 'separate lines' — Fit a separate line to each group, with no
constraints
h = aoctool(...) returns a vector of handles to the line objects in
the plot.
[h,atab,ctab] = aoctool(...) returns cell arrays containing the
entries in ANOVA table (atab) and the table of coefficient estimates
(ctab). (You can copy a text version of either table to the clipboard by
using the Copy Text item on the Edit menu.)
[h,atab,ctab,stats] = aoctool(...) returns a stats structure
that you can use to perform a follow-up multiple comparison test. The
ANOVA table output includes tests of the hypotheses that the slopes
or intercepts are all the same, against a general alternative that they
are not all the same. Sometimes it is preferable to perform a test to
determine which pairs of values are significantly different, and which
are not. You can use the multcompare function to perform such tests by
supplying the stats structure as input. You can test either the slopes,
the intercepts, or population marginal means (the heights of the curves
at the mean x value).
18-37
aoctool
Examples This example illustrates how to fit different models non-interactively.
After loading the smaller car data set and fitting a separate-slopes
model, you can examine the coefficient estimates.
load carsmall
[h,a,c,s] = aoctool(Weight,MPG,Model_Year,0.05,...
'','','','off','separate lines');
c(:,1:2)
ans =
'Term' 'Estimate'
'Intercept' [45.97983716833132]
' 70' [-8.58050531454973]
' 76' [-3.89017396094922]
' 82' [12.47067927549897]
'Slope' [-0.00780212907455]
' 70' [ 0.00195840368824]
' 76' [ 0.00113831038418]
' 82' [-0.00309671407243]
Roughly speaking, the lines relating MPG to Weight have an intercept
close to 45.98 and a slope close to -0.0078. Each group’s coefficients are
offset from these values somewhat. For instance, the intercept for the
cars made in 1970 is 45.98-8.58 = 37.40.
Next, try a fit using parallel lines. (The ANOVA table shows that the
parallel-lines fit is significantly worse than the separate-lines fit.)
[h,a,c,s] = aoctool(Weight,MPG,Model_Year,0.05,...
'','','','off','parallel lines');
c(:,1:2)
ans =
'Term' 'Estimate'
'Intercept' [43.38984085130596]
' 70' [-3.27948192983761]
' 76' [-1.35036234809006]
18-38
aoctool
' 82' [ 4.62984427792768]
'Slope' [-0.00664751826198]
Again, there are different intercepts for each group, but this time the
slopes are constrained to be the same.
See Also anova1 | multcompare | polytool
How To • “Grouped Data” on page 2-34
18-39
TreeBagger.append
Purpose Append new trees to ensemble
Syntax B = append(B,other)
Description B = append(B,other) appends the trees from the other ensemble to
those in B. This method checks for consistency of the X and Y properties
of the two ensembles, as well as consistency of their compact objects and
out-of-bag indices, before appending the trees. The output ensemble B
takes training parameters such as FBoot, Prior, Cost, and other from
the B input. There is no attempt to check if these training parameters
are consistent between the two objects.
See Also CompactTreeBagger.combine
18-40
ProbDistKernel.BandWidth property
Purpose Read-only value specifying bandwidth of kernel smoothing function
for ProbDistKernel object
Description BandWidth is a read-only property of the ProbDistKernel class.
BandWidth is a value specifying the width of the kernel smoothing
function used to compute a nonparametric estimate of the probability
distribution when creating a ProbDistKernel object.
Values For a distribution specified to cover only the positive numbers or only
a finite interval, the data are transformed before the kernel density is
applied, and the bandwidth is on the scale of the transformed data.
Use this information to view and compare the width of the kernel
smoothing function used to create distributions.
See Also ksdensity
18-41
barttest
Purpose Bartlett’s test
Syntax ndim = barttest(X,alpha)
[ndim,prob,chisquare] = barttest(X,alpha)
Description ndim = barttest(X,alpha) returns the number of dimensions
necessary to explain the nonrandom variation in the data matrix X,
using the significance probability alpha. The dimension is determined
by a series of hypothesis tests. The test for ndim=1 tests the hypothesis
that the variances of the data values along each principal component
are equal, the test for ndim=2 tests the hypothesis that the variances
along the second through last components are equal, and so on.
[ndim,prob,chisquare] = barttest(X,alpha) returns the number of
dimensions, the significance values for the hypothesis tests, and the χ2
values associated with the tests.
Examples X = mvnrnd([0 0],[1 0.99; 0.99 1],20);
X(:,3:4) = mvnrnd([0 0],[1 0.99; 0.99 1],20);
X(:,5:6) = mvnrnd([0 0],[1 0.99; 0.99 1],20);
[ndim, prob] = barttest(X,0.05)
ndim =
3
prob =
0
0
0
0.5081
0.6618
See Also princomp | pcacov | pcares
18-42
bbdesign
Purpose Box-Behnken design
Syntax dBB = bbdesign(n)
[dBB,blocks] = bbdesign(n)
[...] = bbdesign(n,param,val)
Description dBB = bbdesign(n) generates a Box-Behnken design for n factors. n
must be an integer 3 or larger. The output matrix dBB is m-by-n, where
m is the number of runs in the design. Each row represents one run,
with settings for all factors represented in the columns. Factor values
are normalized so that the cube points take values between -1 and 1.
[dBB,blocks] = bbdesign(n) requests a blocked design. The output
blocks is an m-by-1 vector of block numbers for each run. Blocks
indicate runs that are to be measured under similar conditions
to minimize the effect of inter-block differences on the parameter
estimates.
[...] = bbdesign(n,param,val) specifies one or more optional
parameter/value pairs for the design. The following table lists valid
parameter/value pairs.
Parameter Description Values
'center' Number of Integer. The default depends on
center points. n.
'blocksize' Maximum Integer. The default is Inf.
number of
points per block.
Examples The following creates a 3-factor Box-Behnken design:
dBB = bbdesign(3)
dBB =
-1 -1 0
-1 1 0
1 -1 0
18-43
bbdesign
1 1 0
-1 0 -1
-1 0 1
1 0 -1
1 0 1
0 -1 -1
0 -1 1
0 1 -1
0 1 1
0 0 0
0 0 0
0 0 0
The center point is run 3 times to allow for a more uniform estimate of
the prediction variance over the entire design space.
Visualize the design as follows:
plot3(dBB(:,1),dBB(:,2),dBB(:,3),'ro',...
'MarkerFaceColor','b')
X = [1 -1 -1 -1 1 -1 -1 -1 1 1 -1 -1; ...
1 1 1 -1 1 1 1 -1 1 1 -1 -1];
Y = [-1 -1 1 -1 -1 -1 1 -1 1 -1 1 -1; ...
1 -1 1 1 1 -1 1 1 1 -1 1 -1];
Z = [1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1; ...
1 1 1 1 -1 -1 -1 -1 1 1 1 1];
line(X,Y,Z,'Color','b')
axis square equal
18-44
bbdesign
See Also ccdesign
18-45
betacdf
Purpose Beta cumulative distribution function
Syntax p = betacdf(X,A,B)
Description p = betacdf(X,A,B) returns the beta cdf at each of the values in X
using the corresponding parameters in A and B. X, A, and B can be
vectors, matrices, or multidimensional arrays that all have the same
size. A scalar input is expanded to a constant array with the same
dimensions as the other inputs. The parameters in A and B must all be
positive, and the values in X must lie on the interval [0,1].
The beta cdf for a given value x and given pair of parameters a and b is
x
1
p = F ( x | a, b) = t a−1 (1 − t)b−1 dt
B(a, b) ∫
0
where B( · ) is the Beta function.
Examples x = 0.1:0.2:0.9;
a = 2;
b = 2;
p = betacdf(x,a,b)
p =
0.0280 0.2160 0.5000 0.7840 0.9720
a = [1 2 3];
p = betacdf(0.5,a,a)
p =
0.5000 0.5000 0.5000
See Also cdf | betapdf | betainv | betastat | betalike | betarnd | betafit
How To • “Beta Distribution” on page B-4
18-46
betafit
Purpose Beta parameter estimates
Syntax phat = betafit(data)
[phat,pci] = betafit(data,alpha)
Description phat = betafit(data) computes the maximum likelihood estimates
of the beta distribution parameters a and b from the data in the vector
data and returns a column vector containing the a and b estimates,
where the beta cdf is given by
x
1
t a−1 (1 − t)b−1 dt
B(a, b) ∫
F ( x | a, b) =
0
and B( · ) is the Beta function. The elements of data must lie in the
open interval (0, 1), where the beta distribution is defined. However,
it is sometimes also necessary to fit a beta distribution to data that
include exact zeros or ones. For such data, the beta likelihood function
is unbounded, and standard maximum likelihood estimation is not
possible. In that case, betafit maximizes a modified likelihood that
incorporates the zeros or ones by treating them as if they were values
that have been left-censored at sqrt(realmin) or right-censored at
1-eps/2, respectively.
[phat,pci] = betafit(data,alpha) returns confidence intervals on
the a and b parameters in the 2-by-2 matrix pci. The first column of the
matrix contains the lower and upper confidence bounds for parameter
a, and the second column contains the confidence bounds for parameter
b. The optional input argument alpha is a value in the range [0, 1]
specifying the width of the confidence intervals. By default, alpha is
0.05, which corresponds to 95% confidence intervals. The confidence
intervals are based on a normal approximation for the distribution of
the logs of the parameter estimates.
Examples This example generates 100 beta distributed observations. The true
a and b parameters are 4 and 3, respectively. Compare these to the
18-47
betafit
values returned in p by the beta fit. Note that the columns of ci both
bracket the true parameters.
data = betarnd(4,3,100,1);
[p,ci] = betafit(data,0.01)
p =
5.5328 3.8097
ci =
3.6538 2.6197
8.3781 5.5402
References [1] Hahn, Gerald J., and S. S. Shapiro. Statistical Models in
Engineering. Hoboken, NJ: John Wiley & Sons, Inc., 1994, p. 95.
See Also mle | betapdf | betainv | betastat | betalike | betarnd | betacdf
How To • “Beta Distribution” on page B-4
18-48
betainv
Purpose Beta inverse cumulative distribution function
Syntax X = betainv(P,A,B)
Description X = betainv(P,A,B) computes the inverse of the beta cdf with
parameters specified by A and B for the corresponding probabilities in P.
P, A, and B can be vectors, matrices, or multidimensional arrays that are
all the same size. A scalar input is expanded to a constant array with
the same dimensions as the other inputs. The parameters in A and B
must all be positive, and the values in P must lie on the interval [0, 1].
The inverse beta cdf for a given probability p and a given pair of
parameters a and b is
x = F −1 ( p| a, b) = { x : F ( x | a, b) = p}
where
x
1
t a−1 (1 − t)b−1 dt
B(a, b) ∫
p = F ( x | a, b) =
0
and B( · ) is the Beta function. Each element of output X is the value
whose cumulative probability under the beta cdf defined by the
corresponding parameters in A and B is specified by the corresponding
value in P.
Algorithm The betainv function uses Newton’s method with modifications to
constrain steps to the allowable range for x, i.e., [0 1].
Examples p = [0.01 0.5 0.99];
x = betainv(p,10,5)
x =
0.3726 0.6742 0.8981
According to this result, for a beta cdf with a = 10 and b = 5, a value
less than or equal to 0.3726 occurs with probability 0.01. Similarly,
18-49
betainv
values less than or equal to 0.6742 and 0.8981 occur with respective
probabilities 0.5 and 0.99.
See Also icdf | betapdf | betafit | betainv | betastat | betalike | betarnd
| betacdf
How To • “Beta Distribution” on page B-4
18-50
betalike
Purpose Beta negative log-likelihood
Syntax nlogL = betalike(params,data)
[nlogL,AVAR] = betalike(params,data)
Description nlogL = betalike(params,data) returns the negative of the beta
log-likelihood function for the beta parameters a and b specified in
vector params and the observations specified in the column vector data.
The elements of data must lie in the open interval (0, 1), where the beta
distribution is defined. However, it is sometimes also necessary to fit a
beta distribution to data that include exact zeros or ones. For such data,
the beta likelihood function is unbounded, and standard maximum
likelihood estimation is not possible. In that case, betalike computes a
modified likelihood that incorporates the zeros or ones by treating them
as if they were values that have been left-censored at sqrt(realmin) or
right-censored at 1-eps/2, respectively.
[nlogL,AVAR] = betalike(params,data) also returns AVAR, which is
the asymptotic variance-covariance matrix of the parameter estimates
if the values in params are the maximum likelihood estimates. AVAR is
the inverse of Fisher’s information matrix. The diagonal elements of
AVAR are the asymptotic variances of their respective parameters.
betalike is a utility function for maximum likelihood estimation of
the beta distribution. The likelihood assumes that all the elements in
the data sample are mutually independent. Since betalike returns
the negative beta log-likelihood function, minimizing betalike using
fminsearch is the same as maximizing the likelihood.
Examples This example continues the betafit example, which calculates
estimates of the beta parameters for some randomly generated beta
distributed data.
r = betarnd(4,3,100,1);
[nlogl,AVAR] = betalike(betafit(r),r)
nlogl =
18-51
betalike
-27.5996
AVAR =
0.2783 0.1316
0.1316 0.0867
See Also betapdf | betafit | betainv | betastat | betarnd | betacdf
How To • “Beta Distribution” on page B-4
18-52
betapdf
Purpose Beta probability density function
Syntax Y = betapdf(X,A,B)
Description Y = betapdf(X,A,B) computes the beta pdf at each of the values in
X using the corresponding parameters in A and B. X, A, and B can be
vectors, matrices, or multidimensional arrays that all have the same
size. A scalar input is expanded to a constant array with the same
dimensions of the other inputs. The parameters in A and B must all be
positive, and the values in X must lie on the interval [0, 1].
The beta probability density function for a given value x and given pair
of parameters a and b is
1
y = f ( x | a, b) = x a−1 (1 − x)b−1 I(0,1) ( x)
B(a, b)
where B( · ) is the Beta function. The indicator function I(0,1) ( x)
ensures that only values of x in the range (0 1) have nonzero probability.
The uniform distribution on (0 1) is a degenerate case of the beta pdf
where a = 1 and b = 1.
A likelihood function is the pdf viewed as a function of the parameters.
Maximum likelihood estimators (MLEs) are the values of the
parameters that maximize the likelihood function for a fixed value of x.
Examples a = [0.5 1; 2 4]
a =
0.5000 1.0000
2.0000 4.0000
y = betapdf(0.5,a,a)
y =
0.6366 1.0000
1.5000 2.1875
See Also pdf | betafit | betainv | betastat | betalike | betarnd | betacdf
18-53
betapdf
How To • “Beta Distribution” on page B-4
18-54
betarnd
Purpose Beta random numbers
Syntax R = betarnd(A,B)
R = betarnd(A,B,v)
R = betarnd(A,B,m,n)
R = betarnd(A,B,m,n,o,...)
Description R = betarnd(A,B) generates random numbers from the beta
distribution with parameters specified by A and B. A and B can be
vectors, matrices, or multidimensional arrays that have the same size,
which is also the size of R. A scalar input for A or B is expanded to a
constant array with the same dimensions as the other input.
R = betarnd(A,B,v) generates an array R of size v containing random
numbers from the beta distribution with parameters A and B, where v is
a row vector. If v is a 1-by-2 vector, R is a matrix with v(1) rows and
v(2) columns. If v is 1-by-n, R is an n-dimensional array.
R = betarnd(A,B,m,n) generates an m-by-n matrix containing random
numbers from the beta distribution with parameters A and B.
R = betarnd(A,B,m,n,o,...) generates an m-by-n-by-o-by-...
multidimensional array containing random numbers from the beta
distribution with parameters A and B.
Examples a = [1 1;2 2];
b = [1 2;1 2];
r = betarnd(a,b)
r =
0.6987 0.6139
0.9102 0.8067
r = betarnd(10,10,[1 5])
r =
0.5974 0.4777 0.5538 0.5465 0.6327
r = betarnd(4,2,2,3)
18-55
betarnd
r =
0.3943 0.6101 0.5768
0.5990 0.2760 0.5474
See Also random | betapdf | betafit | betainv | betastat | betalike |
betacdf
How To • “Beta Distribution” on page B-4
18-56
betastat
Purpose Beta mean and variance
Syntax [M,V] = betastat(A,B)
Description [M,V] = betastat(A,B), with A>0 and B>0, returns the mean of and
variance for the beta distribution with parameters specified by A and
B. A and B can be vectors, matrices, or multidimensional arrays that
have the same size, which is also the size of M and V. A scalar input
for A or B is expanded to a constant array with the same dimensions
as the other input.
The mean of the beta distribution with parameters a and b is a / (a + b)
and the variance is
ab
(a + b + 1)(a + b)2
Examples If parameters a and b are equal, the mean is 1/2.
a = 1:6;
[m,v] = betastat(a,a)
m =
0.5000 0.5000 0.5000 0.5000 0.5000 0.5000
v =
0.0833 0.0500 0.0357 0.0278 0.0227 0.0192
See Also betapdf | betafit | betainv | betalike | betarnd | betacdf
How To • “Beta Distribution” on page B-4
18-57
gmdistribution.BIC property
Purpose Bayes Information Criterion
Description The Bayes Information Criterion: 2*NlogL+m*log(n), where n is the
number of observations and m is the number of estimated parameters.
18-58
binocdf
Purpose Binomial cumulative distribution function
Syntax Y = binocdf(X,N,P)
Description Y = binocdf(X,N,P) computes a binomial cdf at each of the values
in X using the corresponding number of trials in N and probability of
success for each trial in P. X, N, and P can be vectors, matrices, or
multidimensional arrays that are all the same size. A scalar input is
expanded to a constant array with the same dimensions of the other
inputs. The values in N must all be positive integers, the values in X
must lie on the interval [0,N], and the values in P must lie on the
interval [0, 1].
The binomial cdf for a given value x and a given pair of parameters n
and p is
x
⎛ n⎞
y = F ( x | n, p) = ∑ ⎜⎝ i ⎟⎠ pi q(n−i) I(0,1,...,n) (i)
i =0
The result, y, is the probability of observing up to x successes in n
independent trials, where the probability of success in any given trial is
p. The indicator function I(0,1,...,n) (i) ensures that x only adopts values
of 0,1,...,n.
Examples If a baseball team plays 162 games in a season and has a 50-50 chance
of winning any game, then the probability of that team winning more
than 100 games in a season is:
1 - binocdf(100,162,0.5)
The result is 0.001 (i.e., 1-0.999). If a team wins 100 or more games
in a season, this result suggests that it is likely that the team’s true
probability of winning any game is greater than 0.5.
See Also cdf | binopdf | binoinv | binostat | binofit | binornd
18-59
binocdf
How To • “Binomial Distribution” on page B-7
18-60
binofit
Purpose Binomial parameter estimates
Syntax phat = binofit(x,n)
[phat,pci] = binofit(x,n)
[phat,pci] = binofit(x,n,alpha)
Description phat = binofit(x,n) returns a maximum likelihood estimate of the
probability of success in a given binomial trial based on the number of
successes, x, observed in n independent trials. If x = (x(1), x(2),
... x(k)) is a vector, binofit returns a vector of the same size as
x whose ith entry is the parameter estimate for x(i). All k estimates
are independent of each other. If n = (n(1), n(2), ..., n(k)) is a
vector of the same size as x, the binomial fit, binofit, returns a vector
whose ith entry is the parameter estimate based on the number of
successes x(i) in n(i) independent trials. A scalar value for x or n is
expanded to the same size as the other input.
[phat,pci] = binofit(x,n) returns the probability estimate,
phat, and the 95% confidence intervals, pci. binofit uses the
Clopper-Pearson method to calculate confidence intervals.
[phat,pci] = binofit(x,n,alpha) returns the 100(1 - alpha)%
confidence intervals. For example, alpha = 0.01 yields 99% confidence
intervals.
Note binofit behaves differently than other Statistics Toolbox
functions that compute parameter estimates, in that it returns
independent estimates for each entry of x. By comparison, expfit
returns a single parameter estimate based on all the entries of x.
Unlike most other distribution fitting functions, the binofit function
treats its input x vector as a collection of measurements from separate
samples. If you want to treat x as a single sample and compute a single
parameter estimate for it, you can use binofit(sum(x),sum(n)) when
n is a vector, and binofit(sum(X),N*length(X)) when n is a scalar.
18-61
binofit
Examples This example generates a binomial sample of 100 elements, where the
probability of success in a given trial is 0.6, and then estimates this
probability from the outcomes in the sample.
r = binornd(100,0.6);
[phat,pci] = binofit(r,100)
phat =
0.5800
pci =
0.4771 0.6780
The 95% confidence interval, pci, contains the true value, 0.6.
References [1] Johnson, N. L., S. Kotz, and A. W. Kemp. Univariate Discrete
Distributions. Hoboken, NJ: Wiley-Interscience, 1993.
See Also mle | binopdf | binocdf | binoinv | binostat | binornd
How To • “Binomial Distribution” on page B-7
18-62
binoinv
Purpose Binomial inverse cumulative distribution function
Syntax X = binoinv(Y,N,P)
Description X = binoinv(Y,N,P) returns the smallest integer X such that the
binomial cdf evaluated at X is equal to or exceeds Y. You can think of
Y as the probability of observing X successes in N independent trials
where P is the probability of success in each trial. Each X is a positive
integer less than or equal to N.
Y, N, and P can be vectors, matrices, or multidimensional arrays that
all have the same size. A scalar input is expanded to a constant array
with the same dimensions as the other inputs. The parameters in N
must be positive integers, and the values in both P and Y must lie on
the interval [0 1].
Examples If a baseball team has a 50-50 chance of winning any game, what is a
reasonable range of games this team might win over a season of 162
games?
binoinv([0.05 0.95],162,0.5)
ans =
71 91
This result means that in 90% of baseball seasons, a .500 team should
win between 71 and 91 games.
See Also icdf | binopdf | binocdf | binofit | binostat | binornd
How To • “Binomial Distribution” on page B-7
18-63
binopdf
Purpose Binomial probability density function
Syntax Y = binopdf(X,N,P)
Description Y = binopdf(X,N,P) computes the binomial pdf at each of the values
in X using the corresponding number of trials in N and probability of
success for each trial in P. Y, N, and P can be vectors, matrices, or
multidimensional arrays that all have the same size. A scalar input is
expanded to a constant array with the same dimensions of the other
inputs.
The parameters in N must be positive integers, and the values in P must
lie on the interval [0, 1].
The binomial probability density function for a given value x and given
pair of parameters n and p is
⎛ n⎞
y = f ( x | n, p) = ⎜ ⎟ px q(n− x) I(0,1,...,n) ( x)
⎝ x⎠
where q = 1 – p. The result, y, is the probability of observing x successes
in n independent trials, where the probability of success in any given
trial is p. The indicator function I(0,1,...,n)(x) ensures that x only adopts
values of 0, 1, ..., n.
Examples A Quality Assurance inspector tests 200 circuit boards a day. If 2% of
the boards have defects, what is the probability that the inspector will
find no defective boards on any given day?
binopdf(0,200,0.02)
ans =
0.0176
What is the most likely number of defective boards the inspector will
find?
defects=0:200;
y = binopdf(defects,200,.02);
18-64
binopdf
[x,i]=max(y);
defects(i)
ans =
4
See Also pdf | binoinv | binocdf | binofit | binostat | binornd
How To • “Binomial Distribution” on page B-7
18-65
binornd
Purpose Binomial random numbers
Syntax R = binornd(N,P)
R = binornd(N,P,v)
R = binornd(N,p,m,n)
Description R = binornd(N,P) generates random numbers from the binomial
distribution with parameters specified by the number of trials, N, and
probability of success for each trial, P. N and P can be vectors, matrices,
or multidimensional arrays that have the same size, which is also the
size of R. A scalar input for N or P is expanded to a constant array with
the same dimensions as the other input.
R = binornd(N,P,v) generates an array R of size v containing random
numbers from the binomial distribution with parameters N and P, where
v is a row vector. If v is a 1-by-2 vector, R is a matrix with v(1) rows
and v(2) columns. If v is 1-by-n, R is an n-dimensional array.
R = binornd(N,p,m,n) generates an m-by-n matrix containing random
numbers from the binomial distribution with parameters N and P.
Algorithm The binornd function uses the direct method using the definition of the
binomial distribution as a sum of Bernoulli random variables.
Examples n = 10:10:60;
r1 = binornd(n,1./n)
r1 =
2 1 0 1 1 2
r2 = binornd(n,1./n,[1 6])
r2 =
0 1 2 1 3 1
r3 = binornd(n,1./n,1,6)
r3 =
0 1 1 1 0 3
18-66
binornd
See Also random | binoinv | binocdf | binofit | binostat | binopdf
How To • “Binomial Distribution” on page B-7
18-67
binostat
Purpose Binomial mean and variance
Syntax [M,V] = binostat(N,P)
Description [M,V] = binostat(N,P) returns the mean of and variance for the
binomial distribution with parameters specified by the number of trials,
N, and probability of success for each trial, P. N and P can be vectors,
matrices, or multidimensional arrays that have the same size, which
is also the size of M and V. A scalar input for N or P is expanded to a
constant array with the same dimensions as the other input.
The mean of the binomial distribution with parameters n and p is np.
The variance is npq, where q = 1–p.
Examples n = logspace(1,5,5)
n =
10 100 1000 10000 100000
[m,v] = binostat(n,1./n)
m =
1 1 1 1 1
v =
0.9000 0.9900 0.9990 0.9999 1.0000
[m,v] = binostat(n,1/2)
m =
5 50 500 5000 50000
v =
1.0e+04 *
0.0003 0.0025 0.0250 0.2500 2.5000
See Also binoinv | binocdf | binofit | binornd | binopdf
How To • “Binomial Distribution” on page B-7
18-68
biplot
Purpose Biplot
Syntax biplot(coefs)
h = biplot(coefs,'Name',Value)
Description biplot(coefs) creates a biplot of the coefficients in the matrix coefs.
The biplot is 2-D if coefs has two columns or 3-D if it has three
columns. coefs usually contains principal component coefficients
created with princomp, pcacov, or factor loadings estimated with
factoran. The axes in the biplot represent the principal components or
latent factors (columns of coefs), and the observed variables (rows of
coefs) are represented as vectors.
A biplot allows you to visualize the magnitude and sign of each
variable’s contribution to the first two or three principal components,
and how each observation is represented in terms of those components.
biplot imposes a sign convention, forcing the element with largest
magnitude in each column of coefs to be positive. This flips some of the
vectors in coefs to the opposite direction, but often makes the plot easier
to read. Interpretation of the plot is unaffected, because changing the
sign of a coefficient vector does not change its meaning.
h = biplot(coefs,'Name',Value) specifies one or more name/value
input pairs and returns a column vector of handles to the graphics
objects created by biplot. The h contains, in order, handles
corresponding to variables (line handles, followed by marker handles,
followed by text handles), to observations (if present, marker handles
followed by text handles), and to the axis lines.
Input Argument Name/Value Pairs
Arguments
Scores
Plots both coefs and the scores in the matrix scores in the
biplot. scores usually contains principal component scores
created with princomp or factor scores estimated with factoran.
18-69
biplot
Each observation (row of scores) is represented as a point in the
biplot.
VarLabels
Labels each vector (variable) with the text in the character array
or cell array varlabels.
ObsLabels
Uses the text in the character array or cell array obslabels as
observation names when displaying data cursors.
Positive
• 'true' — restricts the biplot to the positive quadrant (in 2-D)
or octant (in 3-D).
• 'false' — makes the biplot over the range +/- max(coefs(:))
for all coordinates.
Default: false
PropertyName
Specifies optional property name/value pairs for all line graphics
objects created by biplot.
Examples Perform a principal component analysis of the data in carsmall.mat:
load carsmall
x = [Acceleration Displacement Horsepower MPG Weight];
x = x(all(~isnan(x),2),:);
[coefs,score] = princomp(zscore(x));
View the data and the original variables in the space of the first three
principal components:
18-70
biplot
vbls = {'Accel','Disp','HP','MPG','Wgt'};
biplot(coefs(:,1:3),'scores',score(:,1:3),...
'varlabels',vbls);
See Also factoran | nnmf | princomp | pcacov | rotatefactors
18-71
bootci
Purpose Bootstrap confidence interval
Syntax ci = bootci(nboot,bootfun,...)
ci = bootci(nboot,{bootfun,...},'alpha',alpha)
ci = bootci(nboot,{bootfun,...},...,'type',type)
ci = bootci(nboot,{bootfun,...},...,'type','student',
'nbootstd',nbootstd)
ci = bootci(nboot,{bootfun,...},...,'type','student','stderr',
stderr)
ci = bootci(nboot,{bootfun,...},...,'Weights',weights)
ci = bootci(nboot,{bootfun,...},...,'Options',options)
Description ci = bootci(nboot,bootfun,...) computes the 95% bootstrap
confidence interval of the statistic computed by the function bootfun.
nboot is a positive integer indicating the number of bootstrap samples
used in the computation. bootfun is a function handle specified with
@, and must return a scalar. The third and later input arguments to
bootci are data (scalars, column vectors, or matrices) that are used
to create inputs to bootfun. bootci creates each bootstrap sample
by sampling with replacement from the rows of the non-scalar data
arguments (these must have the same number of rows). Scalar data
are passed to bootfun unchanged.
If bootfun returns a scalar, ci is a vector containing the lower and
upper bounds of the confidence interval. If bootfun returns a vector of
length m, ci is an array of size 2-by-m, where ci(1,:) are lower bounds
and ci(2,:) are upper bounds. If bootfun returns an array of size
m-by-n-by-p-by-..., ci is an array of size 2-by-m-by-n-by-p-by-..., where
ci(1,:,:,:,...) is an array of lower bounds and ci(2,:,:,:,...) is
an array of upper bounds.
ci = bootci(nboot,{bootfun,...},'alpha',alpha) computes the
100*(1-alpha) bootstrap confidence interval of the statistic defined by
the function bootfun. bootfun and the data that bootci passes to it
are contained in a single cell array. alpha is a scalar between 0 and 1.
The default value of alpha is 0.05.
18-72
bootci
ci = bootci(nboot,{bootfun,...},...,'type',type) computes the
bootstrap confidence interval of the statistic defined by the function
bootfun. type is the confidence interval type, chosen from among the
following strings:
• 'norm' or 'normal' — Normal approximated interval with
bootstrapped bias and standard error.
• 'per' or 'percentile' — Basic percentile method.
• 'cper' or 'corrected percentile' — Bias corrected percentile
method.
• 'bca' — Bias corrected and accelerated percentile method. This
is the default.
• 'stud' or 'student' — Studentized confidence interval.
ci =
bootci(nboot,{bootfun,...},...,'type','student','nbootstd',nbootstd)
computes the studentized bootstrap confidence interval of the statistic
defined by the function bootfun. The standard error of the
bootstrap statistics is estimated using bootstrap, with nbootstd
bootstrap data samples. nbootstd is a positive integer value.
The default value of nbootstd is 100.
ci =
bootci(nboot,{bootfun,...},...,'type','student','stderr',stderr)
computes the studentized bootstrap confidence interval of statistics
defined by the function bootfun. The standard error of the bootstrap
statistics is evaluated by the function stderr. stderr is a function
handle. stderr takes the same arguments as bootfun and returns the
standard error of the statistic computed by bootfun.
ci = bootci(nboot,{bootfun,...},...,'Weights',weights)
specifies observation weights. weights must be a vector of non-negative
numbers with at least one positive element. The number of elements
in weights must be equal to the number of rows in non-scalar
input arguments to bootfun. To obtain one bootstrap replicate,
18-73
bootci
bootstrp samples N out of N with replacement using these weights as
multinomial sampling probabilities.
ci = bootci(nboot,{bootfun,...},...,'Options',options)
specifies options that govern the computation of bootstrap iterations.
One option requests that bootci perform bootstrap iterations using
multiple processors, if the Parallel Computing Toolbox is available. Two
options specify the random number streams to be used in bootstrap
resampling. This argument is a struct that you can create with a call to
statset. You can retrieve values of the individual fields with a call to
statget. Applicable statset parameters are:
• 'UseParallel' — If 'always' and if a matlabpool of the Parallel
Computing Toolbox is open, compute bootstrap iterations in parallel.
If the Parallel Computing Toolbox is not installed, or a matlabpool is
not open, computation occurs in serial mode. Default is 'never', or
serial computation.
• 'UseSubstreams' — If 'always', perform each bootstrap iteration
using a separate Substream of the random number generator (aka
Stream). This option is available only with RandStream types that
support Substreams. Default is 'never', or do not use a different
Substream for each bootstrap iteration.
• 'Streams' — An object of the RandStream class,or a cell array
of RandStream objects. Defaults is an empty cell array. If you do
not supply a value for this parameter, bootci uses the default
RandStream on each MATLAB executable to generate boostrap
samples. Otherwise, bootci selects bootstrap samples using the
supplied RandStream object(s).
If you select 'UseSubstreams', the Streams parameter, if present,
must be a scalar RandStream object.
If you do not select 'UseSubstreams', then the Streams parameter,
if present, must match the number of processors used for the
computation. For serial computation, the Streams parameter must
be a scalar.
18-74
bootci
If computation is distributed ('UseParallel' is 'always' and a
matlabpool is open), then the Streams parameter must be a cell
array of the same length as the matlabpool size. In this case, each
element of the cell array supplies the random number generator for
bootstrap sampling on one of the parallel workers.
Examples Compute the confidence interval for the capability index in statistical
process control:
y = normrnd(1,1,30,1); % Simulated process data
LSL = -3; USL = 3; % Process specifications
capable = @(x)(USL-LSL)./(6* std(x)); % Process capability
ci = bootci(2000,capable,y) % BCa confidence interval
ci =
0.8122
1.2657
sci = bootci(2000,{capable,y},'type','student') % Studentized ci
sci =
0.7739
1.2707
See Also bootstrp | jackknife | statget | statset | randsample | parfor
18-75
bootstrp
Purpose Bootstrap sampling
Syntax bootstat = bootstrp(nboot,bootfun,d1,...)
[bootstat,bootsam] = bootstrp(...)
bootstat = bootstrp(...,'Options',options)
Description bootstat = bootstrp(nboot,bootfun,d1,...) draws nboot
bootstrap data samples, computes statistics on each sample using
bootfun, and returns the results in the matrix bootstat. nboot must
be a positive integer. bootfun is a function handle specified with
@. Each row of bootstat contains the results of applying bootfun to
one bootstrap sample. If bootfun returns a matrix or array, then this
output is converted to a row vector for storage in bootstat.
The third and later input arguments (d1,...) are data (scalars, column
vectors, or matrices) used to create inputs to bootfun. bootstrp creates
each bootstrap sample by sampling with replacement from the rows of
the non-scalar data arguments (these must have the same number of
rows). bootfun accepts scalar data unchanged.
[bootstat,bootsam] = bootstrp(...) returns an n-by-nboot matrix
of bootstrap indices, bootsam. Each column in bootsam contains indices
of the values that were drawn from the original data sets to constitute
the corresponding bootstrap sample. For example, if d1,... each
contain 16 values, and nboot = 4, then bootsam is a 16-by-4 matrix.
The first column contains the indices of the 16 values drawn from
d1,..., for the first of the four bootstrap samples, the second column
contains the indices for the second of the four bootstrap samples, and so
on. (The bootstrap indices are the same for all input data sets.) To get
the output samples bootsam without applying a function, set bootfun
to empty ([]).
bootstat = bootstrp(...,'Options',options) specifies options that
govern the computation of bootstrap iterations. One option requests that
bootstrap iterations use multiple processors, if the Parallel Computing
Toolbox is available. Two options specify the random number streams
used in bootstrap resampling. This argument is a struct you can create
18-76
bootstrp
with a call to statset. You can retrieve values of the individual fields
with a call to statget. Applicable statset parameters are:
• 'Weights' — Observation weights. This must be a vector of
non-negative numbers with at least one positive element. The
number of elements in weights must be equal to the number of rows
in non-scalar input arguments to bootstrp. To obtain one bootstrap
replicate, bootstrp samples N out of N with replacement using these
weights as multinomial sampling probabilities.
• 'Options' — Options governing the computation of bootstrap
iterations. One option requests that bootstrap iterations be performed
using multiple processors, if the Parallel Computing Toolbox is
available. Two other options specify the random number streams to
be used in bootstrap resampling. This argument is a struct that can
be created by a call to statset, and values of the individual fields can
be retrieved by a call to statget. Applicable statset parameters are:
• 'UseParallel' — If 'always' and if a matlabpool of the Parallel
Computing Toolbox is open, compute bootstrap iterations in parallel.
If the Parallel Computing Toolbox is not installed, or a matlabpool is
not open, computation occurs in serial mode. Default is 'never', or
serial computation.
• 'UseSubstreams' — If 'always', perform each bootstrap iteration
using a separate Substream of the random number generator (aka
Stream). This option is available only with RandStream types that
support Substreams. Default is 'never', or do not use a different
Substream for each bootstrap iteration.
• 'Streams' — An object of the RandStream class,or a cell array
of RandStream objects. Defaults is an empty cell array. If you do
not supply a value for this parameter, bootci uses the default
RandStream on each MATLAB executable to generate boostrap
samples. Otherwise, bootci selects bootstrap samples using the
supplied RandStream object(s).
If you select 'UseSubstreams', the Streams parameter, if present,
must be a scalar RandStream object.
18-77
bootstrp
If you do not select 'UseSubstreams', then the Streams parameter,
if present, must match the number of processors used for the
computation. For serial computation, the Streams parameter must
be a scalar.
If computation is distributed ('UseParallel' is 'always' and a
matlabpool is open), then the Streams parameter must be a cell
array of the same length as the matlabpool size. In this case, each
element of the cell array supplies the random number generator for
bootstrap sampling on one of the parallel workers.
Examples Bootstrapping a Correlation Coefficient Standard Error
Load a data set containing the LSAT scores and law-school GPA for 15
students. These 15 data points are resampled to create 1000 different
data sets, and the correlation between the two variables is computed
for each data set.
load lawdata
[bootstat,bootsam] = bootstrp(1000,@corr,lsat,gpa);
Display the first 5 bootstrapped correlation coefficients.
bootstat(1:5,:)
ans =
0.6600
0.7969
0.5807
0.8766
0.9197
Display the indices of the data selected for the first 5 bootstrap samples.
bootsam(:,1:5)
ans =
9 8 15 11 15
14 7 6 7 14
4 6 10 3 11
3 10 11 9 2
18-78
bootstrp
15 4 13 4 14
9 4 5 2 10
8 5 4 3 13
1 9 1 15 11
10 8 6 12 3
1 4 5 2 8
1 1 10 6 2
3 10 15 10 8
14 6 10 3 8
13 12 1 2 4
12 6 4 9 8
hist(bootstat)
18-79
bootstrp
The histogram shows the variation of the correlation coefficient across
all the bootstrap samples. The sample minimum is positive, indicating
that the relationship between LSAT score and GPA is not accidental.
Finally, compute a bootstrap standard of error for the estimated
correlation coefficient.
se = std(bootstat)
se =
0.1327
Estimating the Density of Bootstrapped Statistic
Compute a sample of 100 bootstrapped means of random samples
taken from the vector Y, and plot an estimate of the density of these
bootstrapped means:
y = exprnd(5,100,1);
m = bootstrp(100,@mean,y);
[fi,xi] = ksdensity(m);
plot(xi,fi);
18-80
bootstrp
Bootstrapping More Than One Statistic
Compute a sample of 100 bootstrapped means and standard deviations
of random samples taken from the vector Y, and plot the bootstrap
estimate pairs:
y = exprnd(5,100,1);
stats = bootstrp(100,@(x)[mean(x) std(x)],y);
plot(stats(:,1),stats(:,2),'o')
18-81
bootstrp
Bootstrapping a Regression Model
Estimate the standard errors for a coefficient vector in a linear
regression by bootstrapping residuals:
load hald
x = [ones(size(heat)),ingredients];
y = heat;
b = regress(y,x);
yfit = x*b;
resid = y - yfit;
18-82
bootstrp
se = std(bootstrp(...
1000,@(bootr)regress(yfit+bootr,x),resid));
See Also hist | ksdensity | parfor | random | randsample | RandStream |
statget | statset
18-83
boxplot
Purpose Box plot
Syntax boxplot(X)
boxplot(X,G)
boxplot(axes,X,...)
boxplot(...,'Name',value)
Description boxplot(X) produces a box plot of the data in X. If X is a matrix, there is
one box per column; if X is a vector, there is just one box. On each box,
the central mark is the median, the edges of the box are the 25th and
75th percentiles, the whiskers extend to the most extreme data points
not considered outliers, and outliers are plotted individually.
boxplot(X,G) specifies one or more grouping variables G, producing
a separate box for each set of X values sharing the same G value or
values (see “Grouped Data” on page 2-34). Grouping variables must
have one row per element of X, or one row per column of X. Specify a
single grouping variable in G using a vector, a character array, a cell
array of strings, or a vector categorical array; specify multiple grouping
variables in G using a cell array of these variable types, such as {G1 G2
G3}, or by using a matrix. If multiple grouping variables are used, they
must all be the same length. Groups that contain a NaN value or an
empty string in a grouping variable are omitted, and are not counted in
the number of groups considered by other parameters.
By default, character and string grouping variables are sorted in the
order they initially appear in the data, categorical grouping variables
are sorted by the order of their levels, and numeric grouping variables
are sorted in numeric order. To control the order of groups, do one of
the following:
• Use categorical variables in G and specify the order of their levels.
• Use the 'grouporder' parameter described below.
• Pre-sort your data.
boxplot(axes,X,...) creates the plot in the axes with handle axes.
18-84
boxplot
boxplot(...,'Name',value) specifies one or more optional parameter
name/value pairs, as described in the following table. Specify Name in
single quotes.
Name Value
plotstyle • 'traditional' — Traditional box style.
This is the default.
• 'compact' — Box style designed for plots
with many groups. This style changes the
defaults for some other parameters, as
described in the following table.
boxstyle • 'outline' — Draws an unfilled box with
dashed whiskers. This is the default.
• 'filled' — Draws a narrow filled box with
lines for whiskers.
colorgroup One or more grouping variables, of the same
type as permitted for G, specifying that the
box color should change when the specified
variables change. The default is [] for no box
color change.
colors Colors for boxes, specified as a single color
(such as 'r' or [1 0 0]) or multiple colors
(such as 'rgbm' or a three-column matrix of
RGB values). The sequence is replicated or
truncated as required, so for example 'rb'
gives boxes that alternate in color. The default
when no 'colorgroup' is specified is to use
the same color scheme for all boxes. The
default when 'colorgroup' is specified is a
modified hsv colormap.
18-85
boxplot
Name Value
datalim A two-element vector containing lower and
upper limits, used by 'extrememode' to
determine which points are extreme. The
default is [-Inf Inf].
extrememode • 'clip' — Moves data outside the datalim
limits to the limit. This is the default.
• 'compress' — Evenly distributes data
outside the datalim limits in a region just
outside the limit, retaining the relative
order of the points.
A dotted line marks the limit if any points
are outside it, and two gray lines mark
the compression region if any points are
compressed. Values at +/–Inf can be clipped
or compressed, but NaN values still do not
appear on the plot. Box notches are drawn to
scale and may extend beyond the bounds if the
median is inside the limit; they are not drawn
if the median is outside the limits.
factordirection • 'data' — Arranges factors with the first
value next to the origin. This is the default.
• 'list' — Arranges factors left-to-right if on
the x axis or top-to-bottom if on the y axis.
• 'auto' — Uses 'data' for numeric
grouping variables and 'list' for strings.
18-86
boxplot
Name Value
fullfactors • 'off' — One group for each unique row of
G. This is the default.
• 'on' — Create a group for each possible
combination of group variable values,
including combinations that do not appear
in the data.
factorseparator Specifies which factors should have their
values separated by a grid line. The value
may be 'auto' or a vector of grouping variable
numbers. For example, [1 2] adds a separator
line when the first or second grouping variable
changes value. 'auto' is [] for one grouping
variable and [1] for two or more grouping
variables. The default is [].
factorgap Specifies an extra gap to leave between boxes
when the corresponding grouping factor
changes value, expressed as a percentage of
the width of the plot. For example, with [3 1],
the gap is 3% of the width of the plot between
groups with different values of the first
grouping variable, and 1% between groups
with the same value of the first grouping
variable but different values for the second.
'auto' specifies that boxplot should choose a
gap automatically. The default is [].
grouporder Order of groups for plotting, specified as a
cell array of strings. With multiple grouping
variables, separate values within each string
with a comma. Using categorical arrays as
grouping variables is an easier way to control
the order of the boxes. The default is [], which
does not reorder the boxes.
18-87
boxplot
Name Value
jitter Maximum distance d to displace outliers along
the factor axis by a uniform random amount, in
order to make duplicate points visible. A d of
1 makes the jitter regions just touch between
the closest adjacent groups. The default is 0.
labels A character array, cell array of strings, or
numeric vector of box labels. There may be
one label per group or one label per X value.
Multiple label variables may be specified via a
numeric matrix or a cell array containing any
of these types.
Tip To remove labels from a plot, use the
following command:
set(gca,'XTickLabel',{' '})
labelorientation • 'inline' — Rotates the labels to be vertical.
This is the default when plotstyle is
'compact'.
• 'horizontal' — Leaves the labels
horizontal. This is the default when
plotstyle has the default value of
'traditional'.
When the labels are on the y axis, both settings
leave the labels horizontal.
18-88
boxplot
Name Value
labelverbosity • 'all' — Displays every label. This is the
default.
• 'minor' — Displays a label for a factor only
when that factor has a different value from
the previous group.
• 'majorminor' — Displays a label for a
factor when that factor or any factor major
to it has a different value from the previous
group.
medianstyle • 'line' — Draws a line for the median. This
is the default.
• 'target' — Draws a black dot inside a
white circle for the median.
notch • 'on' — Draws comparison intervals using
notches when plotstyle is 'traditional',
or triangular markers when plotstyle is
'compact'.
• 'marker' — Draws comparison intervals
using triangular markers.
• 'off' — Omits notches. This is the default.
Two medians are significantly different at the
5% significance level if their intervals do not
overlap. Interval endpoints are the extremes
of the notches or the centers of the triangular
markers. When the sample size is small,
notches may extend beyond the end of the box.
orientation • 'vertical' — Plots X on the y axis. This is
the default.
• 'horizontal' — Plots X on the x axis.
18-89
boxplot
Name Value
outliersize Size of the marker used for outliers, in points.
The default is 6 (6/72 inch).
positions Box positions specified as a numeric vector
with one entry per group or X value. The
default is 1:numGroups, where numGroups is
the number of groups.
symbol Symbol and color to use for outliers, using
the same values as the LineSpec parameter
in plot. The default is 'r+'. If the symbol
is omitted then the outliers are invisible; if
the color is omitted then the outliers have the
same color as their corresponding box.
whisker Maximum whisker length w. The default is a
w of 1.5. Points are drawn as outliers if they
are larger than q3 + w(q3 – q1) or smaller than
q1 – w(q3 – q1), where q1 and q3 are the 25th
and 75th percentiles, respectively. The default
of 1.5 corresponds to approximately +/–2.7σ
and 99.3 coverage if the data are normally
distributed. The plotted whisker extends to
the adjacent value, which is the most extreme
data value that is not an outlier. Set whisker
to 0 to give no whiskers and to make every
point outside of q1 and q3 an outlier.
widths A scalar or vector of box widths for when
boxstyle is 'outline'. The default is half
of the minimum separation between boxes,
which is 0.5 when the positions argument
takes its default value. The list of values is
replicated or truncated as necessary.
When the plotstyle parameter takes the value 'compact', the
following default values for other parameters apply.
18-90
boxplot
Parameter Default when plotstyle is compact
boxstyle 'filled'
factorseparator 'auto'
factorgap 'auto'
jitter 0.5
labelorientation 'inline'
labelverbosity 'majorminor'
medianstyle 'target'
outliersize 4
symbol 'o'
You can see data values and group names using the data cursor in the
figure window. The cursor shows the original values of any points
affected by the datalim parameter. You can label the group to which
an outlier belongs using the gname function.
To modify graphics properties of a box plot component, use findobj
with the Tag property to find the component’s handle. Tag values for
box plot components depend on parameter settings, and are listed in
the table below.
Parameter Tag Values
Settings
All settings • 'Box'
• 'Outliers'
When plotstyle is • 'Median'
'traditional'
• 'Upper Whisker'
• 'Lower Whisker'
• 'Upper Adjacent Value'
• 'Lower Adjacent Value'
18-91
boxplot
Parameter Tag Values
Settings
When plotstyle is • 'Whisker'
'compact'
• 'MedianOuter'
• 'MedianInner'
When notch is • 'NotchLo'
'marker'
• 'NotchHi'
Examples Example 1
Create a box plot of car mileage, grouped by country:
load carsmall
boxplot(MPG,Origin)
18-92
boxplot
Example 2
Create notched box plots for two groups of sample data:
x1 = normrnd(5,1,100,1);
x2 = normrnd(6,1,100,1);
boxplot([x1,x2],'notch','on')
18-93
boxplot
The difference between the medians of the two groups is approximately
1. Since the notches in the box plot do not overlap, you can conclude,
with 95% confidence, that the true medians do differ.
The following figure shows the box plot for the same data with the
length of the whiskers specified as 1.0 times the interquartile range.
Points beyond the whiskers are displayed using +.
boxplot([x1,x2],'notch','on','whisker',1)
18-94
boxplot
Example 3
A plotstyle of 'compact' is useful for large numbers of groups:
X = randn(100,25);
subplot(2,1,1)
boxplot(X)
subplot(2,1,2)
boxplot(X,'plotstyle','compact')
18-95
boxplot
References [1] McGill, R., J. W. Tukey, and W. A. Larsen. “Variations of Boxplots.”
The American Statistician. Vol. 32, No. 1, 1978, pp. 12–16.
[2] Velleman, P.F., and D.C. Hoaglin. Applications, Basics, and
Computing of Exploratory Data Analysis. Pacific Grove, CA: Duxbury
Press, 1981.
[3] Nelson, L. S. “Evaluating Overlapping Confidence Intervals.”
Journal of Quality Technology. Vol. 21, 1989, pp. 140–141.
18-96
boxplot
See Also anova1 | axes_props | kruskalwallis | multcompare
How To • “Grouped Data” on page 2-34
18-97
piecewisedistribution.boundary
Purpose Piecewise distribution boundaries
Syntax [p,q] = boundary(obj)
[p,q] = boundary(obj,i)
Description [p,q] = boundary(obj) returns the boundary points between
segments of the piecewise distribution object, obj. p is a vector of
cumulative probabilities at each boundary. q is a vector of quantiles at
each boundary.
[p,q] = boundary(obj,i) returns p and q for the ith boundary.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
[p,q] = boundary(obj)
p =
0.1000
0.9000
q =
-1.7766
1.8432
See Also paretotails | cdf | icdf | nsegments
18-98
candexch
Purpose Candidate set row exchange
Syntax treatments = candexch(C,nruns)
treatments = candexch(C,nruns,'Name',value)
Description treatments = candexch(C,nruns) uses a row-exchange algorithm
to select treatments from a candidate design matrix C to produce a
D-optimal design with nruns runs. The columns of C represent model
terms evaluated at candidate treatments. treatments is a vector of
length nruns giving indices of the rows in C used in the D-optimal
design. The function selects a starting design at random.
treatments = candexch(C,nruns,'Name',value) specifies one or
more additional name/value pairs for the design. Valid parameters and
their values are listed in the following table. Specify Name in single
quotes.
Parameter Value
display Either 'on' or 'off' to control display of the
iteration counter. The default is 'on'.
init Initial design as an nruns-by-p matrix, where p is
the number of model terms. The default is a random
subset of the rows of C.
maxiter Maximum number of iterations. The default is 10.
18-99
candexch
Parameter Value
start A matrix of treatments as a nobs-by-p matrix, where
p is the number of model terms, specifying a set of
nobs fixed treatments to include in the design. The
default matrix is empty. candexch finds nruns-nobs
additional rows to add to the 'start' design. The
parameter provides the same functionality as the
daugment function, using a row-exchange algorithm
rather than a coordinate-exchange algorithm.
tries Number of times to try to generate a design from
a new starting point. The algorithm uses random
points for each try, except possibly the first. The
default is 1.
Note The rowexch function automatically generates a candidate set
using candgen, and then creates a D-optimal design from that candidate
set using candexch. Call candexch separately to specify your own
candidate set to the row-exchange algorithm.
Examples The following example uses rowexch to generate a five-run design for a
two-factor pure quadratic model using a candidate set that is produced
internally:
dRE1 = rowexch(2,5,'purequadratic','tries',10)
dRE1 =
-1 1
0 0
1 -1
1 0
1 1
The same thing can be done using candgen and candexch in sequence:
[dC,C] = candgen(2,'purequadratic') % Candidate set
18-100
candexch
dC =
-1 -1
0 -1
1 -1
-1 0
0 0
1 0
-1 1
0 1
1 1
C =
1 -1 -1 1 1
1 0 -1 0 1
1 1 -1 1 1
1 -1 0 1 0
1 0 0 0 0
1 1 0 1 0
1 -1 1 1 1
1 0 1 0 1
1 1 1 1 1
treatments = candexch(C,5,'tries',10) % D-opt subset
treatments =
2
1
7
3
4
dRE2 = dC(treatments,:) % Display design
dRE2 =
0 -1
-1 -1
-1 1
1 -1
-1 0
You can replace C in this example with a design matrix evaluated
at your own candidate set. For example, suppose your experiment
18-101
candexch
is constrained so that the two factors cannot have extreme settings
simultaneously. The following produces a restricted candidate set:
constraint = sum(abs(dC),2) < 2; % Feasible treatments
my_dC = dC(constraint,:)
my_dC =
0 -1
-1 0
0 0
1 0
0 1
Use the x2fx function to convert the candidate set to a design matrix:
my_C = x2fx(my_dC,'purequadratic')
my_C =
1 0 -1 0 1
1 -1 0 1 0
1 0 0 0 0
1 1 0 1 0
1 0 1 0 1
Find the required design in the same manner:
my_treatments = candexch(my_C,5,'tries',10) % D-opt subset
my_treatments =
2
4
5
1
3
my_dRE = my_dC(my_treatments,:) % Display design
my_dRE =
-1 0
1 0
0 1
0 -1
0 0
18-102
candexch
See Also candgen | rowexch | cordexch | daugment | x2fx
18-103
candgen
Purpose Candidate set generation
Syntax dC = candgen(nfactors,'model')
[dC,C] = candgen(nfactors,'model')
[...] = candgen(nfactors,'model','Name',value)
Description dC = candgen(nfactors,'model') generates a candidate set dC of
treatments appropriate for estimating the parameters in the model
with nfactors factors. dC has nfactors columns and one row for each
candidate treatment. model is one of the following strings, specified
inside single quotes:
• linear — Constant and linear terms. This is the default.
• interaction — Constant, linear, and interaction terms
• quadratic — Constant, linear, interaction, and squared terms
• purequadratic — Constant, linear, and squared terms
Alternatively, model can be a matrix specifying polynomial terms of
arbitrary order. In this case, model should have one column for each
factor and one row for each term in the model. The entries in any row
of model are powers for the factors in the columns. For example, if a
model has factors X1, X2, and X3, then a row [0 1 2] in model specifies
the term (X1.^0).*(X2.^1).*(X3.^2). A row of all zeros in model
specifies a constant term, which can be omitted.
[dC,C] = candgen(nfactors,'model') also returns the design matrix
C evaluated at the treatments in dC. The order of the columns of C for a
full quadratic model with n terms is:
1 The constant term
2 The linear terms in order 1, 2, ..., n
3 The interaction terms in order (1, 2), (1, 3), ..., (1, n), (2, 3), ..., (n–1, n)
4 The squared terms in order 1, 2, ..., n
18-104
candgen
Other models use a subset of these terms, in the same order.
Pass C to candexch to generate a D-optimal design using a
coordinate-exchange algorithm.
[...] = candgen(nfactors,'model','Name',value) specifies one
or more optional name/value pairs for the design. Valid parameters
and their values are listed in the following table. Specify Name inside
single quotes.
Name Value
bounds Lower and upper bounds for each factor, specified as
a 2-by-nfactors matrix. Alternatively, this value
can be a cell array containing nfactors elements,
each element specifying the vector of allowable
values for the corresponding factor.
categorical Indices of categorical predictors.
levels Vector of number of levels for each factor.
Note The rowexch function automatically generates a candidate set
using candgen, and then creates a D-optimal design from that candidate
set using candexch. Call candexch separately to specify your own
candidate set to the row-exchange algorithm.
Examples The following example uses rowexch to generate a five-run design for a
two-factor pure quadratic model using a candidate set that is produced
internally:
dRE1 = rowexch(2,5,'purequadratic','tries',10)
dRE1 =
-1 1
0 0
1 -1
1 0
18-105
candgen
1 1
The same thing can be done using candgen and candexch in sequence:
[dC,C] = candgen(2,'purequadratic') % Candidate set, C
dC =
-1 -1
0 -1
1 -1
-1 0
0 0
1 0
-1 1
0 1
1 1
C =
1 -1 -1 1 1
1 0 -1 0 1
1 1 -1 1 1
1 -1 0 1 0
1 0 0 0 0
1 1 0 1 0
1 -1 1 1 1
1 0 1 0 1
1 1 1 1 1
treatments = candexch(C,5,'tries',10) % Find D-opt subset
treatments =
2
1
7
3
4
dRE2 = dC(treatments,:) % Display design
dRE2 =
0 -1
-1 -1
-1 1
18-106
candgen
1 -1
-1 0
See Also candexch | rowexch
18-107
canoncorr
Purpose Canonical correlation
Syntax [A,B] = canoncorr(X,Y)
[A,B,r] = canoncorr(X,Y)
[A,B,r,U,V] = canoncorr(X,Y)
[A,B,r,U,V,stats] = canoncorr(X,Y)
Description [A,B] = canoncorr(X,Y) computes the sample canonical coefficients
for the n-by-d1 and n-by-d2 data matrices X and Y. X and Y must have
the same number of observations (rows) but can have different numbers
of variables (columns). A and B are d1-by-d and d2-by-d matrices, where
d = min(rank(X),rank(Y)). The jth columns of A and B contain the
canonical coefficients, i.e., the linear combination of variables making
up the jth canonical variable for X and Y, respectively. Columns of
A and B are scaled to make the covariance matrices of the canonical
variables the identity matrix (see U and V below). If X or Y is less than
full rank, canoncorr gives a warning and returns zeros in the rows of A
or B corresponding to dependent columns of X or Y.
[A,B,r] = canoncorr(X,Y) also returns a 1-by-d vector containing the
sample canonical correlations. The jth element of r is the correlation
between the jth columns of U and V (see below).
[A,B,r,U,V] = canoncorr(X,Y) also returns the canonical variables,
scores. U and V are n-by-d matrices computed as
U = (X-repmat(mean(X),N,1))*A
V = (Y-repmat(mean(Y),N,1))*B
[A,B,r,U,V,stats] = canoncorr(X,Y) also returns a structure
stats containing information relating to the sequence of d null
hypotheses H0(k) , that the (k+1)st through dth correlations are all zero,
for k = 0:(d-1). stats contains seven fields, each a 1-by-d vector with
elements corresponding to the values of k, as described in the following
table:
18-108
canoncorr
Field Description
Wilks Wilks’ lambda (likelihood ratio) statistic
chisq
Bartlett’s approximate chi-squared statistic for H0(k)
with Lawley’s modification
pChisq Right-tail significance level for chisq
F
Rao’s approximate F statistic for H0(k)
pF Right-tail significance level for F
df1 Degrees of freedom for the chi-squared statistic, and
the numerator degrees of freedom for the F statistic
df2 Denominator degrees of freedom for the F statistic
Examples load carbig;
X = [Displacement Horsepower Weight Acceleration MPG];
nans = sum(isnan(X),2) > 0;
[A B r U V] = canoncorr(X(~nans,1:3),X(~nans,4:5));
plot(U(:,1),V(:,1),'.')
xlabel('0.0025*Disp+0.020*HP-0.000025*Wgt')
ylabel('-0.17*Accel-0.092*MPG')
18-109
canoncorr
References [1] Krzanowski, W. J. Principles of Multivariate Analysis: A User’s
Perspective. New York: Oxford University Press, 1988.
[2] Seber, G. A. F. Multivariate Observations. Hoboken, NJ: John Wiley
& Sons, Inc., 1984.
See Also manova1 | princomp
18-110
capability
Purpose Process capability indices
Syntax S = capability(data,specs)
Description S = capability(data,specs) estimates capability indices for
measurements in data given the specifications in specs. data can be
either a vector or a matrix of measurements. If data is a matrix, indices
are computed for the columns. specs can be either a two-element vector
of the form [L,U] containing lower and upper specification limits, or (if
data is a matrix) a two-row matrix with the same number of columns as
data. If there is no lower bound, use -Inf as the first element of specs.
If there is no upper bound, use Inf as the second element of specs.
The output S is a structure with the following fields:
• mu — Sample mean
• sigma — Sample standard deviation
• P — Estimated probability of being within limits
• Pl — Estimated probability of being below L
• Pu — Estimated probability of being above U
• Cp — (U-L)/(6*sigma)
• Cpl — (mu-L)./(3.*sigma)
• Cpu — (U-mu)./(3.*sigma)
• Cpk — min(Cpl,Cpu)
Indices are computed under the assumption that data values are
independent samples from a normal population with constant mean
and variance.
Indices divide a “specification width” (between specification limits) by
a “process width” (between control limits). Higher ratios indicate a
process with fewer measurements outside of specification.
18-111
capability
Examples Simulate a sample from a process with a mean of 3 and a standard
deviation of 0.005:
data = normrnd(3,0.005,100,1);
Compute capability indices if the process has an upper specification
limit of 3.01 and a lower specification limit of 2.99:
S = capability(data,[2.99 3.01])
S =
mu: 3.0006
sigma: 0.0047
P: 0.9669
Pl: 0.0116
Pu: 0.0215
Cp: 0.7156
Cpl: 0.7567
Cpu: 0.6744
Cpk: 0.6744
Visualize the specification and process widths:
capaplot(data,[2.99 3.01]);
grid on
18-112
capability
References [1] Montgomery, D. Introduction to Statistical Quality Control.
Hoboken, NJ: John Wiley & Sons, 1991, pp. 369–374.
See Also capaplot | histfit
18-113
capaplot
Purpose Process capability plot
Syntax p = capaplot(data,specs)
[p,h] = capaplot(data,specs)
Description p = capaplot(data,specs) estimates the mean of and variance for
the observations in input vector data, and plots the pdf of the resulting
T distribution. The observations in data are assumed to be normally
distributed. The output, p, is the probability that a new observation
from the estimated distribution will fall within the range specified by
the two-element vector specs. The portion of the distribution between
the lower and upper bounds specified in specs is shaded in the plot.
[p,h] = capaplot(data,specs) additionally returns handles to the
plot elements in h.
capaplot treats NaN values in data as missing, and ignores them.
Examples Simulate a sample from a process with a mean of 3 and a standard
deviation of 0.005:
data = normrnd(3,0.005,100,1);
Compute capability indices if the process has an upper specification
limit of 3.01 and a lower specification limit of 2.99:
S = capability(data,[2.99 3.01])
S =
mu: 3.0006
sigma: 0.0047
P: 0.9669
Pl: 0.0116
Pu: 0.0215
Cp: 0.7156
Cpl: 0.7567
Cpu: 0.6744
Cpk: 0.6744
18-114
capaplot
Visualize the specification and process widths:
capaplot(data,[2.99 3.01]);
grid on
See Also capability | histfit
18-115
caseread
Purpose Read case names from file
Syntax names = caseread('filename')
names = caseread
Description names = caseread('filename') reads the contents of filename and
returns a string matrix of names. filename is the name of a file in
the current folder, or the complete path name of any file elsewhere.
caseread treats each line as a separate case.
names = caseread displays the Select File to Open dialog box for
interactive selection of the input file.
Examples Read the file months.dat created using the casewrite function.
type months.dat
January
February
March
April
May
names = caseread('months.dat')
names =
January
February
March
April
May
See Also casewrite | gname | tdfread | tblread
18-116
casewrite
Purpose Write case names to file
Syntax casewrite(strmat,'filename')
casewrite(strmat)
Description casewrite(strmat,'filename') writes the contents of string matrix
strmat to filename. Each row of strmat represents one case name.
filename is the name of a file in the current folder, or the complete
path name of any file elsewhere. casewrite writes each name to a
separate line in filename.
casewrite(strmat) displays the Select File to Write dialog box for
interactive specification of the output file.
Examples strmat = char('January','February',...
'March','April','May')
strmat =
January
February
March
April
May
casewrite(strmat,'months.dat')
type months.dat
January
February
March
April
May
See Also gname | caseread | tblwrite | tdfread
18-117
categorical.cat
Purpose Concatenate categorical arrays
Syntax c = cat(dim,A,B,...)
Description c = cat(dim,A,B,...) concatenates the categorical arrays A,B,...
along dimension dim. All inputs must have the same size except along
dimension dim. The set of categorical levels for C is the sorted union of
the sets of levels of the inputs, as determined by their labels.
See Also cat | horzcat | vertcat
18-118
categorical class
Purpose Arrays for categorical data
Description categorical is an abstract class, and you cannot create instances of it
directly. You must create nominal or ordinal arrays.
Categorical arrays store data with values in a discrete set of levels.
Each level is meant to capture a single, defining characteristic of an
observation. If you do not encode ordering in the levels, the data and
the array are nominal. If you do encode an ordering, the data and the
array are ordinal.
Construction categorical Create categorical array
Methods addlevels Add levels to categorical array
cat Concatenate categorical arrays
cellstr Convert categorical array to cell
array of strings
char Convert categorical array to
character array
circshift Shift categorical array circularly
ctranspose Transpose categorical matrix
disp Display categorical array
display Display categorical array
double Convert categorical array to
double array
droplevels Drop levels
end Last index in indexing expression
for categorical array
18-119
categorical class
flipdim Flip categorical array along
specified dimension
fliplr Flip categorical matrix in
left/right direction
flipud Flip categorical matrix in
up/down direction
getlabels Access categorical array labels
getlevels Get categorical array levels
hist Plot histogram of categorical data
horzcat Horizontal concatenation for
categorical arrays
int16 Convert categorical array to
signed 16-bit integer array
int32 Convert categorical array to
signed 32-bit integer array
int64 Convert categorical array to
signed 64-bit integer array
int8 Convert categorical array to
signed 8-bit integer array
intersect Set intersection for categorical
arrays
ipermute Inverse permute dimensions of
categorical array
isempty True for empty categorical array
isequal True if categorical arrays are
equal
islevel Test for levels
ismember True for elements of categorical
array in set
18-120
categorical class
isscalar True if categorical array is scalar
isundefined Test for undefined elements
isvector True if categorical array is vector
length Length of categorical array
levelcounts Element counts by level
ndims Number of dimensions of
categorical array
numel Number of elements in categorical
array
permute Permute dimensions of
categorical array
reorderlevels Reorder levels
repmat Replicate and tile categorical
array
reshape Resize categorical array
rot90 Rotate categorical matrix 90
degrees
setdiff Set difference for categorical
arrays
setlabels Label levels
setxor Set exclusive-or for categorical
arrays
shiftdim Shift dimensions of categorical
array
single Convert categorical array to
single array
size Size of categorical array
18-121
categorical class
squeeze Squeeze singleton dimensions
from categorical array
subsasgn Subscripted assignment for
categorical array
subsindex Subscript index for categorical
array
subsref Subscripted reference for
categorical array
summary Summary statistics for categorical
array
times Product of categorical arrays
transpose Transpose categorical matrix
uint16 Convert categorical array to
unsigned 16-bit integers
uint32 Convert categorical array to
unsigned 32-bit integers
uint64 Convert categorical array to
unsigned 64-bit integers
uint8 Convert categorical array to
unsigned 8-bit integers
union Set union for categorical arrays
unique Unique values in categorical
array
vertcat Vertical concatenation for
categorical arrays
Properties labels Text labels for levels
undeflabel Text label for undefined levels
18-122
categorical class
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
How To • “Categorical Arrays” on page 2-13
18-123
categorical
Purpose Create categorical array
Description categorical is an abstract class, and you cannot create instances of it
directly. You must create nominal or ordinal arrays.
See Also nominal | ordinal
18-124
dataset.cat
Purpose Concatenate dataset arrays
Syntax ds = cat(dim, ds1, ds2, ...)
Description ds = cat(dim, ds1, ds2, ...) concatenates the dataset arrays ds1,
ds2, ... along dimension dim by calling the dataset/horzcat or
dataset/vertcat method. dim must be 1 or 2.
See Also horzcat | vertcat
18-125
classregtree.catsplit
Purpose Categorical splits used for branches in decision tree
Syntax v=catsplit(t)
v=catsplit(t,j)
Description v=catsplit(t) returns an n-by-2 cell array v. Each row in v gives left
and right values for a categorical split. For each branch node j based on
a categorical predictor variable z, the left child is chosen if z is in v(j,1)
and the right child is chosen if z is in v(j,2). The splits are in the
same order as nodes of the tree. Nodes for these splits can be found by
running cuttype and selecting 'categorical' cuts from top to bottom.
v=catsplit(t,j) takes an array j of rows and returns the splits for
the specified rows.
See Also classregtree
18-126
gmdistribution.cdf
Purpose Cumulative distribution function for Gaussian mixture distribution
Syntax y = cdf(obj,X)
Description y = cdf(obj,X) returns a vector y of length n containing the values of
the cumulative distribution function (cdf) for the gmdistribution object
obj, evaluated at the n-by-d data matrix X, where n is the number of
observations and d is the dimension of the data. obj is an object created
by gmdistribution or fit. y(I) is the cdf of observation I.
Examples Create a gmdistribution object defining a two-component mixture of
bivariate Gaussian distributions:
MU = [1 2;-3 -5];
SIGMA = cat(3,[2 0;0 .5],[1 0;0 1]);
p = ones(1,2)/2;
obj = gmdistribution(MU,SIGMA,p);
ezsurf(@(x,y)cdf(obj,[x y]),[-10 10],[-10 10])
18-127
gmdistribution.cdf
See Also gmdistribution | fit | pdf | mvncdf
18-128
ccdesign
Purpose Central composite design
Syntax dCC = ccdesign(n)
[dCC,blocks] = ccdesign(n)
[...] = ccdesign(n,'Name',value)
Description dCC = ccdesign(n) generates a central composite design for n factors.
n must be an integer 2 or larger. The output matrix dCC is m-by-n, where
m is the number of runs in the design. Each row represents one run,
with settings for all factors represented in the columns. Factor values
are normalized so that the cube points take values between -1 and 1.
[dCC,blocks] = ccdesign(n) requests a blocked design. The output
blocks is an m-by-1 vector of block numbers for each run. Blocks
indicate runs that are to be measured under similar conditions
to minimize the effect of inter-block differences on the parameter
estimates.
[...] = ccdesign(n,'Name',value) specifies one or more optional
name/value pairs for the design. Valid parameters and their values are
listed in the following table. Specify Name in single quotes.
Parameter Description Values
center Number of • Integer — Number of center
center points. points to include.
• 'uniform' — Select number of
center points to give uniform
precision.
• 'orthogonal' — Select
number of center points to give
an orthogonal design. This is
the default.
fraction Fraction of • 0 — Whole design. This is the
full-factorial default.
cube, expressed
• 1 — 1/2 fraction.
18-129
ccdesign
Parameter Description Values
as an exponent
of 1/2. • 2 — 1/4 fraction.
type Type of CCD. • 'circumscribed' —
Circumscribed (CCC). This is
the default.
• 'inscribed' — Inscribed
(CCI).
• 'faced' — Faced (CCF).
blocksize Maximum Integer. The default is Inf.
number of
points per block.
Examples The following creates a 2-factor CCC:
dCC = ccdesign(2,'type','circumscribed')
dCC =
-1.0000 -1.0000
-1.0000 1.0000
1.0000 -1.0000
1.0000 1.0000
-1.4142 0
1.4142 0
0 -1.4142
0 1.4142
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
18-130
ccdesign
The center point is run 8 times to reduce the correlations among the
coefficient estimates.
Visualize the design as follows:
plot(dCC(:,1),dCC(:,2),'ro','MarkerFaceColor','b')
X = [1 -1 -1 -1; 1 1 1 -1];
Y = [-1 -1 1 -1; 1 -1 1 1];
line(X,Y,'Color','b')
axis square equal
See Also bbdesign
18-131
cdf
Purpose Cumulative distribution functions
Syntax Y = cdf('name',X,A)
Y = cdf('name',X,A,B)
Y = cdf('name',X,A,B,C)
Description Y = cdf('name',X,A) computes the cumulative distribution function
for the one-parameter family of distributions specified by name. A
contains parameter values for the distribution. The cumulative
distribution function is evaluated at the values in X and its values are
returned in Y.
If X and A are arrays, they must be the same size. If X is a scalar, it is
expanded to a constant matrix the same size as A. If A is a scalar, it is
expanded to a constant matrix the same size as X.
Y is the common size of X and A after any necessary scalar expansion.
Y = cdf('name',X,A,B) computes the cumulative distribution function
for two-parameter families of distributions, where parameter values
are given in A and B.
If X, A, and B are arrays, they must be the same size. If X is a scalar, it is
expanded to a constant matrix the same size as A and B. If either A or B
are scalars, they are expanded to constant matrices the same size as X.
Y is the common size of X, A, and B after any necessary scalar expansion.
Y = cdf('name',X,A,B,C) computes the cumulative distribution
function for three-parameter families of distributions, where parameter
values are given in A, B, and C.
If X, A, B, and C are arrays, they must be the same size. If X is a scalar,
it is expanded to a constant matrix the same size as A, B, and C. If
any of A, B or C are scalars, they are expanded to constant matrices
the same size as X.
Y is the common size of X, A, B, and C after any necessary scalar
expansion.
Acceptable strings for name (specified in single quotes) are:
18-132
cdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
beta or “Beta a b —
Beta Distribution”
on page B-4
bino or “Binomial n: number p: —
Binomial Distribution” of trials probability
on page B-7 of success
for each
trial
chi2 or “Chi-Square ν: degrees — —
Chisquare Distribution” of freedom
on page
B-12
exp or “Exponential : mean — —
Exponential Distribution”
on page
B-16
ev or “Extreme : location σ: scale —
Extreme Value parameter parameter
Value Distribution”
on page
B-19
f or F “F ν 1: ν 2: —
Distribution” numerator denominator
gam or on page
“Gamma degrees
a : shapeof degrees
b : scale of —
Gamma B-25
Distribution” freedom
parameter freedom
parameter
on page
B-27
18-133
cdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
gev or “Generalized K: shape : location σ: scale
Generalized Extreme parameter parameter parameter
Extreme Value
Value Distribution”
on page
B-32
gp or “Generalized k: tail index σ: scale : threshold
Generalized Pareto (shape) parameter (location)
Pareto Distribution” parameter parameter
on page
B-37
geo or “Geometric p: — —
Geometric Distribution” probability
hyge or on page parameter
“Hypergeometric
M : size of the K: number n: number
B-41
Distribution”
Hypergeometric population of items of samples
on page with the drawn
B-43 desired
characteristic
in the
population
logn or “Lognormal σ —
Lognormal Distribution”
on page
B-51
nbin or “Negative r: number p: —
Negative Binomial of successes probability
Binomial Distribution” of success
on page in a single
B-72 trial
18-134
cdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
ncf or “Noncentral ν 1: ν 2: δ:
Noncentral F numerator denominator noncentrality
F Distribution” degrees of degrees of parameter
on page freedom freedom
B-78
nct or “Noncentral ν: degrees δ: —
Noncentral t of freedom noncentrality
t Distribution” parameter
ncx2 or on page
“Noncentral ν: degrees δ: —
Noncentral B-80
Chi-Square of freedom noncentrality
Chi-square Distribution” parameter
on page
B-76
norm or “Normal : mean σ: standard —
Normal Distribution” deviation
on page
B-83
poiss or “Poisson λ: mean — —
Poisson Distribution”
on page
B-89
rayl or “Rayleigh b: scale — —
Rayleigh Distribution” parameter
on page
B-91
t or T “Student’s t ν: degrees — —
Distribution” of freedom
on page
B-95
18-135
cdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
unif or “Uniform a: lower b: upper —
Uniform Distribution endpoint endpoint
(Continuous)” (minimum) (maximum)
on page
B-99
unid or “Uniform N: — —
Discrete Distribution maximum
Uniform (Discrete)” observable
on page value
B-101
wbl or “Weibull a: scale b: shape —
Weibull Distribution” parameter parameter
on page
B-103
Examples Compute the cdf of the normal distribution with mean 0 and standard
deviation 1 at inputs –2, –1, 0, 1, 2:
p1 = cdf('Normal',-2:2,0,1)
p1 =
0.0228 0.1587 0.5000 0.8413 0.9772
The order of the parameters is the same as for normcdf.
Compute the cdfs of Poisson distributions with rate parameters 0, 1, ...,
4 at inputs 1, 2, ..., 5, respectively:
p2 = cdf('Poisson',0:4,1:5)
p2 =
0.3679 0.4060 0.4232 0.4335 0.4405
The order of the parameters is the same as for poisscdf.
18-136
cdf
See Also pdf | icdf
18-137
piecewisedistribution.cdf
Purpose Cumulative distribution function for piecewise distribution
Syntax P = cdf(obj,X)
Description P = cdf(obj,X) returns an array P of values of the cumulative
distribution function for the piecewise distribution object obj, evaluated
at the values in the array X.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
[p,q] = boundary(obj)
p =
0.1000
0.9000
q =
-1.7766
1.8432
cdf(obj,q)
ans =
0.1000
0.9000
See Also paretotails | pdf | icdf
18-138
ProbDist.cdf
Purpose Return cumulative distribution function (CDF) for ProbDist object
Syntax Y = cdf(PD, X)
Description Y = cdf(PD, X) returns Y, an array containing the cumulative
distribution function (CDF) for the ProbDist object PD, evaluated at
values in X.
Input PD An object of the class ProbDistUnivParam or
Arguments ProbDistUnivKernel.
X A numeric array of values where you want to
evaluate the CDF.
Output Y An array containing the cumulative
Arguments distribution function (CDF) for the ProbDist
object PD.
See Also cdf
18-139
cdfplot
Purpose Empirical cumulative distribution function plot
Syntax cdfplot(X)
h = cdfplot(X)
[h,stats] = cdfplot(X)
Description cdfplot(X) displays a plot of the empirical cumulative distribution
function (cdf) for the data in the vector X. The empirical cdf F(x) is
defined as the proportion of X values less than or equal to x.
This plot, like those produced by hist and normplot, is useful for
examining the distribution of a sample of data. You can overlay a
theoretical cdf on the same plot to compare the empirical distribution
of the sample to the theoretical distribution.
The kstest, kstest2, and lillietest functions compute test statistics
that are derived from the empirical cdf. You may find the empirical
cdf plot produced by cdfplot useful in helping you to understand the
output from those functions.
h = cdfplot(X) returns a handle to the cdf curve.
[h,stats] = cdfplot(X) also returns a stats structure with the
following fields.
Field Description
stats.min Minimum value
stats.max Maximum value
stats.mean Sample mean
stats.median Sample median (50th percentile)
stats.std Sample standard deviation
Examples The following example compares the empirical cdf for a sample from
an extreme value distribution with a plot of the cdf for the sampling
distribution. In practice, the sampling distribution would be unknown,
and would be chosen to match the empirical cdf.
18-140
cdfplot
y = evrnd(0,3,100,1);
cdfplot(y)
hold on
x = -20:0.1:10;
f = evcdf(x,0,3);
plot(x,f,'m')
legend('Empirical','Theoretical','Location','NW')
See Also ecdf
18-141
categorical.cellstr
Purpose Convert categorical array to cell array of strings
Syntax B = cellstr(A)
Description B = cellstr(A) converts the categorical array A to a cell array of
strings. Each element of B contains the categorical level label for the
corresponding element of A.
See Also char | getlabels
18-142
dataset.cellstr
Purpose Create cell array of strings from dataset array
Syntax B = cellstr(A)
B = cellstr(A,VARS)
Description B = cellstr(A) returns the contents of the dataset A, converted to a
cell array of strings. The variables in the dataset must support the
conversion and must have compatible sizes.
B = cellstr(A,VARS) returns the contents of the dataset variables
specified by VARS. VARS is a positive integer, a vector of positive integers,
a variable name, a cell array containing one or more variable names, or
a logical vector.
See Also dataset.double | dataset.replacedata
18-143
categorical.char
Purpose Convert categorical array to character array
Syntax B = char(A)
Description B = char(A) converts the categorical array A to a 2-D character matrix.
char does not preserve the shape of A. B contains numel(A) rows, and
each row of B contains the categorical level label for the corresponding
element of A(:).
See Also cellstr | getlabels
18-144
chi2cdf
Purpose Chi-square cumulative distribution function
Syntax P = chi2cdf(X,V)
Description P = chi2cdf(X,V) computes the chi-square cdf at each of the values
in X using the corresponding degrees of freedom in V. X and V can
be vectors, matrices, or multidimensional arrays that have the same
size. A scalar input is expanded to a constant array with the same
dimensions as the other input.
The degrees of freedom parameters in V must be positive integers, and
the values in X must lie on the interval [0 Inf].
The χ2 cdf for a given value x and degrees-of-freedom ν is
x t( −2)/ 2 e− t / 2
p = F ( x | ) = ∫0 2 / 2 Γ( / 2)
dt
where Γ( · ) is the Gamma function.
The chi-square density function with ν degrees-of-freedom is the same
as the gamma density function with parameters ν/2 and 2.
Examples probability = chi2cdf(5,1:5)
probability =
0.9747 0.9179 0.8282 0.7127 0.5841
probability = chi2cdf(1:5,1:5)
probability =
0.6827 0.6321 0.6084 0.5940 0.5841
See Also cdf | chi2pdf | chi2inv | chi2stat | chi2rnd
How To • “Chi-Square Distribution” on page B-12
18-145
chi2gof
Purpose Chi-square goodness-of-fit test
Syntax h = chi2gof(x)
[h,p] = chi2gof(...)
[h,p,stats] = chi2gof(...)
[...] = chi2gof(X,'Name',value)
Description h = chi2gof(x) performs a chi-square goodness-of-fit test of the
default null hypothesis that the data in vector x are a random sample
from a normal distribution with mean and variance estimated from
x, against the alternative that the data are not normally distributed
with the estimated mean and variance. The result h is 1 if the null
hypothesis can be rejected at the 5% significance level. The result h is 0
if the null hypothesis cannot be rejected at the 5% significance level.
The null distribution can be changed from a normal distribution to
an arbitrary discrete or continuous distribution. See the syntax for
specifying optional argument name/value pairs below.
The test is performed by grouping the data into bins, calculating
the observed and expected counts for those bins, and computing the
chi-square test statistic
N
χ2 = ∑ ( Oi − Ei ) / Ei
2
i=1
where Oi are the observed counts and Ei are the expected counts. The
statistic has an approximate chi-square distribution when the counts
are sufficiently large. Bins in either tail with an expected count less
than 5 are pooled with neighboring bins until the count in each extreme
bin is at least 5. If bins remain in the interior with counts less than 5,
chi2gof displays a warning. In this case, you should use fewer bins,
or provide bin centers or edges, to increase the expected counts in all
bins. (See the syntax for specifying optional argument name/value pairs
below.) chi2gof sets the number of bins, nbins, to 10 by default, and
compares the test statistic to a chi-square distribution with nbins – 3
degrees of freedom to take into account the two estimated parameters.
18-146
chi2gof
[h,p] = chi2gof(...) also returns the p value of the test, p. The p
value is the probability, under assumption of the null hypothesis, of
observing the given statistic or one more extreme.
[h,p,stats] = chi2gof(...) also returns a structure stats with
the following fields:
• chi2stat — The chi-square statistic
• df — Degrees of freedom
• edges — Vector of bin edges after pooling
• O — Observed count in each bin
• E — Expected count in each bin
[...] = chi2gof(X,'Name',value) specifies one or more optional
argument name/value pairs chosen from the following lists. Argument
names are case insensitive and partial matches are allowed. Specify
Name in single quotes.
The following name/value pairs control the initial binning of the data
before pooling. You should not specify more than one of these options.
• nbins — The number of bins to use. Default is 10.
• ctrs — A vector of bin centers
• edges — A vector of bin edges
The following name/value pairs determine the null distribution for the
test. Do not specify both cdf and expected.
• cdf — A fully specified cumulative distribution function. This can
be a function name, a function handle, or a ProbDist object of the
ProbDistUnivParam class or ProbDistUnivKernel class. When
'cdf' is a function name or handle, the distribution function must
take x as its only argument. Alternately, you can provide a cell array
whose first element is a function name or handle, and whose later
18-147
chi2gof
elements are parameter values, one per cell. The function must take
x as its first argument, and other parameters as later arguments.
• expected — A vector with one element per bin specifying the
expected counts for each bin.
• nparams — The number of estimated parameters; used to adjust
the degrees of freedom to be nbins – 1 – nparams, where nbins is
the number of bins.
If your cdf or expected input depends on estimated parameters, you
should use nparams to ensure that the degrees of freedom for the test is
correct. If cdfis a cell array, the default value of nparams is the number
of parameters in the array; otherwise the default is 0.
The following name/value pairs control other aspects of the test.
• emin — The minimum allowed expected value for a bin; any bin in
either tail having an expected value less than this amount is pooled
with a neighboring bin. Use the value 0 to prevent pooling. The
default is 5.
• frequency — A vector the same length as x containing the frequency
of the corresponding xvalues
• alpha — Significance level for the test. The default is 0.05.
Examples Example 1
Equivalent ways to test against an unspecified normal distribution
with estimated parameters:
x = normrnd(50,5,100,1);
[h,p] = chi2gof(x)
h =
0
p =
0.7532
18-148
chi2gof
[h,p] = chi2gof(x,'cdf',@(z)normcdf(z,mean(x),std(x)),'nparams',2)
h =
0
p =
0.7532
[h,p] = chi2gof(x,'cdf',{@normcdf,mean(x),std(x)})
h =
0
p =
0.7532
Example 2
Test against the standard normal:
x = randn(100,1);
[h,p] = chi2gof(x,'cdf',@normcdf)
h =
0
p =
0.9443
Example 3
Test against the standard uniform:
x = rand(100,1);
n = length(x);
edges = linspace(0,1,11);
expectedCounts = n * diff(edges);
[h,p,st] = chi2gof(x,'edges',edges,...
'expected',expectedCounts)
h =
0
p =
0.3191
18-149
chi2gof
st =
chi2stat: 10.4000
df: 9
edges: [1x11 double]
O: [6 11 4 12 15 8 14 9 11 10]
E: [1x10 double]
Example 4
Test against the Poisson distribution by specifying observed and
expected counts:
bins = 0:5;
obsCounts = [6 16 10 12 4 2];
n = sum(obsCounts);
lambdaHat = sum(bins.*obsCounts)/n;
expCounts = n*poisspdf(bins,lambdaHat);
[h,p,st] = chi2gof(bins,'ctrs',bins,...
'frequency',obsCounts, ...
'expected',expCounts,...
'nparams',1)
h =
0
p =
0.4654
st =
chi2stat: 2.5550
df: 3
edges: [1x6 double]
O: [6 16 10 12 6]
E: [7.0429 13.8041 13.5280 8.8383 6.0284]
See Also crosstab | lillietest | kstest | chi2cdf | chi2pdf | chi2inv |
chi2stat | chi2rnd
How To • “Chi-Square Distribution” on page B-12
18-150
chi2inv
Purpose Chi-square inverse cumulative distribution function
Syntax X = chi2inv(P,V)
Description X = chi2inv(P,V) computes the inverse of the chi-square cdf with
degrees of freedom specified by V for the corresponding probabilities in
P. P and V can be vectors, matrices, or multidimensional arrays that
have the same size. A scalar input is expanded to a constant array with
the same dimensions as the other inputs.
The degrees of freedom parameters in V must be positive integers, and
the values in P must lie in the interval [0 1].
The inverse chi-square cdf for a given probability p and ν degrees of
freedom is
x = F −1 ( p| ) = { x : F ( x | ) = p}
where
x t( −2)/ 2 e− t / 2
p = F ( x | ) = ∫0 2 / 2 Γ( / 2)
dt
and Γ( · ) is the Gamma function. Each element of output X is the value
whose cumulative probability under the chi-square cdf defined by the
corresponding degrees of freedom parameter in V is specified by the
corresponding value in P.
Examples Find a value that exceeds 95% of the samples from a chi-square
distribution with 10 degrees of freedom.
x = chi2inv(0.95,10)
x =
18.3070
You would observe values greater than 18.3 only 5% of the time by
chance.
18-151
chi2inv
See Also icdf | chi2cdf | chi2pdf | chi2stat | chi2rnd
How To • “Chi-Square Distribution” on page B-12
18-152
chi2pdf
Purpose Chi-square probability density function
Syntax Y = chi2pdf(X,V)
Description Y = chi2pdf(X,V) computes the chi-square pdf at each of the values
in X using the corresponding degrees of freedom in V. X and V can be
vectors, matrices, or multidimensional arrays that have the same size,
which is also the size of the output Y. A scalar input is expanded to a
constant array with the same dimensions as the other input.
The degrees of freedom parameters in V must be positive integers, and
the values in X must lie on the interval [0 Inf].
The chi-square pdf for a given value x and ν degrees of freedom is
x( −2)/ 2 e− x / 2
y = f ( x | ) =
2 / 2 Γ( / 2)
where Γ( · ) is the Gamma function.
If x is standard normal, then x2 is distributed chi-square with one
degree of freedom. If x1, x2, ..., xn are n independent standard normal
observations, then the sum of the squares of the x’s is distributed
chi-square with n degrees of freedom (and is equivalent to the gamma
density function with parameters ν/2 and 2).
Examples nu = 1:6;
x = nu;
y = chi2pdf(x,nu)
y =
0.2420 0.1839 0.1542 0.1353 0.1220 0.1120
The mean of the chi-square distribution is the value of the degrees of
freedom parameter, nu. The above example shows that the probability
density of the mean falls as nu increases.
See Also pdf | chi2cdf | chi2inv | chi2stat | chi2rnd
18-153
chi2pdf
How To • “Chi-Square Distribution” on page B-12
18-154
chi2rnd
Purpose Chi-square random numbers
Syntax R = chi2rnd(V)
R = chi2rnd(V,u)
R = chi2rnd(V,m,n)
Description R = chi2rnd(V) generates random numbers from the chi-square
distribution with degrees of freedom parameters specified by V. V can be
a vector, a matrix, or a multidimensional array. R is the same size as V.
R = chi2rnd(V,u) generates an array R of size u containing random
numbers from the chi-square distribution with degrees of freedom
parameters specified by V, where u is a row vector. If u is a 1-by-2
vector, R is a matrix with u(1) rows and u(2) columns. If u is 1-by-n, R
is an n-dimensional array.
R = chi2rnd(V,m,n) generates an m-by-n matrix containing random
numbers from the chi-square distribution with degrees of freedom
parameter V.
Examples Note that the first and third commands are the same, but are different
from the second command.
r = chi2rnd(1:6)
r =
0.0037 3.0377 7.8142 0.9021 3.2019 9.0729
r = chi2rnd(6,[1 6])
r =
6.5249 2.6226 12.2497 3.0388 6.3133 5.0388
r = chi2rnd(1:6,1,6)
r =
0.7638 6.0955 0.8273 3.2506 1.5469 10.9197
See Also random | chi2cdf | chi2pdf | chi2inv | chi2stat
18-155
chi2rnd
How To • “Chi-Square Distribution” on page B-12
18-156
chi2stat
Purpose Chi-square mean and variance
Syntax [M,V] = chi2stat(NU)
Description [M,V] = chi2stat(NU) returns the mean of and variance for the
chi-square distribution with degrees of freedom parameters specified
by NU.
The mean of the chi-square distribution is ν, the degrees of freedom
parameter, and the variance is 2ν.
Examples nu = 1:10;
nu = nu'*nu;
[m,v] = chi2stat(nu)
m =
1 2 3 4 5 6 7 8 9 10
2 4 6 8 10 12 14 16 18 20
3 6 9 12 15 18 21 24 27 30
4 8 12 16 20 24 28 32 36 40
5 10 15 20 25 30 35 40 45 50
6 12 18 24 30 36 42 48 54 60
7 14 21 28 35 42 49 56 63 70
8 16 24 32 40 48 56 64 72 80
9 18 27 36 45 54 63 72 81 90
10 20 30 40 50 60 70 80 90 100
v =
2 4 6 8 10 12 14 16 18 20
4 8 12 16 20 24 28 32 36 40
6 12 18 24 30 36 42 48 54 60
8 16 24 32 40 48 56 64 72 80
10 20 30 40 50 60 70 80 90 100
12 24 36 48 60 72 84 96 108 120
14 28 42 56 70 84 98 112 126 140
16 32 48 64 80 96 112 128 144 160
18 36 54 72 90 108 126 144 162 180
20 40 60 80 100 120 140 160 180 200
18-157
chi2stat
See Also chi2cdf | chi2pdf | chi2inv | chi2rnd
How To • “Chi-Square Distribution” on page B-12
18-158
classregtree.children
Purpose Child nodes
Syntax C = children(t)
C = children(t,nodes)
Description C = children(t) returns an n-by-2 array C containing the numbers
of the child nodes for each node in the tree t, where n is the number of
nodes. Leaf nodes have child node 0.
C = children(t,nodes) takes a vector nodes of node numbers and
returns the children for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-159
classregtree.children
C = children(t)
C =
2 3
0 0
4 5
6 7
0 0
8 9
0 0
0 0
18-160
classregtree.children
0 0
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree | numnodes | parent
18-161
cholcov
Purpose Cholesky-like covariance decomposition
Syntax T = cholcov(SIGMA)
[T,num] = cholcov(SIGMA)
[T,num] = cholcov(SIGMA,0)
Description T = cholcov(SIGMA) computes T such that SIGMA = T'*T. SIGMA
must be square, symmetric, and positive semi-definite. If SIGMA is
positive definite, then T is the square, upper triangular Cholesky factor.
If SIGMA is not positive definite, T is computed from an eigenvalue
decomposition of SIGMA. T is not necessarily triangular or square in this
case. Any eigenvectors whose corresponding eigenvalue is close to zero
(within a small tolerance) are omitted. If any remaining eigenvalues
are negative, T is empty.
[T,num] = cholcov(SIGMA) returns the number num of negative
eigenvalues of SIGMA, and T is empty if num is positive. If num is zero,
SIGMA is positive semi-definite. If SIGMA is not square and symmetric,
num is NaN and T is empty.
[T,num] = cholcov(SIGMA,0) returns num equal to zero if SIGMA
is positive definite, and T is the Cholesky factor. If SIGMA is not
positive definite, num is a positive integer and T is empty. [...] =
cholcov(SIGMA,1) is equivalent to [...] = cholcov(SIGMA).
Examples The following 4-by-4 covariance matrix is rank-deficient:
C1 = [2 1 1 2;1 2 1 2;1 1 2 2;2 2 2 3]
C1 =
2 1 1 2
1 2 1 2
1 1 2 2
2 2 2 3
rank(C1)
ans =
3
Use cholcov to factor C1:
18-162
cholcov
T = cholcov(C1)
T =
-0.2113 0.7887 -0.5774 0
0.7887 -0.2113 -0.5774 0
1.1547 1.1547 1.1547 1.7321
C2 = T'*T
C2 =
2.0000 1.0000 1.0000 2.0000
1.0000 2.0000 1.0000 2.0000
1.0000 1.0000 2.0000 2.0000
2.0000 2.0000 2.0000 3.0000
Use T to generate random data with the specified covariance:
C3 = cov(randn(1e6,3)*T)
C3 =
1.9973 0.9982 0.9995 1.9975
0.9982 1.9962 0.9969 1.9956
0.9995 0.9969 1.9980 1.9972
1.9975 1.9956 1.9972 2.9951
See Also chol | cov
18-163
categorical.circshift
Purpose Shift categorical array circularly
Syntax B = circshift(A,shiftsize)
Description B = circshift(A,shiftsize) circularly shifts the values in the
categorical array A by shiftsize elements. shiftsize is a vector of
integer scalars where the n-th element specifies the shift amount for the
n-th dimension of array A. If an element in shiftsize is positive, the
values of A are shifted down (or to the right). If it is negative, the values
of A are shifted up (or to the left).
See Also permute | shiftdim
18-164
NaiveBayes.CIsNonEmpty property
Purpose Flag for non-empty classes
Description The CIsNonEmpty property is a logical vector of length NClasses
specifying which classes are not empty. When the grouping variable
is categorical, it may contain categorical levels that don’t appear in
the elements of the grouping variable. Those levels are empty and
NaiveBayes ignores them for the purposes of training the classifier.
18-165
classregtree.classcount
Purpose Class counts
Syntax P = classcount(t)
P = classcount(t,nodes)
Description P = classcount(t) returns an n-by-m array P of class counts for the
nodes in the classification tree t, where n is the number of nodes and
m is the number of classes. For any node number i, the class counts
P(i,:) are counts of observations (from the data used in fitting the
tree) from each class satisfying the conditions for node i.
P = classcount(t,nodes) takes a vector nodes of node numbers and
returns the class counts for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-166
classregtree.classcount
P = classcount(t)
P =
50 50 50
50 0 0
0 50 50
0 49 5
0 1 45
0 47 1
0 2 4
0 47 0
18-167
classregtree.classcount
0 0 1
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree | numnodes
18-168
classify
Purpose Discriminant analysis
Syntax class = classify(sample,training,group)
class = classify(sample,training,group,'type')
class = classify(sample,training,group,'type',prior)
[class,err] = classify(...)
[class,err,POSTERIOR] = classify(...)
[class,err,POSTERIOR,logp] = classify(...)
[class,err,POSTERIOR,logp,coeff] = classify(...)
Description class = classify(sample,training,group) classifies each row of
the data in sample into one of the groups in training. sample and
training must be matrices with the same number of columns. group is
a grouping variable for training. Its unique values define groups; each
element defines the group to which the corresponding row of training
belongs. group can be a categorical variable, a numeric vector, a string
array, or a cell array of strings. training and group must have the
same number of rows. (See “Grouped Data” on page 2-34.) classify
treats NaNs or empty strings in group as missing values, and ignores
the corresponding rows of training. The output class indicates the
group to which each row of sample has been assigned, and is of the
same type as group.
class = classify(sample,training,group,'type') allows you to
specify the type of discriminant function. Specify type inside single
quotes. type is one of:
• linear — Fits a multivariate normal density to each group, with a
pooled estimate of covariance. This is the default.
• diaglinear — Similar to linear, but with a diagonal covariance
matrix estimate (naive Bayes classifiers).
• quadratic — Fits multivariate normal densities with covariance
estimates stratified by group.
• diagquadratic — Similar to quadratic, but with a diagonal
covariance matrix estimate (naive Bayes classifiers).
18-169
classify
• mahalanobis — Uses Mahalanobis distances with stratified
covariance estimates.
class = classify(sample,training,group,'type',prior) allows
you to specify prior probabilities for the groups. prior is one of:
• A numeric vector the same length as the number of unique values
in group (or the number of levels defined for group, if group is
categorical). If group is numeric or categorical, the order of prior
must correspond to the ordered values in group, or, if group contains
strings, to the order of first occurrence of the values in group.
• A 1-by-1 structure with fields:
- prob — A numeric vector.
- group — Of the same type as group, containing unique values
indicating the groups to which the elements of prob correspond.
As a structure, prior can contain groups that do not appear in
group. This can be useful if training is a subset a larger training
set. classify ignores any groups that appear in the structure but
not in the group array.
• The string 'empirical', indicating that group prior probabilities
should be estimated from the group relative frequencies in training.
prior defaults to a numeric vector of equal probabilities, i.e., a uniform
distribution. prior is not used for discrimination by Mahalanobis
distance, except for error rate calculation.
[class,err] = classify(...) also returns an estimate err of the
misclassification error rate based on the training data. classify
returns the apparent error rate, i.e., the percentage of observations in
training that are misclassified, weighted by the prior probabilities
for the groups.
[class,err,POSTERIOR] = classify(...) also returns a matrix
POSTERIOR of estimates of the posterior probabilities that the jth
training group was the source of the ith sample observation, i.e.,
18-170
classify
Pr(group j|obs i). POSTERIOR is not computed for Mahalanobis
discrimination.
[class,err,POSTERIOR,logp] = classify(...) also returns a
vector logp containing estimates of the logarithms of the unconditional
predictive probability density of the sample observations, p(obs i) =
∑p(obs i|group j)Pr(group j) over all groups. logp is not computed for
Mahalanobis discrimination.
[class,err,POSTERIOR,logp,coeff] = classify(...) also returns
a structure array coeff containing coefficients of the boundary
curves between pairs of groups. Each element coeff(I,J) contains
information for comparing group I to group J in the following fields:
• type — Type of discriminant function, from the type input.
• name1 — Name of the first group.
• name2 — Name of the second group.
• const — Constant term of the boundary equation (K)
• linear — Linear coefficients of the boundary equation (L)
• quadratic — Quadratic coefficient matrix of the boundary equation
(Q)
For the linear and diaglinear types, the quadratic field is absent,
and a row x from the sample array is classified into group I rather than
group J if 0 < K+x*L. For the other types, x is classified into group I if
0 < K+x*L+x*Q*x'.
Examples For training data, use Fisher’s sepal measurements for iris versicolor
and virginica:
load fisheriris
SL = meas(51:end,1);
SW = meas(51:end,2);
group = species(51:end);
h1 = gscatter(SL,SW,group,'rb','v^',[],'off');
18-171
classify
set(h1,'LineWidth',2)
legend('Fisher versicolor','Fisher virginica',...
'Location','NW')
Classify a grid of measurements on the same scale:
[X,Y] = meshgrid(linspace(4.5,8),linspace(2,4));
X = X(:); Y = Y(:);
[C,err,P,logp,coeff] = classify([X Y],[SL SW],...
group,'quadratic');
Visualize the classification:
18-172
classify
hold on;
gscatter(X,Y,C,'rb','.',1,'off');
K = coeff(1,2).const;
L = coeff(1,2).linear;
Q = coeff(1,2).quadratic;
% Function to compute K + L*v + v'*Q*v for multiple vectors
% v=[x;y]. Accepts x and y as scalars or column vectors.
f = @(x,y) K + [x y]*L + sum(([x y]*Q) .* [x y], 2);
h2 = ezplot(f,[4.5 8 2 4]);
set(h2,'Color','m','LineWidth',2)
axis([4.5 8 2 4])
xlabel('Sepal Length')
ylabel('Sepal Width')
title('{\bf Classification with Fisher Training Data}')
18-173
classify
References [1] Krzanowski, W. J. Principles of Multivariate Analysis: A User’s
Perspective. New York: Oxford University Press, 1988.
[2] Seber, G. A. F. Multivariate Observations. Hoboken, NJ: John Wiley
& Sons, Inc., 1984.
See Also classregtree | mahal | NaiveBayes
How To • “Grouped Data” on page 2-34
18-174
CompactTreeBagger.ClassNames property
Purpose Names of classes
Description The ClassNames property is a cell array containing the class names for
the response variable Y supplied to TreeBagger. This property is empty
for regression trees.
18-175
TreeBagger.ClassNames property
Purpose Names of classes
Description The ClassNames property is a cell array containing the class names for
the response variable Y. This property is empty for regression trees.
18-176
classregtree.classprob
Purpose Class probabilities
Syntax P = classprob(t)
P = classprob(t,nodes)
Description P = classprob(t) returns an n-by-m array P of class probabilities for
the nodes in the classification tree t, where n is the number of nodes
and m is the number of classes. For any node number i, the class
probabilities P(i,:) are the estimated probabilities for each class for a
point satisfying the conditions for node i.
P = classprob(t,nodes) takes a vector nodes of node numbers and
returns the class probabilities for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-177
classregtree.classprob
P = classprob(t)
P =
0.3333 0.3333 0.3333
1.0000 0 0
0 0.5000 0.5000
0 0.9074 0.0926
0 0.0217 0.9783
0 0.9792 0.0208
0 0.3333 0.6667
0 1.0000 0
18-178
classregtree.classprob
0 0 1.0000
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree | numnodes
18-179
classregtree class
Purpose Classification and regression trees
Construction classregtree Construct classification and
regression trees
Methods catsplit Categorical splits used for
branches in decision tree
children Child nodes
classcount Class counts
classprob Class probabilities
cutcategories Cut categories
cutpoint Returns decision tree cut point
values
cuttype Cut types
cutvar Cut variable names
disp Display classregtree object
display Display classregtree object
eval Predicted responses
isbranch Test node for branch
nodeerr Return vector of node errors
nodeprob Node probabilities
nodesize Return node size
numnodes Number of nodes
parent Parent node
prune Prune tree
risk Node risks
18-180
classregtree class
subsasgn Subscripted reference for
classregtree object
subsref Subscripted reference for
classregtree object
test Error rate
type Tree type
varimportance Compute embedded estimates of
input feature importance
view Plot tree
Properties Objects of the classregtree class have no properties accessible by dot
indexing, get methods, or set methods. To obtain information about a
classregtree object, use the appropriate method.
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
How To • “Regression and Classification by Bagging Decision Trees” on page
12-30
• “Classification Trees” on page 12-9
• “Regression Trees” on page 9-94
• “Grouped Data” on page 2-34
18-181
classregtree
Purpose Construct classification and regression trees
Syntax t = classregtree(X,y)
t = classregtree(X,y,'Name',value)
Description t = classregtree(X,y) creates a decision tree t for predicting the
response y as a function of the predictors in the columns of X. X is an
n-by-m matrix of predictor values. If y is a vector of n response values,
classregtree performs regression. If y is a categorical variable,
character array, or cell array of strings, classregtree performs
classification. Either way, t is a binary tree where each branching node
is split based on the values of a column of X. NaN values in X or y are
taken to be missing values. Observations with all missing values for
X or missing values for y are not used in the fit. Observations with
some missing values for X are used to find splits on variables for which
these observations have valid values.
t = classregtree(X,y,'Name',value) specifies one or more optional
parameter name/value pairs. Specify Name in single quotes. The
following options are available:
18-182
classregtree
For all trees:
• categorical — Vector of indices of the columns of X that are to be
treated as unordered categorical variables
• method — Either 'classification' (default if y is text or a
categorical variable) or 'regression' (default if y is numeric).
• names — A cell array of names for the predictor variables, in the
order in which they appear in the X from which the tree was created.
• prune — 'on' (default) to compute the full tree and the optimal
sequence of pruned subtrees, or 'off' for the full tree without
pruning.
• minparent — A number k such that impure nodes must have k or
more observations to be split (default is 10).
• minleaf — A minimal number of observations per tree leaf (default
is 1). If you supply both 'minparent' and 'minleaf', classregtree
uses the setting which results in larger leaves: minparent =
max(minparent,2*minleaf)
• nvartosample — Number of predictor variables randomly selected
for each split. By default all variables are considered for each
decision split.
• mergeleaves — 'on' (default) to merge leaves that originate from
the same parent node and give the sum of risk values greater or equal
to the risk associated with the parent node. If 'off', classregtree
does not merge leaves.
• weights — Vector of observation weights. By default the weight of
every observation is 1. The length of this vector must be equal to
the number of rows in X.
18-183
classregtree
For regression trees only:
• qetoler — Defines tolerance on quadratic error per node for
regression trees. Splitting nodes stops when quadratic error per
node drops below qetoler*qed, where qed is the quadratic error for
the entire data computed before the decision tree is grown: qed =
norm(y-ybar) with ybar estimated as the average of the input array
Y. Default value is 1e-6.
For classification trees only:
• cost — Square matrix C, where C(i,j) is the cost of classifying
a point into class j if its true class is i (default has C(i,j)=1 if
i~=j, and C(i,j)=0 if i=j). Alternatively, this value can be a
structure S having two fields: S.groupcontaining the group names as
a categorical variable, character array, or cell array of strings; and
S.cost containing the cost matrix C.
• splitcriterion — Criterion for choosing a split. One of 'gdi'
(default) or Gini’s diversity index, 'twoing' for the twoing rule, or
'deviance' for maximum deviance reduction.
• priorprob — Prior probabilities for each class, specified as a string
('empirical' or 'equal') or as a vector (one value for each distinct
group name) or as a structure S with two fields:
- S.group containing the group names as a categorical variable,
character array, or cell array of strings
- S.prob containing a vector of corresponding probabilities.
If the input value is 'empirical' (default), class probabilities are
determined from class frequencies in Y. If the input value is 'equal',
all class probabilities are set equal. If both observation weights and
class prior probabilities are supplied, the weights are renormalized to
add up to the value of the prior probability in the respective class.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
18-184
classregtree
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-185
classregtree
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also eval | prune | test | view
How To • Grouped Data
• “Regression and Classification by Bagging Decision Trees” on page
12-30
18-186
NaiveBayes.CLevels property
Purpose Class levels
Description The CLevels property is a vector of the same type as the grouping
variable, containing the unique levels of the grouping variable.
18-187
cluster
Purpose Construct agglomerative clusters from linkages
Syntax T = cluster(Z,'cutoff',c)
T = cluster(Z,'cutoff',c,'depth',d)
T = cluster(Z,'cutoff',c,'criterion',criterion)
T = cluster(Z,'maxclust',n)
Description T = cluster(Z,'cutoff',c) constructs clusters from the
agglomerative hierarchical cluster tree, Z, as generated by the linkage
function. Z is a matrix of size (m – 1)-by-3, where m is the number of
observations in the original data. c is a threshold for cutting Z into
clusters. Clusters are formed when a node and all of its subnodes have
inconsistent value less than c. All leaves at or below the node are
grouped into a cluster. t is a vector of size m containing the cluster
assignments of each observation.
If c is a vector, T is a matrix of cluster assignments with one column
per cutoff value.
T = cluster(Z,'cutoff',c,'depth',d) evaluates inconsistent values
by looking to a depth d below each node. The default depth is 2.
T = cluster(Z,'cutoff',c,'criterion',criterion) uses the
specified criterion for forming clusters, where criterion is one of the
strings 'inconsistent' (default) or 'distance'. The 'distance'
criterion uses the distance between the two subnodes merged at a node
to measure node height. All leaves at or below a node with height less
than c are grouped into a cluster.
T = cluster(Z,'maxclust',n) constructs a maximum of n clusters
using the 'distance' criterion. cluster finds the smallest height at
which a horizontal cut through the tree leaves n or fewer clusters.
If n is a vector, T is a matrix of cluster assignments with one column
per maximum value.
Examples Compare clusters from Fisher iris data with species:
load fisheriris
18-188
cluster
d = pdist(meas);
Z = linkage(d);
c = cluster(Z,'maxclust',3:5);
crosstab(c(:,1),species)
ans =
0 0 2
0 50 48
50 0 0
crosstab(c(:,2),species)
ans =
0 0 1
0 50 47
0 0 2
50 0 0
crosstab(c(:,3),species)
ans =
0 4 0
0 46 47
0 0 1
0 0 2
50 0 0
See Also clusterdata | cophenet | inconsistent | linkage | pdist
18-189
gmdistribution.cluster
Purpose Construct clusters from Gaussian mixture distribution
Syntax idx = cluster(obj,X)
[idx,nlogl] = cluster(obj,X)
[idx,nlogl,P] = cluster(obj,X)
[idx,nlogl,P,logpdf] = cluster(obj,X)
[idx,nlogl,P,logpdf,M] = cluster(obj,X)
Description idx = cluster(obj,X) partitions data in the n-by-d matrix X, where n
is the number of observations and d is the dimension of the data, into
k clusters determined by the k components of the Gaussian mixture
distribution defined by obj. obj is an object created by gmdistribution
or fit. idx is an n-by-1 vector, where idx(I) is the cluster index of
observation I. The cluster index gives the component with the largest
posterior probability for the observation, weighted by the component
probability.
Note The data in X is typically the same as the data used to create
the Gaussian mixture distribution defined by obj. Clustering with
cluster is treated as a separate step, apart from density estimation.
For cluster to provide meaningful clustering with new data, X should
come from the same population as the data used to create obj.
cluster treats NaN values as missing data. Rows of X with NaN values
are excluded from the partition.
[idx,nlogl] = cluster(obj,X) also returns nlogl, the negative
log-likelihood of the data.
[idx,nlogl,P] = cluster(obj,X) also returns the posterior
probabilities of each component for each observation in the n-by-k
matrix P. P(I,J) is the probability of component J given observation I.
[idx,nlogl,P,logpdf] = cluster(obj,X) also returns the n-by-1
vector logpdf containing the logarithm of the estimated probability
density function for each observation. The density estimate for
18-190
gmdistribution.cluster
observation I is a sum over all components of the component density at
I times the component probability.
[idx,nlogl,P,logpdf,M] = cluster(obj,X) also returns an n-by-k
matrix M containing Mahalanobis distances in squared units. M(I,J) is
the Mahalanobis distance of observation I from the mean of component
J.
Examples Generate data from a mixture of two bivariate Gaussian distributions
using the mvnrnd function:
MU1 = [1 2];
SIGMA1 = [2 0; 0 .5];
MU2 = [-3 -5];
SIGMA2 = [1 0; 0 1];
X = [mvnrnd(MU1,SIGMA1,1000);mvnrnd(MU2,SIGMA2,1000)];
scatter(X(:,1),X(:,2),10,'.')
hold on
18-191
gmdistribution.cluster
Fit a two-component Gaussian mixture model:
obj = gmdistribution.fit(X,2);
h = ezcontour(@(x,y)pdf(obj,[x y]),[-8 6],[-8 6]);
18-192
gmdistribution.cluster
Use the fit to cluster the data:
idx = cluster(obj,X);
cluster1 = X(idx == 1,:);
cluster2 = X(idx == 2,:);
delete(h)
h1 = scatter(cluster1(:,1),cluster1(:,2),10,'r.');
h2 = scatter(cluster2(:,1),cluster2(:,2),10,'g.');
legend([h1 h2],'Cluster 1','Cluster 2','Location','NW')
18-193
gmdistribution.cluster
See Also fit | gmdistribution | mahal | posterior
18-194
clusterdata
Purpose Construct agglomerative clusters from data
Syntax T = clusterdata(X,cutoff)
T = clusterdata(X,'Name',value)
Description T = clusterdata(X,cutoff) uses the pdist, linkage, and cluster
functions to construct agglomerative clusters from data X. X is an m-by-n
matrix, treated as m observations of n variables. cutoff is a threshold
for cutting the hierarchical tree generated by linkage into clusters.
When 0 < cutoff < 2, clusterdata forms clusters when inconsistent
values are greater than cutoff (see inconsistent). When cutoff is an
integer ≥ 2, clusterdata interprets cutoff as the maximum number
of clusters to keep in the hierarchical tree generated by linkage. The
output T is a vector of size m containing a cluster number for each
observation.
When 0 < cutoff < 2, T = clusterdata(X,cutoff) is equivalent to:
Y = pdist(X,'euclid');
Z = linkage(Y,'single');
T = cluster(Z,'cutoff',cutoff);
When cutoff is an integer ≥ 2, T = clusterdata(X,cutoff) is
equivalent to:
Y = pdist(X,'euclid');
Z = linkage(Y,'single');
T = cluster(Z,'maxclust',cutoff);
T = clusterdata(X,'Name',value) provides more control over the
clustering through one or more optional name/value pairs. Specify Name
in single quotes. Valid names are:
18-195
clusterdata
Parameter Value
distance Any of the distance metric names allowed by pdist
(follow the 'minkowski' option by the value of the
exponent p)
linkage Any of the linkage methods allowed by the linkage
function
cutoff Cutoff for inconsistent or distance measure
maxclust Maximum number of clusters to form
criterion Either 'inconsistent' or 'distance'
depth Depth for computing inconsistent values
Examples The example first creates a sample data set of random numbers. It then
uses clusterdata to compute the distances between items in the data
set and create a hierarchical cluster tree from the data set. Finally,
the clusterdata function groups the items in the data set into three
clusters. The example uses the find function to list all the items in
cluster 2, and the scatter3 function to plot the data with each cluster
shown in a different color.
X = [gallery('uniformdata',[10 3],12);...
gallery('uniformdata',[10 3],13)+1.2;...
gallery('uniformdata',[10 3],14)+2.5];
T = clusterdata(X,'maxclust',3);
find(T==2)
ans =
11
12
13
14
15
16
17
18
18-196
clusterdata
19
20
scatter3(X(:,1),X(:,2),X(:,3),100,T,'filled')
See Also cluster | inconsistent | kmeans | linkage | pdist
18-197
cmdscale
Purpose Classical multidimensional scaling
Syntax Y = cmdscale(D)
[Y,e] = cmdscale(D)
Description Y = cmdscale(D) takes an n-by-n distance matrix D, and returns an
n-by-p configuration matrix Y. Rows of Y are the coordinates of n points
in p-dimensional space for some p < n. When D is a Euclidean distance
matrix, the distances between those points are given by D. p is the
dimension of the smallest space in which the n points whose inter-point
distances are given by D can be embedded.
[Y,e] = cmdscale(D) also returns the eigenvalues of Y*Y'. When D is
Euclidean, the first p elements of e are positive, the rest zero. If the first
k elements of e are much larger than the remaining (n-k), then you can
use the first k columns of Y as k-dimensional points whose inter-point
distances approximate D. This can provide a useful dimension reduction
for visualization, e.g., for k = 2.
D need not be a Euclidean distance matrix. If it is non-Euclidean or a
more general dissimilarity matrix, then some elements of e are negative,
and cmdscale chooses p as the number of positive eigenvalues. In this
case, the reduction to p or fewer dimensions provides a reasonable
approximation to D only if the negative elements of e are small in
magnitude.
You can specify D as either a full dissimilarity matrix, or in upper
triangle vector form such as is output by pdist. A full dissimilarity
matrix must be real and symmetric, and have zeros along the diagonal
and positive elements everywhere else. A dissimilarity matrix in upper
triangle form must have real, positive entries. You can also specify D
as a full similarity matrix, with ones along the diagonal and all other
elements less than one. cmdscale transforms a similarity matrix to a
dissimilarity matrix in such a way that distances between the points
returned in Y equal or approximate sqrt(1-D). To use a different
transformation, you must transform the similarities prior to calling
cmdscale.
18-198
cmdscale
Examples Generate some points in 4-D space, but close to 3-D space, then reduce
them to distances only.
X = [normrnd(0,1,10,3) normrnd(0,.1,10,1)];
D = pdist(X,'euclidean');
Find a configuration with those inter-point distances.
[Y,e] = cmdscale(D);
% Four, but fourth one small
dim = sum(e > eps^(3/4))
% Poor reconstruction
maxerr2 = max(abs(pdist(X)-pdist(Y(:,1:2))))
% Good reconstruction
maxerr3 = max(abs(pdist(X)-pdist(Y(:,1:3))))
% Exact reconstruction
maxerr4 = max(abs(pdist(X)-pdist(Y)))
% D is now non-Euclidean
D = pdist(X,'cityblock');
[Y,e] = cmdscale(D);
% One is large negative
min(e)
% Poor reconstruction
maxerr = max(abs(pdist(X)-pdist(Y)))
References [1] Seber, G. A. F. Multivariate Observations. Hoboken, NJ: John Wiley
& Sons, Inc., 1984.
See Also mdscale | pdist | procrustes
18-199
NaiveBayes.CNames property
Purpose Class names
Description The CNames property is an NClasses-by-1 cell array containing the
group names, where NClasses number of groups in the grouping
variable used to create the Naive Bayes classifier.
18-200
CompactTreeBagger.combine
Purpose Combine two ensembles
Syntax B1 = combine(B1,B2)
Description B1 = combine(B1,B2) appends decision trees from ensemble B2 to
those stored in B1 and returns ensemble B1. This method requires that
the class and variable names be identical in both ensembles.
See Also TreeBagger.append
18-201
combnk
Purpose Enumeration of combinations
Syntax C = combnk(v,k)
Description C = combnk(v,k) returns all combinations of the n elements in v taken
k at a time.
C = combnk(v,k) produces a matrix C with k columns and n!/k!(n–k)!
rows, where each row contains k of the elements in the vector v.
It is not practical to use this function if v has more than about 15
elements.
Examples Combinations of characters from a string.
C = combnk('tendril',4);
last5 = C(31:35,:)
last5 =
tedr
tenl
teni
tenr
tend
Combinations of elements from a numeric vector.
c = combnk(1:4,2)
c =
3 4
2 4
2 3
1 4
1 3
1 2
See Also perms
18-202
TreeBagger.compact
Purpose Compact ensemble of decision trees
Description Return an object of class CompactTreeBagger holding the structure
of the trained ensemble. The class is more compact than the full
TreeBagger class because it does not contain information for growing
more trees for the ensemble. In particular, it does not contain X and
Y used for training.
See Also CompactTreeBagger
18-203
CompactTreeBagger class
Purpose Compact ensemble of decision trees grown by bootstrap aggregation
Description CompactTreeBagger class is a lightweight class that contains the trees
grown using TreeBagger. CompactTreeBagger does not preserve
any information about how TreeBagger grew the decision trees. It
does not contain the input data used for growing trees, nor does it
contain training parameters such as minimal leaf size or number of
variables sampled for each decision split at random. You can only use
CompactTreeBagger for predicting the response of the trained ensemble
given new data X, and other related functions.
CompactTreeBagger lets you save the trained ensemble to disk, or
use it in any other way, while discarding training data and various
parameters of the training configuration irrelevant for predicting
response of the fully grown ensemble. This reduces storage and memory
requirements, especially for ensembles trained on large datasets.
Construction CompactTreeBagger Create CompactTreeBagger
object
Methods combine Combine two ensembles
error Error (misclassification
probability or MSE)
margin Classification margin
mdsProx Multidimensional scaling of
proximity matrix
meanMargin Mean classification margin
outlierMeasure Outlier measure for data
predict Predict response
18-204
CompactTreeBagger class
proximity Proximity matrix for data
SetDefaultYfit Set default value for predict
Properties ClassNames Names of classes
DefaultYfit Default value returned by
predict
DeltaCritDecisionSplit Split criterion contributions for
each predictor
Method Method used by trees
(classification or regression)
NTrees Number of decision trees in
ensemble
Trees Decision trees in ensemble
VarNames Variable names
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
See Also classregtree
How To • “Classification Trees” on page 12-9
• “Regression and Classification by Bagging Decision Trees” on page
12-30
• “Regression Trees” on page 9-94
• “Grouped Data” on page 2-34
18-205
CompactTreeBagger
Purpose Create CompactTreeBagger object
Description When you use the TreeBagger constructor to grow trees, it creates a
CompactTreeBagger object. You can obtain the compact object from the
full TreeBagger object using the TreeBagger/compact method. You do
not create an instance of CompactTreeBagger directly.
See Also TreeBagger
How To • Grouped Data
• “Regression and Classification by Bagging Decision Trees” on page
12-30
18-206
TreeBagger.ComputeOOBPrediction property
Purpose Flag to compute out-of-bag predictions
Description The ComputeOOBPrediction property is a logical flag specifying
whether out-of-bag predictions for training observations should be
computed. The default is false.
If this flag is true, the following properties are available:
• OOBIndices
• OOBInstanceWeight
If this flag is true, the following methods can be called:
• oobError
• oobMargin
• oobMeanMargin
See Also oobError | OOBIndices | OOBInstanceWeight | oobMargin |
oobMeanMargin
18-207
TreeBagger.ComputeOOBVarImp property
Purpose Flag to compute out-of-bag variable importance
Description The ComputeOOBVarImp property is a logical flag specifying whether
TreeBagger should compute out-of-bag estimates of variable
importance. The default is false.
If this flag is true, the following properties are available:
• OOBPermutedVarDeltaError
• OOBPermutedVarDeltaMeanMargin
• OOBPermutedVarCountRaiseMargin
See Also ComputeOOBPrediction | OOBPermutedVarDeltaError
| OOBPermutedVarDeltaMeanMargin |
OOBPermutedVarCountRaiseMargin | oobMeanMargin | TreeBagger
18-208
confusionmat
Purpose Confusion matrix
Syntax C = confusionmat(group,grouphat)
C = confusionmat(group,grouphat,'order',grouporder)
[C,order] = confusionmat(...)
Description C = confusionmat(group,grouphat) returns the confusion matrix C
determined by the known and predicted groups in group and grouphat,
respectively. group and grouphat are grouping variables with the
same number of observations, as described in “Grouped Data” on page
2-34. Input vectors must be of the same type. C is a square matrix
with size equal to the total number of distinct elements in group and
grouphat. C(i,j) is a count of observations known to be in group
i but predicted to be in group j. Group indices and their order are
the same for the rows and columns of C, computed by grp2idx using
grp2idx(group;grouphat). NaN, empty, or 'undefined' groups are
not counted.
C = confusionmat(group,grouphat,'order',grouporder) uses
grouporder to order the rows and columns of C. grouporder is a
grouping variable containing all of the distinct elements in group and
grouphat. If grouporder contains elements that are not in group or
grouphat, the corresponding entries in C will be 0.
[C,order] = confusionmat(...) also returns the order of the rows
and columns of C in a variable order the same type as group and
grouphat.
Examples Example 1
Display the confusion matrix for data with two misclassifications and
one missing classification:
g1 = [1 1 2 2 3 3]'; % Known groups
g2 = [1 1 2 3 4 NaN]'; % Predicted groups
[C,order] = confusionmat(g1,g2)
C =
18-209
confusionmat
2 0 0 0
0 1 1 0
0 0 0 1
0 0 0 0
order =
1
2
3
4
Example 2
Randomize the measurements and groups in Fisher’s iris data:
load fisheriris
numObs = length(species);
p = randperm(numObs);
meas = meas(p,:);
species = species(p);
Use classify to classify measurements in the second half of the data,
using the first half of the data for training:
half = floor(numObs/2);
training = meas(1:half,:);
trainingSpecies = species(1:half);
sample = meas(half+1:end,:);
grouphat = classify(sample,training,trainingSpecies);
Display the confusion matrix for the resulting classification:
group = species(half+1:end);
[C,order] = confusionmat(group,grouphat)
C =
22 0 0
2 22 0
0 0 29
order =
'virginica'
18-210
confusionmat
'versicolor'
'setosa'
See Also crosstab | grp2idx
How To • “Grouped Data” on page 2-34
18-211
controlchart
Purpose Shewhart control charts
Syntax controlchart(X)
controlchart(x,group)
controlchart(X,group)
[stats,plotdata] = controlchart(x,[group])
controlchart(x,group,'name',value)
Description controlchart(X) produces an xbar chart of the measurements in
matrix X. Each row of X is considered to be a subgroup of measurements
containing replicate observations taken at the same time. The rows
should be in time order. If X is a time series object, the time samples
should contain replicate observations.
The chart plots the means of the subgroups in time order, a center
line (CL) at the average of the means, and upper and lower control
limits (UCL, LCL) at three standard errors from the center line. The
standard error is the estimated process standard deviation divided by
the square root of the subgroup size. Process standard deviation is
estimated from the average of the subgroup standard deviations. Out of
control measurements are marked as violations and drawn with a red
circle. Data cursor mode is enabled, so clicking any data point displays
information about that point.
controlchart(x,group) accepts a grouping variable group for a vector
of measurements x. (See “Grouped Data” on page 2-34.) group is a
categorical variable, vector, string array, or cell array of strings the
same length as x. Consecutive measurements x(n) sharing the same
value of group(n) for 1 ≤ n ≤ length(x) are defined to be a subgroup.
Subgroups can have different numbers of observations.
controlchart(X,group) accepts a grouping variable group for a matrix
of measurements in X. In this case, group is only used to label the time
axis; it does not change the default grouping by rows.
[stats,plotdata] = controlchart(x,[group]) returns a structure
stats of subgroup statistics and parameter estimates, and a structure
plotdata of plotted values. plotdata contains one record for each chart.
18-212
controlchart
The fields in stats and plotdata depend on the chart type.
The fields in stats are selected from the following:
• mean — Subgroup means
• std — Subgroup standard deviations
• range — Subgroup ranges
• n — Subgroup size, or total inspection size or area
• i — Individual data values
• ma — Moving averages
• mr — Moving ranges
• count — Count of defects or defective items
• mu — Estimated process mean
• sigma — Estimated process standard deviation
• p — Estimated proportion defective
• m — Estimated mean defects per unit
The fields in plotdata are the following:
• pts — Plotted point values
• cl — Center line
• lcl — Lower control limit
• ucl — Upper control limit
• se — Standard error of plotted point
• n — Subgroup size
• ooc — Logical that is true for points that are out of control
18-213
controlchart
controlchart(x,group,'name',value) specifies one or more of the
following optional parameter name/value pairs, with name in single
quotes:
• charttype — The name of a chart type chosen from among the
following:
- 'xbar' — Xbar or mean
- 's' — Standard deviation
- 'r' — Range
- 'ewma' — Exponentially weighted moving average
- 'i' — Individual observation
- 'mr' — Moving range of individual observations
- 'ma' — Moving average of individual observations
- 'p' — Proportion defective
- 'np' — Number of defectives
- 'u' — Defects per unit
- 'c' — Count of defects
Alternatively, a parameter can be a cell array listing multiple
compatible chart types. There are four sets of compatible types:
- 'xbar', 's', 'r', and 'ewma'
- 'i', 'mr', and 'ma'
- 'p' and 'np'
- 'u' and 'c'
• display — Either 'on' (default) to display the control chart, or
'off' to omit the display
• label — A string array or cell array of strings, one per subgroup.
This label is displayed as part of the data cursor for a point on the
plot.
18-214
controlchart
• lambda — A parameter between 0 and 1 controlling how much the
current prediction is influenced by past observations in an EWMA
plot. Higher values of 'lambda' give less weight to past observations
and more weight to the current observation. The default is 0.4.
• limits' — A three-element vector specifying the values of the lower
control limit, center line, and upper control limits. Default is to
estimate the center line and to compute control limits based on the
estimated value of sigma. Not permitted if there are multiple chart
types.
• mean — Value for the process mean, or an empty value (default)
to estimate the mean from X. This is the p parameter for p and np
charts, the mean defects per unit for u and c charts, and the normal
mu parameter for other charts.
• nsigma — The number of sigma multiples from the center line to a
control limit. Default is 3.
• parent — The handle of the axes to receive the control chart plot.
Default is to create axes in a new figure. Not permitted if there are
multiple chart types.
• rules — The name of a control rule, or a cell array containing
multiple control rule names. These rules, together with the control
limits, determine if a point is marked as out of control. The default is
to apply no control rules, and to use only the control limits to decide
if a point is out of control. See controlrules for more information.
Control rules are applied to charts that measure the process level
(xbar, i, c, u, p, and np) rather than the variability (r, s), and they
are not applied to charts based on moving statistics (ma, mr, ewma).
• sigma — Either a value for sigma, or a method of estimating
sigma chosen from among 'std' (the default) to use the average
within-subgroup standard deviation, 'range' to use the average
subgroup range, and 'variance' to use the square root of the pooled
variance. When creating i, mr, or ma charts for data not in subgroups,
the estimate is always based on a moving range.
18-215
controlchart
• specs — A vector specifying specification limits. Typically this is a
two-element vector of lower and upper specification limits. Since
specification limits typically apply to individual measurements, this
parameter is primarily suitable for i charts. These limits are not
plotted on r, s, or mr charts.
• unit — The total number of inspected items for p and np charts, and
the size of the inspected unit for u and c charts. In both cases X must
be the count of the number of defects or defectives found. Default is
1 for u and c charts. This argument is required (no default) for p
and np charts.
• width — The width of the window used for computing the moving
ranges and averages in mr and ma charts, and for computing the
sigma estimate in i, mr, and ma charts. Default is 5.
Examples Create xbar and r control charts for the data in parts.mat:
load parts
st = controlchart(runout,'chart',{'xbar' 'r'});
18-216
controlchart
Display the process mean and standard deviation:
fprintf('Parameter estimates: mu = %g, sigma = %g\n',st.mu,st.sigma);
Parameter estimates: mu = -0.0863889, sigma = 0.130215
See Also controlrules
How To • “Grouped Data” on page 2-34
18-217
controlrules
Purpose Western Electric and Nelson control rules
Syntax R = controlrules('rules',x,cl,se)
[R,RULES] = controlrules('rules',x,cl,se)
Description R = controlrules('rules',x,cl,se) determines which points in the
vector x violate the control rules in rules. cl is a vector of center-line
values. se is a vector of standard errors. (Typically, control limits on a
control chart are at the values cl – 3*se and cl + 3*se.) rules is the
name of a control rule, or a cell array containing multiple control rule
names, from the list below. If x has n values and rules contains m
rules, then R is an n-by-m logical array, with R(i,j) assigned the value
1 if point i violates rule j, 0 if it does not.
The following are accepted values for rules (specified inside single
quotes):
• we1 — 1 point above cl + 3*se
• we2 — 2 of 3 above cl + 2*se
• we3 — 4 of 5 above cl + se
• we4 — 8 of 8 above cl
• we5 — 1 below cl 3*se
• we6 — 2 of 3 below cl 2*se
• we7 — 4 of 5 below cl se
• we8 — 8 of 8 below cl
• we9 — 15 of 15 between cl se and cl + se
• we10 — 8 of 8 below cl se or above cl + se
• n1 — 1 point below cl 3*se or above cl + 3*se
• n2 — 9 of 9 on the same side of cl
• n3 — 6 of 6 increasing or decreasing
• n4 — 14 alternating up/down
18-218
controlrules
• n5 — 2 of 3 below cl 2*se or above cl + 2*se, same side
• n6 — 4 of 5 below cl se or above cl + se, same side
• n7 — 15 of 15 between cl se and cl + se
• n8 — 8 of 8 below cl se or above cl + se, either side
• we — All Western Electric rules
• n — All Nelson rules
For multi-point rules, a rule violation at point i indicates that the set
of points ending at point i triggered the rule. Point i is considered to
have violated the rule only if it is one of the points violating the rule’s
condition.
Any points with NaN as their x, cl, or se values are not considered to
have violated rules, and are not counted in the rules for other points.
Control rules can be specified in the controlchart function as values
for the 'rules' parameter.
[R,RULES] = controlrules('rules',x,cl,se) returns a cell array
of text strings RULES listing the rules applied.
Examples Create an xbar chart using the we2 rule to mark out of control
measurements:
load parts;
st = controlchart(runout,'rules','we2');
x = st.mean;
cl = st.mu;
se = st.sigma./sqrt(st.n);
hold on
plot(cl+2*se,'m')
18-219
controlrules
Use controlrules to identify the measurements that violate the
control rule:
R = controlrules('we2',x,cl,se);
I = find(R)
I =
21
23
18-220
controlrules
24
25
26
27
See Also controlchart
18-221
gmdistribution.Converged property
Purpose Determine if algorithm converged
Description Logical true if the algorithm has converged; logical false if the
algorithm has not converged.
Note This property applies only to gmdistribution objects constructed
with fit.
18-222
cophenet
Purpose Cophenetic correlation coefficient
Syntax c = cophenet(Z,Y)
[c,d] = cophenet(Z,Y)
Description c = cophenet(Z,Y) computes the cophenetic correlation coefficient
for the hierarchical cluster tree represented by Z. Z is the output of
the linkage function. Y contains the distances or dissimilarities used
to construct Z, as output by the pdist function. Z is a matrix of size
(m–1)-by-3, with distance information in the third column. Y is a vector
of size m*(m–1)/2.
[c,d] = cophenet(Z,Y) returns the cophenetic distances d in the same
lower triangular distance vector format as Y.
The cophenetic correlation for a cluster tree is defined as the linear
correlation coefficient between the cophenetic distances obtained from
the tree, and the original distances (or dissimilarities) used to construct
the tree. Thus, it is a measure of how faithfully the tree represents the
dissimilarities among observations.
The cophenetic distance between two observations is represented in a
dendrogram by the height of the link at which those two observations
are first joined. That height is the distance between the two subclusters
that are merged by that link.
The output value, c, is the cophenetic correlation coefficient. The
magnitude of this value should be very close to 1 for a high-quality
solution. This measure can be used to compare alternative cluster
solutions obtained using different algorithms.
The cophenetic correlation between Z(:,3) and Y is defined as
∑i< j (Yij − y)( Zij − z)
c=
∑i< j (Yij − y)2 ∑i< j ( Zij − z)2
where:
18-223
cophenet
• Yij is the distance between objects i and j in Y.
• Zij is the cophenetic distance between objects i and j, from Z(:,3).
• y and z are the average of Y and Z(:,3), respectively.
Examples X = [rand(10,3); rand(10,3)+1; rand(10,3)+2];
Y = pdist(X);
Z = linkage(Y,'average');
% Compute Spearman's rank correlation between the
% dissimilarities and the cophenetic distances
[c,D] = cophenet(Z,Y);
r = corr(Y',D','type','spearman')
r =
0.8279
See Also cluster | dendrogram | inconsistent | linkage | pdist |
squareform
18-224
copulacdf
Purpose Copula cumulative distribution function
Syntax Y = copulacdf('Gaussian',U,rho)
Y = copulacdf('t',U,rho,NU)
Y = copulacdf('family',U,alpha)
Description Y = copulacdf('Gaussian',U,rho) returns the cumulative
probability of the Gaussian copula with linear correlation parameters
rho, evaluated at the points in U. U is an n-by-p matrix of values in
[0,1], representing n points in the p-dimensional unit hypercube. rho
is a p-by-p correlation matrix. If U is an n-by-2 matrix, rho may also
be a scalar correlation coefficient.
Y = copulacdf('t',U,rho,NU) returns the cumulative probability
of the t copula with linear correlation parameters rho and degrees
of freedom parameter NU, evaluated at the points in U. U is an n-by-p
matrix of values in [0,1]. rho is a p-by-p correlation matrix. If U is an
n-by-2 matrix, rho may also be a scalar correlation coefficient.
Y = copulacdf('family',U,alpha) returns the cumulative
probability of the bivariate Archimedean copula determined by family,
with scalar parameter alpha, evaluated at the points in U. family is
Clayton, Frank, or Gumbel. U is an n-by-2 matrix of values in [0,1].
Examples u = linspace(0,1,10);
[U1,U2] = meshgrid(u,u);
F = copulacdf('Clayton',[U1(:) U2(:)],1);
surf(U1,U2,reshape(F,10,10))
xlabel('u1')
ylabel('u2')
18-225
copulacdf
See Also copulapdf | copularnd | copulastat | copulaparam
18-226
copulafit
Purpose Fit copula to data
Syntax RHOHAT = copulafit('Gaussian',U)
[RHOHAT,nuhat] = copulafit('t',U)
[RHOHAT,nuhat,nuci] = copulafit('t',U)
paramhat = copulafit('family',U)
[paramhat,paramci] = copulafit('family',U)
[...] = copulafit(...,'alpha',alpha)
[...] = copulafit('t',U,'Method','ApproximateML')
[...] = copulafit(...,'Options',options)
Description RHOHAT = copulafit('Gaussian',U) returns an estimate RHOHAT of
the matrix of linear correlation parameters for a Gaussian copula, given
data in U. U is an n-by-p matrix of values in the open interval (0,1)
representing n points in the p-dimensional unit hypercube.
[RHOHAT,nuhat] = copulafit('t',U) returns an estimate RHOHAT
of the matrix of linear correlation parameters for a t copula and an
estimate nuhat of the degrees of freedom parameter, given data in U.
U is an n-by-p matrix of values in the open interval (0,1) representing
n points in the p-dimensional unit hypercube.
[RHOHAT,nuhat,nuci] = copulafit('t',U) also returns an
approximate 95% confidence interval nuci for the degrees of freedom
parameter estimated in nuhat.
paramhat = copulafit('family',U) returns an estimate paramhat of
the copula parameter for an Archimedean copula specified by family,
given data in U. U is an n-by-2 matrix of values in the open interval (0,1)
representing n points in the unit square. family is one of Clayton,
Frank, or Gumbel.
[paramhat,paramci] = copulafit('family',U) also returns an
approximate 95% confidence interval paramci for the copula parameter
estimated in paramhat.
[...] = copulafit(...,'alpha',alpha) returns approximate
100*(1-alpha)% confidence intervals in nuci or paramci.
18-227
copulafit
Note By default, copulafit uses maximum likelihood to fit a
copula to U. When U contains data transformed to the unit hypercube
by parametric estimates of their marginal cumulative distribution
functions, this is known as the Inference Functions for Margins (IFM)
method. When U contains data transformed by the empirical cdf (see
ecdf), this is known as Canonical Maximum Likelihood (CML).
[...] = copulafit('t',U,'Method','ApproximateML') fits a t
copula for large samples U by maximizing an objective function that
approximates the profile log-likelihood for the degrees of freedom
parameter (see [1]). This method can be significantly faster than
maximum likelihood, but the estimates and confidence limits may not
be accurate for small to moderate sample sizes.
[...] = copulafit(...,'Options',options) specifies control
parameters for the iterative parameter estimation algorithm
using an options structure options as created by statset. Type
statset('copulafit') at the command prompt for fields and default
values used by copulafit. This argument is not applicable to the
'Gaussian' family.
References [1] Bouye, E., V. Durrleman, A. Nikeghbali, G. Riboulet, and T. Roncalli.
“Copulas for Finance: A Reading Guide and Some Applications.”
Working Paper. Groupe de Recherche Operationnelle, Credit Lyonnais,
2000.
Examples Load and plot simulated stock return data:
load stockreturns
x = stocks(:,1);
y = stocks(:,2);
scatterhist(x,y)
18-228
copulafit
Transform the data to the copula scale (unit square) using a kernel
estimator of the cumulative distribution function:
u = ksdensity(x,x,'function','cdf');
v = ksdensity(y,y,'function','cdf');
scatterhist(u,v)
xlabel('u')
ylabel('v')
18-229
copulafit
Fit a t copula:
[Rho,nu] = copulafit('t',[u v],'Method','ApproximateML')
Rho =
1.0000 0.7220
0.7220 1.0000
nu =
2.8934e+006
Generate a random sample from the t copula:
r = copularnd('t',Rho,nu,1000);
18-230
copulafit
u1 = r(:,1);
v1 = r(:,2);
scatterhist(u1,v1)
xlabel('u')
ylabel('v')
set(get(gca,'children'),'marker','.')
Transform the random sample back to the original scale of the data:
x1 = ksdensity(x,u1,'function','icdf');
y1 = ksdensity(y,v1,'function','icdf');
18-231
copulafit
scatterhist(x1,y1)
set(get(gca,'children'),'marker','.')
See Also ecdf | copulacdf | copulaparam | copulapdf | copularnd |
copulastat
18-232
copulaparam
Purpose Copula parameters as function of rank correlation
Syntax rho = copulaparam('Gaussian',R)
rho = copulaparam('t',R,NU)
alpha = copulaparam('family',R)
[...] = copulaparam(...,'type',type)
Description rho = copulaparam('Gaussian',R) returns the linear correlation
parameters rho corresponding to a Gaussian copula having Kendall’s
rank correlation R. If R is a scalar correlation coefficient, rho is a scalar
correlation coefficient corresponding to a bivariate copula. If R is a
p-by-p correlation matrix, rho is a p-by-p correlation matrix.
rho = copulaparam('t',R,NU) returns the linear correlation
parameters rho corresponding to a t copula having Kendall’s rank
correlation R and degrees of freedom NU. If R is a scalar correlation
coefficient, rho is a scalar correlation coefficient corresponding to a
bivariate copula. If R is a p-by-p correlation matrix, rho is a p-by-p
correlation matrix.
alpha = copulaparam('family',R) returns the copula parameter
alpha corresponding to a bivariate Archimedean copula having
Kendall’s rank correlation R. R is a scalar. family is one of Clayton,
Frank, or Gumbel.
[...] = copulaparam(...,'type',type) assumes R is the specified
type of rank correlation. type is 'Kendall' for Kendall’s tau or
'Spearman' for Spearman’s rho.
copulaparam uses an approximation to Spearman’s rank correlation
for some copula families when no analytic formula exists. The
approximation is based on a smooth fit to values computed using Monte
Carlo simulations.
Examples Get the linear correlation coefficient corresponding to a bivariate
Gaussian copula having a rank correlation of -0.5.
tau = -0.5
rho = copulaparam('gaussian',tau)
18-233
copulaparam
rho =
-0.7071
% Generate dependent beta random values using that copula
u = copularnd('gaussian',rho,100);
b = betainv(u,2,2);
% Verify that the sample has a rank correlation
% approximately equal to tau
tau_sample = corr(b,'type','k')
tau_sample =
1.0000 -0.4638
-0.4638 1.0000
See Also copulacdf | copulapdf | copularnd | copulastat
18-234
copulapdf
Purpose Copula probability density function
Syntax Y = copulapdf('Gaussian',U,rho)
Y = copulapdf('t',U,rho,NU)
Y = copulapdf('family',U,alpha)
Description Y = copulapdf('Gaussian',U,rho) returns the probability density of
the Gaussian copula with linear correlation parameters rho, evaluated
at the points in U. U is an n-by-p matrix of values in [0,1], representing
n points in the p-dimensional unit hypercube. rho is a p-by-p correlation
matrix. If U is an n-by-2 matrix, rho may also be a scalar correlation
coefficient.
Y = copulapdf('t',U,rho,NU) returns the probability density of the t
copula with linear correlation parameters rho and degrees of freedom
parameter NU, evaluated at the points in U. U is an n-by-p matrix of
values in [0,1]. rho is a p-by-p correlation matrix. If U is an n-by-2
matrix, rho may also be a scalar correlation coefficient.
Y = copulapdf('family',U,alpha) returns the probability density of
the bivariate Archimedean copula determined by family, with scalar
parameter alpha, evaluated at the points in U. family is Clayton,
Frank, or Gumbel. U is an n-by-2 matrix of values in [0,1].
Examples u = linspace(0,1,10);
[U1,U2] = meshgrid(u,u);
F = copulapdf('Clayton',[U1(:) U2(:)],1);
surf(U1,U2,reshape(F,10,10))
xlabel('u1')
ylabel('u2')
18-235
copulapdf
See Also copulacdf | copulaparam | copularnd | copulastat
18-236
copulastat
Purpose Copula rank correlation
Syntax R = copulastat('Gaussian',rho)
R = copulastat('t',rho,NU)
R = copulastat('family',alpha)
R = copulastat(...,'type','type')
Description R = copulastat('Gaussian',rho) returns the Kendall’s rank
correlation R that corresponds to a Gaussian copula having linear
correlation parameters rho. If rho is a scalar correlation coefficient, R is
a scalar correlation coefficient corresponding to a bivariate copula. If
rho is a p-by-p correlation matrix, R is a p-by-p correlation matrix.
R = copulastat('t',rho,NU) returns the Kendall’s rank correlation R
that corresponds to a t copula having linear correlation parameters rho
and degrees of freedom NU. If rho is a scalar correlation coefficient, R is
a scalar correlation coefficient corresponding to a bivariate copula. If
rho is a p-by-p correlation matrix, R is a p-by-p correlation matrix.
R = copulastat('family',alpha) returns the Kendall’s rank
correlation R that corresponds to a bivariate Archimedean copula with
scalar parameter alpha. family is one of Clayton, Frank, or Gumbel.
R = copulastat(...,'type','type') returns the specified type
of rank correlation. type is Kendall to compute Kendall’s tau, or
Spearman to compute Spearman’s rho.
copulastat uses an approximation to Spearman’s rank correlation
for some copula families when no analytic formula exists. The
approximation is based on a smooth fit to values computed using
Monte-Carlo simulations.
Examples Get the theoretical rank correlation coefficient for a bivariate.
% Gaussian copula with linear correlation parameter rho
rho = -.7071;
tau = copulastat('gaussian',rho)
tau =
-0.5000
18-237
copulastat
% Generate dependent beta random values using that copula
u = copularnd('gaussian',rho,100);
b = betainv(u,2,2);
% Verify that the sample has a rank correlation
% approximately equal to tau
tau_sample = corr(b,'type','k')
tau_sample =
1.0000 -0.5265
-0.5265 1.0000
See Also copulacdf | copulaparam | copulapdf | copularnd
18-238
copularnd
Purpose Copula random numbers
Syntax U = copularnd('Gaussian',rho,N)
U = copularnd('t',rho,NU,N)
U = copularnd('family',alpha,N)
Description U = copularnd('Gaussian',rho,N) returns N random vectors
generated from a Gaussian copula with linear correlation parameters
rho. If rho is a p-by-p correlation matrix, U is an n-by-p matrix. If rho is
a scalar correlation coefficient, copularnd generates U from a bivariate
Gaussian copula. Each column of U is a sample from a Uniform(0,1)
marginal distribution.
U = copularnd('t',rho,NU,N) returns N random vectors generated
from a t copula with linear correlation parameters rho and degrees of
freedom NU. If rho is a p-by-p correlation matrix, U is an n-by-p matrix.
If rho is a scalar correlation coefficient, copularnd generates U from a
bivariate t copula. Each column of U is a sample from a Uniform(0,1)
marginal distribution.
U = copularnd('family',alpha,N) returns N random vectors
generated from the bivariate Archimedean copula determined by
family, with scalar parameter alpha. family is Clayton, Frank, or
Gumbel. U is an n-by-2 matrix. Each column of U is a sample from a
Uniform(0,1) marginal distribution.
Examples Determine the linear correlation parameter corresponding to a bivariate
Gaussian copula having a rank correlation of -0.5.
tau = -0.5
rho = copulaparam('gaussian',tau)
rho =
-0.7071
% Generate dependent beta random values using that copula
u = copularnd('gaussian',rho,100);
b = betainv(u,2,2);
18-239
copularnd
% Verify that the sample has a rank correlation
% approximately equal to tau
tau_sample = corr(b,'type','kendall')
tau_sample =
1.0000 -0.4537
-0.4537 1.0000
See Also copulacdf | copulaparam | copulapdf | copulastat
18-240
cordexch
Purpose Coordinate exchange
Syntax dCE = cordexch(nfactors,nruns)
[dCE,X] = cordexch(nfactors,nruns)
[dCE,X] = cordexch(nfactors,nruns,'model')
[dCE,X] = cordexch(...,'name',value)
Description dCE = cordexch(nfactors,nruns) uses a coordinate-exchange
algorithm to generate a D-optimal design dCE with nruns runs (the rows
of dCE) for a linear additive model with nfactors factors (the columns
of dCE). The model includes a constant term.
[dCE,X] = cordexch(nfactors,nruns) also returns the associated
design matrix X, whose columns are the model terms evaluated at each
treatment (row) of dCE.
[dCE,X] = cordexch(nfactors,nruns,'model') uses the linear
regression model specified in model. model is one of the following
strings, specified inside single quotes:
• linear — Constant and linear terms. This is the default.
• interaction — Constant, linear, and interaction terms
• quadratic — Constant, linear, interaction, and squared terms
• purequadratic — Constant, linear, and squared terms
The order of the columns of X for a full quadratic model with n terms is:
1 The constant term
2 The linear terms in order 1, 2, ..., n
3 The interaction terms in order (1, 2), (1, 3), ..., (1, n), (2, 3), ..., (n–1, n)
4 The squared terms in order 1, 2, ..., n
Other models use a subset of these terms, in the same order.
18-241
cordexch
Alternatively, model can be a matrix specifying polynomial terms of
arbitrary order. In this case, model should have one column for each
factor and one row for each term in the model. The entries in any row
of model are powers for the factors in the columns. For example, if a
model has factors X1, X2, and X3, then a row [0 1 2] in model specifies
the term (X1.^0).*(X2.^1).*(X3.^2). A row of all zeros in model
specifies a constant term, which can be omitted.
[dCE,X] = cordexch(...,'name',value) specifies one or more
optional name/value pairs for the design. Valid parameters and their
values are listed in the following table. Specify name inside single
quotes.
name Value
bounds Lower and upper bounds for each factor, specified as
a 2-by-nfactors matrix. Alternatively, this value
can be a cell array containing nfactors elements,
each element specifying the vector of allowable
values for the corresponding factor.
categorical Indices of categorical predictors.
display Either 'on' or 'off' to control display of the
iteration counter. The default is 'on'.
excludefun Handle to a function that excludes undesirable
runs. If the function is f, it must support the syntax
b = f(S), where S is a matrix of treatments with
nfactors columns and b is a vector of Boolean
values with the same number of rows as S. b(i) is
true if the method should exclude ith row S.
init Initial design as a nruns-by-nfactors matrix. The
default is a randomly selected set of points.
levels Vector of number of levels for each factor.
18-242
cordexch
name Value
maxiter Maximum number of iterations. The default is 10.
tries Number of times to try to generate a design from
a new starting point. The algorithm uses random
points for each try, except possibly the first. The
default is 1.
Algorithm Both cordexch and rowexch use iterative search algorithms. They
operate by incrementally changing an initial design matrix X to increase
D = |XTX| at each step. In both algorithms, there is randomness
built into the selection of the initial design and into the choice of the
incremental changes. As a result, both algorithms may return locally,
but not globally, D-optimal designs. Run each algorithm multiple times
and select the best result for your final design. Both functions have a
'tries' parameter that automates this repetition and comparison.
Unlike the row-exchange algorithm used by rowexch, cordexch does not
use a candidate set. (Or rather, the candidate set is the entire design
space.) At each step, the coordinate-exchange algorithm exchanges a
single element of X with a new element evaluated at a neighboring
point in design space. The absence of a candidate set reduces demands
on memory, but the smaller scale of the search means that the
coordinate-exchange algorithm is more likely to become trapped in a
local minimum.
Examples Suppose you want a design to estimate the parameters in the following
three-factor, seven-term interaction model:
y = 0 + 1 x 1 + 2 x 2 + 3 x 3 + 12 x 1 x 2 + 13 x 1 x 3 + 23 x 2 x 3 +
Use cordexch to generate a D-optimal design with seven runs:
nfactors = 3;
nruns = 7;
[dCE,X] = cordexch(nfactors,nruns,'interaction','tries',10)
dCE =
18-243
cordexch
-1 1 1
-1 -1 -1
1 1 1
-1 1 -1
1 -1 1
1 -1 -1
-1 -1 1
X =
1 -1 1 1 -1 -1 1
1 -1 -1 -1 1 1 1
1 1 1 1 1 1 1
1 -1 1 -1 -1 1 -1
1 1 -1 1 -1 1 -1
1 1 -1 -1 -1 -1 1
1 -1 -1 1 1 -1 -1
Columns of the design matrix X are the model terms evaluated at each
row of the design dCE. The terms appear in order from left to right:
constant term, linear terms (1, 2, 3), interaction terms (12, 13, 23). Use
X to fit the model, as described in “Linear Regression” on page 9-3, to
response data measured at the design points in dCE.
See Also rowexch | daugment | dcovary
18-244
corr
Purpose Linear or rank correlation
Syntax RHO = corr(X)
RHO = corr(X,Y)
[RHO,PVAL] = corr(X,Y)
[RHO,PVAL] = corr(X,Y,'name',value)
Description RHO = corr(X) returns a p-by-p matrix containing the pairwise linear
correlation coefficient between each pair of columns in the n-by-p
matrix X.
RHO = corr(X,Y) returns a p1-by-p2 matrix containing the pairwise
correlation coefficient between each pair of columns in the n-by-p1 and
n-by-p2 matrices X and Y.
[RHO,PVAL] = corr(X,Y) also returns PVAL, a matrix of p-values for
testing the hypothesis of no correlation against the alternative that
there is a nonzero correlation. Each element of PVAL is the p value for
the corresponding element of RHO. If PVAL(i, j) is small, say less than
0.05, then the correlation RHO(i,j) is significantly different from zero.
[RHO,PVAL] = corr(X,Y,'name',value) specifies one or more optional
name/value pairs. Specify name inside single quotes. The following
table lists valid parameters and their values.
Parameter Values
type • 'Pearson' (the default) computes Pearson’s
linear correlation coefficient
• 'Kendall' computes Kendall’s tau
• 'Spearman' computes Spearman’s rho
rows • 'all' (the default) uses all rows regardless of
missing values (NaNs)
• 'complete' uses only rows with no missing
values
18-245
corr
Parameter Values
• 'pairwise'computes RHO(i,j) using rows
with no missing values in column i or j
tail — The • 'both' — Correlation is not zero (the default)
alternative
• 'right' — Correlation is greater than zero
hypothesis
against which to • 'left' — Correlation is less than zero
compute p-values
for testing the
hypothesis of no
correlation
Using the 'pairwise' option for the rows parameter may return a
matrix that is not positive definite. The 'complete' option always
returns a positive definite matrix, but in general the estimates are
based on fewer observations.
corr computes p-values for Pearson’s correlation using a Student’s
t distribution for a transformation of the correlation. This correlation
is exact when X and Y are normal. corr computes p-values for
Kendall’s tau and Spearman’s rho using either the exact permutation
distributions (for small sample sizes), or large-sample approximations.
corr computes p-values for the two-tailed test by doubling the more
significant of the two one-tailed p-values.
References [1] Gibbons, J.D. (1985) Nonparametric Statistical Inference, 2nd ed.,
M. Dekker.
[2] Hollander, M. and D.A. Wolfe (1973) Nonparametric Statistical
Methods, Wiley.
[3] Kendall, M.G. (1970) Rank Correlation Methods, Griffin.
[4] Best, D.J. and D.E. Roberts (1975) "Algorithm AS 89: The Upper
Tail Probabilities of Spearman’s rho", Applied Statistics, 24:377-379.
18-246
corr
See Also corrcoef | partialcorr | corrcov | tiedrank
18-247
corrcov
Purpose Convert covariance matrix to correlation matrix
Syntax R = corrcov(C)
[R,sigma] = corrcov(C)
Description R = corrcov(C) computes the correlation matrix R corresponding to
the covariance matrix C. C must be square, symmetric, and positive
semi-definite.
[R,sigma] = corrcov(C) also computes the vector of standard
deviations sigma.
Examples Use cov and corrcoef to compute covariances and correlations,
respectively, for sample data on weight and blood pressure (systolic,
diastolic) in hospital.mat:
load hospital
X = [hospital.Weight hospital.BloodPressure];
C = cov(X)
C =
706.0404 27.7879 41.0202
27.7879 45.0622 23.8194
41.0202 23.8194 48.0590
R = corrcoef(X)
R =
1.0000 0.1558 0.2227
0.1558 1.0000 0.5118
0.2227 0.5118 1.0000
Compare R with the correlation matrix computed from C by corrcov:
corrcov(C)
ans =
1.0000 0.1558 0.2227
0.1558 1.0000 0.5118
0.2227 0.5118 1.0000
See Also cov | corrcoef | corr | cholcov
18-248
TreeBagger.Cost property
Purpose Misclassification costs
Description The Cost property is a matrix with misclassification costs. This
property is empty for ensembles of regression trees.
See Also classregtree
18-249
gmdistribution.CovType property
Purpose Type of covariance matrices
Description The string 'diagonal' if the covariance matrices are restricted to be
diagonal; the string 'full' otherwise.
18-250
coxphfit
Purpose Cox proportional hazards regression
Syntax b = coxphfit(X,y)
b = coxphfit(X,y,'name',value)
[b,logl,H,stats] = coxphfit(...)
Description b = coxphfit(X,y) returns a p-by-1 vector b of coefficient estimates
for a Cox proportional hazards regression of the responses in y on
the predictors in X. X is an n-by-p matrix of p predictors at each of n
observations. y is an n-by-1 vector of observed responses.
The phrase h(t)*exp(X*b) models the hazard rate for the distribution
of y, where h(t) is a common baseline hazard function. The model does
not include a constant term, and X cannot contain a column of 1s.
b = coxphfit(X,y,'name',value) specifies one or more optional
parameter name/value pairs chosen from the following table. Specify
name inside single quotes.
Name Value
baseline The X values at which to compute the baseline
hazard. Default is mean(X), so the hazard at X is
h(t)*exp((X-mean(X))*b). Enter 0 to compute
the baseline relative to 0, so the hazard at X is
h(t)*exp(X*b).
censoring A Boolean array of the same size as y that is 1
for observations that are right-censored and 0 for
observations that are observed exactly. Default is all
observations observed exactly.
frequency An array of the same size as y containing nonnegative
integer counts. The jth element of this vector gives
the number of times the method observes the jth
element of y and the jth row of X. Default is one
observation per row of X and y.
18-251
coxphfit
Name Value
init A vector containing initial values for the estimated
coefficients b.
options A structure specifying control parameters for
the iterative algorithm used to estimate b. A
call to statset can create this argument. For
parameter names and default values, type
statset('coxphfit').
[b,logl,H,stats] = coxphfit(...) returns additional results. logl
is the log likelihood. H is a two-column matrix containing y values in the
first column and the estimated baseline cumulative hazard evaluated at
those values in the second column. stats is a structure that contains
the fields:
• beta — Coefficient estimates (same as b)
• se — Standard errors of coefficient estimates b
• z — z statistics for b (b divided by standard error)
• p — p-values for b
• covb — Estimated covariance matrix for b
Examples Generate Weibull data depending on predictor x:
x = 4*rand(100,1);
A = 50*exp(-0.5*x); B = 2;
y = wblrnd(A,B);
Fit a Cox model :
[b,logL,H,stats] = coxphfit(x,y);
Show the Cox estimate of the baseline survivor function together with
the known Weibull function:
stairs(H(:,1),exp(-H(:,2)))
18-252
coxphfit
xx = linspace(0,100);
line(xx,1-wblcdf(xx,50*exp(-0.5*mean(x)),B),'color','r')
xlim([0,50])
legend('Survivor Function','Weibull Function')
References [1] Cox, D.R., and D. Oakes. Analysis of Survival Data. London:
Chapman & Hall, 1984.
[2] Lawless, J. F. Statistical Models and Methods for Lifetime Data.
Hoboken, NJ: Wiley-Interscience, 2002.
See Also ecdf | statset | wblfit
18-253
NaiveBayes.CPrior property
Purpose Class priors
Description The CPrior property is a vector of length NClasses containing the class
priors. The priors for empty classes are zero.
18-254
createns
Purpose Create object to use in k-nearest neighbors search
Syntax NS = createns(X)
NS = createns(X,'Name',Value)
Description NS = createns(X) uses the data observations in an mx-by-n matrix
X to create an object NS. Rows of X correspond to observations and
columns correspond to variables. NS is either an ExhaustiveSearcher
or a KDTreeSearcher object which you can use to find nearest neighbors
in X for desired query points. When NS is an ExhaustiveSearcher
object, knnsearch uses the exhaustive search algorithm to find nearest
neighbors. When NS is a KDTreeSearcher, createns creates and saves
a kd-tree based on X in NS. knnsearch uses the kd-tree to find nearest
neighbors. For information on these search methods, see “k-Nearest
Neighbor Search” on page 12-17.
NS = createns(X,'Name',Value) accepts one or more optional
name/value pairs. Specify Name inside single quotes. Specify NSMethod
to determine which type of object to create. The object’s properties
save the information when you specify other arguments. For more
information on the objects’ properties, see ExhaustiveSearcher and
KDTreeSearcher.
Input Name/Value Pairs
Arguments
NSMethod
Nearest neighbors search method, used to define the type of object
created. Value is either:
• 'kdtree' — Create a KDTreeSearcher object. If you do not
specify NSMethod, this is the default value when the number of
columns of X is less than 10, X is not sparse, and the distance
measure is one of the following measures:
- 'euclidean' (default)
- 'cityblock'
18-255
createns
- 'minkowski'
- 'chebychev'
• 'exhaustive' — Create an ExhaustiveSearcher object. If
you do not specify NSMethod, this is the default value when the
default criteria for 'kdtree' do not apply.
Distance
A string or a function handle specifying the default distance
metric used when you call the knnsearch method to find nearest
neighbors for future query points. If you specify a distance metric
but not an NSMethod, this input determines the type of object
createns creates, according to the default values described in
NSMethod.
For both KDTreeSearcher and ExhaustiveSearcher objects, the
following options apply:
• 'euclidean' (default) — Euclidean distance.
• 'cityblock' — City block distance.
• 'chebychev' — Chebychev distance (maximum coordinate
difference).
• 'minkowski' — Minkowski distance.
The following options apply only to ExhaustiveSearcher objects:
• 'seuclidean' — Standardized Euclidean distance. Each
coordinate difference between rows in X and the query matrix is
scaled by dividing by the corresponding element of the standard
deviation computed from X, S=nanstd(X). To specify another
value for S, use the Scale argument.
• 'mahalanobis' — Mahalanobis distance, which is computed
using a positive definite covariance matrix C. The default value
of C is the sample covariance matrix of X, as computed by
nancov(X). To change the value of C, use the Cov parameter.
18-256
createns
• 'cosine' — One minus the cosine of the included angle
between observations (treated as vectors).
• 'correlation' — One minus the sample linear correlation
between observations (treated as sequences of values).
• 'spearman' — One minus the sample Spearman’s rank
correlation between observations (treated as sequences of
values).
• 'hamming' — Hamming distance, which is percentage of
coordinates that differ.
• 'jaccard' — One minus the Jaccard coefficient, which is the
percentage of nonzero coordinates that differ.
• custom distance function — A distance function specified using
@ (for example, @distfun). A distance function must be of the
form function D2 = distfun(ZI, ZJ), taking as arguments
a 1-by-n vector ZI containing a single row from X or from the
query points Y, and an m2-by-n matrix ZJ containing multiple
rows of X or Y, and returning an m2-by-1 vector of distances d2,
whose jth element is the distance between the observations
ZI and ZJ(j,:).
P
A positive scalar, p, indicating the exponent of the Minkowski
distance. This parameter is only valid when Distance is
'minkowski'. Default is 2.
Cov
A positive definite matrix indicating the covariance matrix when
computing the Mahalanobis distance. This parameter is only
valid when Distance is 'mahalanobis'. Default is nancov(X).
Scale
A vector S with the length equal to the number of columns in
X. Each coordinate of X and each query point is scaled by the
corresponding element of S when computing the standardized
18-257
createns
Euclidean distance. This parameter is only valid when Distance
is 'seuclidean'. Default is nanstd(X).
BucketSize
A positive integer, indicating the maximum number of data points
in each leaf node of the kd-tree. This argument is only meaningful
when using the kd-tree search method. Default is 50.
Examples Create a kd-tree with a Minkowski distance metric and a P value of 5:
load fisheriris
x = meas(:,3:4);
% Since x has only two columns and the Distance is Minkowski,
% createns creates a KDTreeSearcher object by default:
knnobj = createns(x,'Distance','minkowski','P',5)
knnobj =
KDTreeSearcher
Properties:
BucketSize: 50
X: [150x2 double]
Distance: 'minkowski'
DistParameter: 5
See Also ExhaustiveSearcher.knnsearch | KDTreeSearcher.knnsearch |
ExhaustiveSearcher | KDTreeSearcher
How To • “k-Nearest Neighbor Search” on page 12-17
18-258
crosstab
Purpose Cross-tabulation
Syntax table = crosstab(x1,x2)
table = crosstab(x1,x2,...,xn)
[table,chi2,p] = crosstab(x1,...,xn)
[table,chi2,p,labels] = crosstab(x1,...,xn)
Description table = crosstab(x1,x2) returns a cross-tabulation table of two
vectors of the same length x1 and x2. table is m-by-n, where m is the
number of distinct values in x1 and n is the number of distinct values
in x2.
x1 and x2 are grouping variables, as described in “Grouped Data” on
page 2-34. crosstab uses grp2idx to assign a positive integer to each
distinct value. table(i,j) is a count of indices where grp2idx(x1)
is i and grp2idx(x2) is j. The numerical order of grp2idx(x1) and
grp2idx(x2) order rows and columns of table, respectively.
table = crosstab(x1,x2,...,xn) returns a multi-dimensional table
where table(i,j,...,n) is a count of indices where grp2idx(x1) is i,
grp2idx(x2) is j, grp2idx(x3) is k, and so on.
[table,chi2,p] = crosstab(x1,...,xn) also returns the chi-square
statistic chi2 and its p value p for a test that table is independent in
each dimension. The null hypothesis is that the proportion in any entry
of table is the product of the proportions in each dimension.
[table,chi2,p,labels] = crosstab(x1,...,xn) also returns a
cell array labels with one column for each input argument. The
entries in the first column are labels for the rows of table, the entries
in the second column are labels for the columns, and so on, for a
multi-dimensional table.
Examples Example 1
Cross-tabulate two vectors with three and four distinct values,
respectively:
x = [1 1 2 3 1]; y = [1 2 5 3 1];
18-259
crosstab
table = crosstab(x,y)
table =
2 1 0 0
0 0 0 1
0 0 1 0
Example 2
Generate two independent vectors, each containing 50 discrete uniform
random numbers in the range 1:3:
x1 = unidrnd(3,50,1);
x2 = unidrnd(3,50,1);
[table,chi2,p] = crosstab(x1,x2)
table =
1 6 7
5 5 2
11 7 6
chi2 =
7.5449
p =
0.1097
At the 95% confidence level, the p value fails to reject the null
hypothesis that table is independent in each dimension.
Example 3
The file carbig.mat contains measurements of large model cars during
the years 1970-1982:
load carbig
[table,chi2,p,labels] = crosstab(cyl4,when,org)
table(:,:,1) =
82 75 25
12 22 38
table(:,:,2) =
0 4 3
23 26 17
table(:,:,3) =
18-260
crosstab
3 3 4
12 25 32
chi2 =
207.7689
p =
0
label =
'Other' 'Early' 'USA'
'Four' 'Mid' 'Europe'
[] 'Late' 'Japan'
table and label together show that the number of four-cylinder cars
made in the USA during the late period of the data was table(2,3,1)
or 38 cars.
See Also grp2idx | tabulate
How To • “Grouped Data” on page 2-34
18-261
crossval
Purpose Loss estimate using cross-validation
Syntax vals = crossval(fun,X)
vals = crossval(fun,X,Y,...)
mse = crossval('mse',X,y,'Predfun',predfun)
mcr = crossval('mcr',X,y,'Predfun',predfun)
val = crossval(criterion,X1,X2,...,y,'Predfun',predfun)
vals = crossval(...,'name',value)
Description vals = crossval(fun,X) performs 10-fold cross-validation for the
function fun, applied to the data in X.
fun is a function handle to a function with two inputs, the training
subset of X, XTRAIN, and the test subset of X, XTEST, as follows:
testval = fun(XTRAIN,XTEST)
Each time it is called, fun should use XTRAIN to fit a model, then return
some criterion testval computed on XTEST using that fitted model.
X can be a column vector or a matrix. Rows of X correspond to
observations; columns correspond to variables or features. Each row of
vals contains the result of applying fun to one test set. If testval is
a non-scalar value, crossval converts it to a row vector using linear
indexing and stored in one row of vals.
vals = crossval(fun,X,Y,...) is used when data are stored in
separate variables X, Y, ... . All variables (column vectors, matrices, or
arrays) must have the same number of rows. fun is called with the
training subsets of X, Y, ... , followed by the test subsets of X, Y, ... ,
as follows:
testvals = fun(XTRAIN,YTRAIN,...,XTEST,YTEST,...)
mse = crossval('mse',X,y,'Predfun',predfun) returns mse, a
scalar containing a 10-fold cross-validation estimate of mean-squared
error for the function predfun. X can be a column vector, matrix, or
array of predictors. y is a column vector of response values. X and y
must have the same number of rows.
18-262
crossval
predfun is a function handle called with the training subset of X, the
training subset of y, and the test subset of X as follows:
yfit = predfun(XTRAIN,ytrain,XTEST)
Each time it is called, predfun should use XTRAIN and ytrain to fit a
regression model and then return fitted values in a column vector yfit.
Each row of yfit contains the predicted values for the corresponding
row of XTEST. crossval computes the squared errors between yfit
and the corresponding response test set, and returns the overall mean
across all test sets.
mcr = crossval('mcr',X,y,'Predfun',predfun) returns mcr, a
scalar containing a 10-fold cross-validation estimate of misclassification
rate (the proportion of misclassified samples) for the function predfun.
The matrix X contains predictor values and the vector y contains class
labels. predfun should use XTRAIN and YTRAIN to fit a classification
model and return yfit as the predicted class labels for XTEST.
crossval computes the number of misclassifications between yfit
and the corresponding response test set, and returns the overall
misclassification rate across all test sets.
val = crossval(criterion,X1,X2,...,y,'Predfun',predfun),
where criterion is 'mse' or 'mcr', returns a cross-validation estimate
of mean-squared error (for a regression model) or misclassification rate
(for a classification model) with predictor values in X1, X2, ... and,
respectively, response values or class labels in y. X1, X2, ... and y must
have the same number of rows. predfun is a function handle called
with the training subsets of X1, X2, ..., the training subset of y, and the
test subsets of X1, X2, ..., as follows:
yfit=predfun(X1TRAIN,X2TRAIN,...,ytrain,X1TEST,X2TEST,...)
yfit should be a column vector containing the fitted values.
vals = crossval(...,'name',value) specifies one or more optional
parameter name/value pairs from the following table. Specify name
inside single quotes.
18-263
crossval
Name Value
holdout A scalar specifying the ratio or the number
of observations p for holdout cross-validation.
When 0 < p < 1, approximately p*n observations
for the test set are randomly selected. When p
is an integer, p observations for the test set are
randomly selected.
kfold A scalar specifying the number of folds k for
k-fold cross-validation.
leaveout Specifies leave-one-out cross-validation. The
value must be 1.
mcreps A positive integer specifying the number of
Monte-Carlo repetitions for validation. Ifthe
first input of crossval is 'mse' or 'mcr',
crossval returns the mean of mean-squared
error or misclassification rate across all of the
Monte-Carlo repetitions. Otherwise, crossval
concatenates the values vals from all of
the Monte-Carlo repetitions along the first
dimension.
partition An object c of the cvpartition class, specifying
the cross-validation type and partition.
stratify A column vector group specifying groups for
stratification. Both training and test sets have
roughly the same class proportions as in group.
NaNs or empty strings in group are treated as
missing values, and the corresponding rows of
the data are ignored.
options A struct that specifies options that govern the
computation of crossval. One option requests
that crossval conduct multiple function
evaluations using multiple processors, if the
Parallel Computing Toolbox is available. Two
18-264
crossval
Name Value
options specify the random number streams to
use in constructing randomized cvpartition
objects. You can create this argument with
a call to statset You can retrieve values of
the individual fields with a call to statget.
Applicable statset parameters are:
• 'UseParallel' — If 'always' and if a
matlabpool of the Parallel Computing Toolbox
is open, compute separate function evaluations
in parallel. If the Parallel Computing Toolbox
is not installed, or a matlabpool is not open,
computation occurs in serial mode. Default is
'never', or serial computation.
• 'UseSubstreams' — If 'always' perform
each partitioning operation using a separate
Substream of the random number generator
(aka Stream). This option is available
only with RandStream types that support
Substreams. Default is 'never', do not use
a different Substream for each partitioning
operation.
• 'Streams' — An object of the RandStream
class, or a scalar cell array (length==1)
containing a RandStream object. If you do not
supply a value for this parameter, crossval
uses the default RandStream on each MATLAB
executable in constructing randomized
partitions. Otherwise, crossval selects
randomized partitions using the supplied
RandStream object(s).
Only one of kfold, holdout, leaveout, or partition can be specified,
and partition cannot be specified with stratify. If both partition
18-265
crossval
and mcreps are specified, the first Monte-Carlo repetition uses the
partition information in the cvpartition object, and the repartition
method is called to generate new partitions for each of the remaining
repetitions. If no cross-validation type is specified, the default is 10-fold
cross-validation.
Note When using cross-validation with classification algorithms,
stratification is preferred. Otherwise, some test sets may not include
observations from all classes.
Examples Example 1
Compute mean-squared error for regression using 10-fold
cross-validation:
load('fisheriris');
y = meas(:,1);
X = [ones(size(y,1),1),meas(:,2:4)];
regf=@(XTRAIN,ytrain,XTEST)(XTEST*regress(ytrain,XTRAIN));
cvMse = crossval('mse',X,y,'predfun',regf)
cvMse =
0.1015
Example 2
Compute misclassification rate using stratified 10-fold cross-validation:
load('fisheriris');
y = species;
X = meas;
cp = cvpartition(y,'k',10); % Stratified cross-validation
classf = @(XTRAIN, ytrain,XTEST)(classify(XTEST,XTRAIN,...
ytrain));
18-266
crossval
cvMCR = crossval('mcr',X,y,'predfun',classf,'partition',cp)
cvMCR =
0.0200
Example 3
Compute the confusion matrix using stratified 10-fold cross-validation:
load('fisheriris');
y = species;
X = meas;
order = unique(y); % Order of the group labels
cp = cvpartition(y,'k',10); % Stratified cross-validation
f = @(xtr,ytr,xte,yte)confusionmat(yte,...
classify(xte,xtr,ytr),'order',order);
cfMat = crossval(f,X,y,'partition',cp);
cfMat = reshape(sum(cfMat),3,3)
cfMat =
50 0 0
0 48 2
0 1 49
cfMat is the summation of 10 confusion matrices from 10 test sets.
References [1] Hastie, T., R. Tibshirani, and J. Friedman. The Elements of
Statistical Learning. New York: Springer, 2001.
See Also cvpartition
How To • “Grouped Data” on page 2-34
18-267
categorical.ctranspose
Purpose Transpose categorical matrix
Syntax B = ctranspose(A)
Description B = ctranspose(A) returns the transpose of the 2-D categorical matrix
A. Note that ctranspose is identical to transpose for categorical arrays.
See Also transpose | permute
18-268
classregtree.cutcategories
Purpose Cut categories
Syntax C = cutcategories(t)
C = cutcategories(t,nodes)
Description C = cutcategories(t) returns an n-by-2 cell array C of the categories
used at branches in the decision tree t, where n is the number of nodes.
For each branch node i based on a categorical predictor variable x,
the left child is chosen if x is among the categories listed in C{i,1},
and the right child is chosen if x is among those listed in C{i,2}. Both
columns of C are empty for branch nodes based on continuous predictors
and for leaf nodes.
C = cutcategories(t,nodes) takes a vector nodes of node numbers
and returns the categories for the specified nodes.
Examples Create a classification tree for car data:
load carsmall
t = classregtree([MPG Cylinders],Origin,...
'names',{'MPG' 'Cyl'},'cat',2)
t =
Decision tree for classification
1 if Cyl=4 then node 2 elseif Cyl in {6 8} then node 3 else USA
2 if MPG<31.5 then node 4 elseif MPG>=31.5 then node 5 else USA
3 if Cyl=6 then node 6 elseif Cyl=8 then node 7 else USA
4 if MPG<21.5 then node 8 elseif MPG>=21.5 then node 9 else USA
5 if MPG<41 then node 10 elseif MPG>=41 then node 11 else Japan
6 if MPG<17 then node 12 elseif MPG>=17 then node 13 else USA
7 class = USA
8 class = France
9 class = USA
10 class = Japan
11 class = Germany
12 class = Germany
13 class = USA
18-269
classregtree.cutcategories
view(t)
C = cutcategories(t)
C =
[4] [1x2 double]
[] []
[6] [ 8]
[] []
[] []
[] []
18-270
classregtree.cutcategories
[] []
[] []
[] []
[] []
[] []
[] []
[] []
C{1,2}
ans =
6 8
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree | cutpoint | cuttype | cutvar
18-271
classregtree.cutpoint
Purpose Returns decision tree cut point values
Syntax v = cutpoint(t)
v = cutpoint(t,nodes)
Description v = cutpoint(t) returns an n-element vector v of the values used as
cut points in the decision tree t, where n is the number of nodes. For
each branch node i based on a continuous predictor variable x, the
left child is chosen if x < v(i) and the right child is chosen if x >=
v(i). v is NaN for branch nodes based on categorical predictors and
for leaf nodes.
v = cutpoint(t,nodes) takes a vector nodes of node numbers and
returns the cut points for the specified nodes.
Examples Create a classification tree for car data:
load carsmall
t = classregtree([MPG Cylinders],Origin,...
'names',{'MPG' 'Cyl'},'cat',2)
t =
Decision tree for classification
1 if Cyl=4 then node 2 elseif Cyl in {6 8} then node 3 else USA
2 if MPG<31.5 then node 4 elseif MPG>=31.5 then node 5 else USA
3 if Cyl=6 then node 6 elseif Cyl=8 then node 7 else USA
4 if MPG<21.5 then node 8 elseif MPG>=21.5 then node 9 else USA
5 if MPG<41 then node 10 elseif MPG>=41 then node 11 else Japan
6 if MPG<17 then node 12 elseif MPG>=17 then node 13 else USA
7 class = USA
8 class = France
9 class = USA
10 class = Japan
11 class = Germany
12 class = Germany
13 class = USA
18-272
classregtree.cutpoint
view(t)
v = cutpoint(t)
v =
NaN
31.5000
NaN
21.5000
41.0000
17.0000
18-273
classregtree.cutpoint
NaN
NaN
NaN
NaN
NaN
NaN
NaN
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, cutcategories, cuttype, cutvar
18-274
classregtree.cuttype
Purpose Cut types
Syntax c = cuttype(t)
c = cuttype(t,nodes)
Description c = cuttype(t) returns an n-element cell array c indicating the type
of cut at each node in the tree t, where n is the number of nodes. For
each node i, c{i} is:
• 'continuous' — If the cut is defined in the form x < v for a variable
x and cut point v.
• 'categorical' — If the cut is defined by whether a variable x takes
a value in a set of categories.
• '' — If i is a leaf node.
cutvar returns the cut points for 'continuous' cuts, and
cutcategories returns the set of categories.
c = cuttype(t,nodes) takes a vector nodes of node numbers and
returns the cut types for the specified nodes.
Examples Create a classification tree for car data:
load carsmall
t = classregtree([MPG Cylinders],Origin,...
'names',{'MPG' 'Cyl'},'cat',2)
t =
Decision tree for classification
1 if Cyl=4 then node 2 elseif Cyl in {6 8} then node 3 else USA
2 if MPG<31.5 then node 4 elseif MPG>=31.5 then node 5 else USA
3 if Cyl=6 then node 6 elseif Cyl=8 then node 7 else USA
4 if MPG<21.5 then node 8 elseif MPG>=21.5 then node 9 else USA
5 if MPG<41 then node 10 elseif MPG>=41 then node 11 else Japan
6 if MPG<17 then node 12 elseif MPG>=17 then node 13 else USA
7 class = USA
18-275
classregtree.cuttype
8 class = France
9 class = USA
10 class = Japan
11 class = Germany
12 class = Germany
13 class = USA
view(t)
c = cuttype(t)
c =
18-276
classregtree.cuttype
'categorical'
'continuous'
'categorical'
'continuous'
'continuous'
'continuous'
''
''
''
''
''
''
''
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, numnodes, cutvar, cutcategories
18-277
classregtree.cutvar
Purpose Cut variable names
Syntax v = cutvar(t)
v = cutvar(t,nodes)
[v,num] = cutvar(...)
Description v = cutvar(t) returns an n-element cell array v of the names of the
variables used for branching in each node of the tree t, where n is
the number of nodes. These variables are sometimes known as cut
variables. For leaf nodes, v contains an empty string.
v = cutvar(t,nodes) takes a vector nodes of node numbers and
returns the cut variables for the specified nodes.
[v,num] = cutvar(...) also returns a vector num containing the
number of each variable.
Examples Create a classification tree for car data:
load carsmall
t = classregtree([MPG Cylinders],Origin,...
'names',{'MPG' 'Cyl'},'cat',2)
t =
Decision tree for classification
1 if Cyl=4 then node 2 elseif Cyl in {6 8} then node 3 else USA
2 if MPG<31.5 then node 4 elseif MPG>=31.5 then node 5 else USA
3 if Cyl=6 then node 6 elseif Cyl=8 then node 7 else USA
4 if MPG<21.5 then node 8 elseif MPG>=21.5 then node 9 else USA
5 if MPG<41 then node 10 elseif MPG>=41 then node 11 else Japan
6 if MPG<17 then node 12 elseif MPG>=17 then node 13 else USA
7 class = USA
8 class = France
9 class = USA
10 class = Japan
11 class = Germany
12 class = Germany
13 class = USA
18-278
classregtree.cutvar
view(t)
[v,num] = cutvar(t)
v =
'Cyl'
'MPG'
'Cyl'
'MPG'
'MPG'
18-279
classregtree.cutvar
'MPG'
''
''
''
''
''
''
''
num =
2
1
2
1
1
1
0
0
0
0
0
0
0
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, numnodes, children
18-280
cvpartition class
Purpose Data partitions for cross-validation
Description An object of the cvpartition class defines a random partition on a set of
data of a specified size. You can be use this partition to define test and
training sets for validating a statistical model using cross-validation.
Construction cvpartition Create cross-validation partition
for data
Methods disp Display cvpartition object
display Display cvpartition object
repartition Repartition data for
cross-validation
test Test indices for cross-validation
training Training indices for
cross-validation
Properties N Number of observations
(including observations with
missing group values)
NumTestSets Number of test sets
TestSize Size of each test set
TrainSize Size of each training set
Type Type of partition
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
18-281
cvpartition class
Examples Use a 10-fold stratified cross-validation to compute the misclassification
error for classify on iris data.
load('fisheriris');
CVO = cvpartition(species,'k',10);
err = zeros(CVO.NumTestSets,1);
for i = 1:CVO.NumTestSets
trIdx = CVO.training(i);
teIdx = CVO.test(i);
ytest = classify(meas(teIdx,:),meas(trIdx,:),...
species(trIdx,:));
err(i) = sum(~strcmp(ytest,species(teIdx)));
end
cvErr = sum(err)/sum(CVO.TestSize);
See Also “Grouped Data” on page 2-34
crossval
18-282
cvpartition
Purpose Create cross-validation partition for data
Syntax c = cvpartition(n,'kfold',k)
c = cvpartition(group,'kfold',k)
c = cvpartition(n,'holdout',p)
c = cvpartition(group,'holdout',p)
c = cvpartition(n,'leaveout')
c = cvpartition(n,'resubstitution')
Description c = cvpartition(n,'kfold',k) constructs an object c of
the cvpartition class defining a random partition for k-fold
cross-validation on n observations. The partition divides the
observations into k disjoint subsamples (or folds), chosen randomly but
with roughly equal size. The default value of k is 10.
c = cvpartition(group,'kfold',k) creates a random partition for a
stratified k-fold cross-validation. group is a numeric vector, categorical
array, string array, or cell array of strings indicating the class of each
observation. Each subsample has roughly equal size and roughly the
same class proportions as in group. cvpartition treats NaNs or empty
strings in group as missing values.
c = cvpartition(n,'holdout',p) creates a random partition for
holdout validation on n observations. This partition divides the
observations into a training set and a test (or holdout) set. The
parameter p must be a scalar. When 0 < p < 1, cvpartition randomly
selects approximately p*n observations for the test set. When p is an
integer, cvpartition randomly selects p observations for the test set.
The default value of p is 1/10.
c = cvpartition(group,'holdout',p) randomly partitions
observations into a training set and a test set with stratification, using
the class information in group; that is, both training and test sets have
roughly the same class proportions as in group.
c = cvpartition(n,'leaveout') creates a random partition for
leave-one-out cross-validation on n observations. Leave-one-out is a
special case of 'kfold', in which the number of folds equals the number
of observations.
18-283
cvpartition
c = cvpartition(n,'resubstitution') creates an object c that does
not partition the data. Both the training set and the test set contain
all of the original n observations.
Examples Use stratified 10-fold cross-validation to compute misclassification rate:
load fisheriris;
y = species;
c = cvpartition(y,'k',10);
fun = @(xT,yT,xt,yt)(sum(~strcmp(yt,classify(xt,xT,yT))));
rate = sum(crossval(fun,meas,y,'partition',c))...
/sum(c.TestSize)
rate =
0.0200
See Also Grouped Data
crossval, repartition
18-284
dataset class
Purpose Arrays for statistical data
Description Dataset arrays are used to collect heterogeneous data and metadata
including variable and observation names into a single container
variable. Dataset arrays are suitable for storing column-oriented or
tabular data that are often stored as columns in a text file or in a
spreadsheet, and can accommodate variables of different types, sizes,
units, etc.
Dataset arrays can contain different kinds of variables, including
numeric, logical, character, categorical, and cell. However, a dataset
array is a different class than the variables that it contains. For
example, even a dataset array that contains only variables that are
double arrays cannot be operated on as if it were itself a double array.
However, using dot subscripting, you can operate on variable in a
dataset array as if it were a workspace variable.
You can subscript dataset arrays using parentheses much like ordinary
numeric arrays, but in addition to numeric and logical indices, you can
use variable and observation names as indices.
Construction Use the dataset constructor to create a dataset array from variables in
the MATLAB workspace. You can also create a dataset array by reading
data from a text or spreadsheet file. You can access each variable in a
dataset array much like fields in a structure, using dot subscripting. See
the following section for a list of operations available for dataset arrays.
dataset Construct dataset array
Methods cat Concatenate dataset arrays
cellstr Create cell array of strings from
dataset array
datasetfun Apply function to dataset array
variables
18-285
dataset class
disp Display dataset array
display Display dataset array
double Convert dataset variables to
double array
end Last index in indexing expression
for dataset array
export Write dataset array to file
get Access dataset array properties
grpstats Summary statistics by group for
dataset arrays
horzcat Horizontal concatenation for
dataset arrays
isempty True for empty dataset array
join Merge observations
length Length of dataset array
ndims Number of dimensions of dataset
array
numel Number of elements in dataset
array
replacedata Replace dataset variables
set Set and display properties
single Convert dataset variables to
single array
size Size of dataset array
sortrows Sort rows of dataset array
stack Stack data from multiple
variables into single variable
18-286
dataset class
subsasgn Subscripted assignment to
dataset array
subsref Subscripted reference for dataset
array
summary Print summary of dataset array
unique Unique observations in dataset
array
unstack Unstack data from single variable
into multiple variables
vertcat Vertical concatenation for dataset
arrays
Properties A dataset array D has properties that store metadata (information
about your data). Access or assign to a property using P =
D.Properties.PropName or D.Properties.PropName = P, where
PropName is one of the following:
Description String describing data set
DimNames Two-element cell array of strings
giving names of dimensions of
data set
ObsNames Cell array of nonempty,
distinct strings giving names
of observations in data set
Units Units of variables in data set
UserData Variable containing additional
information associated with data
set
18-287
dataset class
VarDescription Cell array of strings giving
descriptions of variables in data
set
VarNames Cell array giving names of
variables in data set
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
Examples Load a dataset array from a .mat file and create some simple subsets:
load hospital
h1 = hospital(1:10,:)
h2 = hospital(:,{'LastName' 'Age' 'Sex' 'Smoker'})
% Access and modify metadata
hospital.Properties.Description
hospital.Properties.VarNames{4} = 'Wgt'
% Create a new dataset variable from an existing one
hospital.AtRisk = hospital.Smoker | (hospital.Age > 40)
% Use individual variables to explore the data
boxplot(hospital.Age,hospital.Sex)
h3 = hospital(hospital.Age<30,...
{'LastName' 'Age' 'Sex' 'Smoker'})
% Sort the observations based on two variables
h4 = sortrows(hospital,{'Sex','Age'})
See Also genvarname | tdfread | textscan | xlsread
How To • “Dataset Arrays” on page 2-23
18-288
dataset
Purpose Construct dataset array
Syntax A = dataset(VAR1,VAR2,...)
A = dataset(...,{VAR,name},...)
A = dataset(...,{VAR,name_1,...,name_m},...)
A = dataset(...,'VarNames',{name_1,...,name_m},...)
A = dataset(...,'ObsNames',{name_1,...,name_n},...)
A = dataset('File',filename,param1,val1,param2,val2,...)
A = dataset('XLSFile',filename,param1,val1,param2,val2,...)
A = dataset('XPTFile',xptfilename, ...)
Description A = dataset(VAR1,VAR2,...) creates dataset array A from workspace
variables VAR1, VAR2, ... using the workspace variable names for the
names of the variables in A. Variables can be arrays of any size, but all
variables must be the same size along dimension 1 (rows).
A = dataset(...,{VAR,name},...) creates a variable in dataset A
from the workspace variable VAR and assigns it the name name in A.
Names must be valid, unique MATLAB identifier strings.
A = dataset(...,{VAR,name_1,...,name_m},...), where VAR is an
array with size m along dimension 2 (columns), creates m variables in
dataset A from the columns of the workspace variable VAR and assigns
them the names name_1, ..., name_m in A.
A = dataset(...,'VarNames',{name_1,...,name_m},...) names the
m variables in A with the specified variable names. Names must be valid,
unique MATLAB identifier strings. The number of names must equal
the number of variables in A. You cannot use the 'VarNames' parameter
if you provide names for individual variables using {VAR,name} pairs.
A = dataset(...,'ObsNames',{name_1,...,name_n},...) names the
n observations in A with the specified observation names. The names
need not be valid MATLAB identifier strings, but must be unique. The
number of names must equal the number of observations (rows) in A.
18-289
dataset
Note Dataset arrays may contain built-in types or array objects as
variables. Array objects must implement each of the following:
• Standard MATLAB parenthesis indexing of the form var(i,...),
where i is a numeric or logical vector corresponding to rows of the
variable
• A size method with a dim argument
• A vertcat method
A = dataset('File',filename,param1,val1,param2,val2,...)
creates dataset array A from column-oriented data in the text file
specified by the string filename. Variables in A are of type double
if data in the corresponding column of the file, following the column
header, are entirely numeric; otherwise the variables in A are cell arrays
of strings. dataset converts empty fields to either NaN (for a numeric
variable) or the empty string (for a string-valued variable). dataset
ignores insignificant white space in the file.
The following optional parameter name/value pairs are available:
'Delimiter' A string indicating the character separating
columns in the file. Values are
• '\t' (tab, the default when no format is
specified)
• ' ' (space, the default when a format is
specified)
• ',' (comma)
• ';' (semicolon)
• '|' (bar)
18-290
dataset
'Format' A format string, as accepted by textscan. dataset
reads the file using textscan, and creates variables
in A according to the conversion specifiers in
the format string. You may also provide any
parameter/value pairs accepted by textscan.
Using the 'format' parameter is much faster for
large files. If 'ReadObsNames' is true, the format
string should include a format specifier for the first
column of the file.
'ReadVarNames' A logical value indicating whether (true) or not
(false) to read variable names from the first row
of the file. The default is true. If 'ReadVarNames'
is true, variable names in the column headers of
the file cannot be empty.
'ReadObsNames' A logical value indicating whether (true) or
not (false) to read observation names from the
first column of the file. The default is false. If
'ReadObsNames' and 'ReadVarNames' are both
true, dataset saves the header of the first column
in the file as the name of the first dimension in
A.Properties.DimNames.
'TreatAsEmpty' Specifies strings to treat as the empty string in a
numeric column. Values may be a character string
or a cell array of strings. The parameter applies
only to numeric columns in the file; dataset does
not accept numeric literals such as '-99'.
A = dataset('XLSFile',filename,param1,val1,param2,val2,...)
creates dataset array A from column-oriented data in the Excel®
spreadsheet specified by the string filename. Variables in A are of
type double if data in the corresponding column of the spreadsheet,
following the column header, are entirely numeric; otherwise the
variables in A are cell arrays of strings. Optional parameter name/value
pairs are as follows:
18-291
dataset
'Sheet' A positive scalar value of type double indicating
the sheet number, or a quoted string indicating
the sheet name.
'Range' A string of the form 'C1:C2' where C1 and C2
are the names of cells at opposing corners of a
rectangular region to be read, as for xlsread.
By default, the rectangular region extends to
the right-most column containing data. If the
spreadsheet contains empty columns between
columns of data, or if the spreadsheet contains
figures or other non-tabular information, specify
a range that contains only data.
'ReadVarNames' A logical value indicating whether (true) or
not (false) to read variable names from the
first row of the range. The default is true. If
'ReadVarNames' is true, variable names in the
column headers of the range cannot be empty.
'ReadObsNames' A logical value indicating whether (true) or not
(false) to read observation names from the first
column of the range. The default is false. If
'ReadObsNames' and 'ReadVarNames' are both
true, the header of the first column in the range
is saved as the name of the first dimension in
A.Properties.DimNames.
A = dataset('XPTFile',xptfilename, ...) creates a dataset array
from a SAS XPORT format file. Variable names from the XPORT format
file are preserved. Numeric data types in the XPORT format file are
preserved but all other data types are converted to cell arrays of
strings. The XPORT format allows for 28 missing data types. dataset
represents these in the file by an upper case letter, '.' or '_'. dataset
converts all missing data to NaN values in A. However, if you need the
specific missing types you can use the xptread function to recover the
information.
18-292
dataset
When reading from an XPT format file, the 'ReadObsNames' parameter
name/value pair determines whether or not to try to use the first
variable in the file as observation names. Specify as a logical value
(default false). If the contents of the first variable are not valid
observation names then the variable will be read into a variable of the
dataset array and observation names will not be set.
Examples Create a dataset array from workspace variables:
load cereal
cereal = dataset(Calores,Protein,Fat,Sodium,Fiber,Carbo,...
Sugars,'ObsNames',Name)
cereal.Properties.VarDescription = Variables(4:10,2);
Create a dataset array from a single workspace variable:
load cities
categories = cellstr(categories);
cities = dataset({ratings,categories{:}},...
'ObsNames',cellstr(names))
Load data from a text or spreadsheet file
patients = dataset('File','hospital.dat',...
'Delimiter',',','ReadObsNames',true)
patients2 = dataset('XLSFile','hospital.xls',...
'ReadObsNames',true)
1 Load patient data from the CSV file hospital.dat and store the
information in a dataset array with observation names given by the
first column in the data (patient identification):
patients = dataset('file','hospital.dat', ...
18-293
dataset
'format','%s%s%s%f%f%f%f%f%f%f%f%f', ...
'Delimiter',',','ReadObsNames','on');
You can also load the data without specifying a format string.
dataset will automatically create dataset variables that are either
double arrays or cell arrays of strings, depending on the contents
of the file:
patients = dataset('file','hospital.dat',...
'delimiter',',',...
'ReadObsNames',true);
2 Make the {0,1}-valued variable smoke nominal, and change the labels
to 'No' and 'Yes':
patients.smoke = nominal(patients.smoke,{'No','Yes'});
3 Add new levels to smoke as placeholders for more detailed histories
of smokers:
patients.smoke = addlevels(patients.smoke,...
{'0-5 Years','5-10 Years','LongTerm'});
4 Assuming the nonsmokers have never smoked, relabel the 'No' level:
patients.smoke = setlabels(patients.smoke,'Never','No');
5 Drop the undifferentiated 'Yes' level from smoke:
patients.smoke = droplevels(patients.smoke,'Yes');
Warning: OLDLEVELS contains categorical levels that
were present in A, caused some array elements to have
undefined levels.
Note that smokers now have an undefined level.
6 Set each smoker to one of the new levels, by observation name:
18-294
dataset
patients.smoke('YPL-320') = '5-10 Years';
See Also tdfread | textscan | xlsread
18-295
dataset.datasetfun
Purpose Apply function to dataset array variables
Syntax b = datasetfun(fun,A)
[b,c,...] = datasetfun(fun,A)
[b,...] = datasetfun(fun,A,...,'UniformOutput',false)
[b,...] = datasetfun(fun,A,...,'DatasetOutput',true)
[b,...] = datasetfun(fun,A,...,'DataVars',vars)
[b,...] = datasetfun(fun,A,...,'ObsNames',obsnames)
[b,...] = datasetfun(fun,A,...,'ErrorHandler',efun)
Description b = datasetfun(fun,A) applies the function specified by fun to each
variable of the dataset array A, and returns the results in the vector b.
The ith element of b is equal to fun applied to the ith dataset variable of
A. fun is a function handle to a function that takes one input argument
and returns a scalar value. fun must return values of the same class
each time it is called, and datasetfun concatenates them into the vector
b. The outputs from fun must be one of the following types: numeric,
logical, character, structure, or cell.
To apply functions that return results that are nonscalar or of different
sizes and types, use the 'UniformOutput' or 'DatasetOutput'
parameters described below.
Do not rely on the order in which datasetfun computes the elements
of b, which is unspecified.
If fun is bound to more than one built-in function or file, (that is, if it
represents a set of overloaded functions), datasetfun follows MATLAB
dispatching rules in calling the function. (See “Determining Which
Function Gets Called”.)
[b,c,...] = datasetfun(fun,A), where fun is a function handle
to a function that returns multiple outputs, returns vectors b, c, ...,
each corresponding to one of the output arguments of fun. datasetfun
calls fun each time with as many outputs as there are in the call to
datasetfun. fun may return output arguments having different classes,
but the class of each output must be the same each time fun is called.
18-296
dataset.datasetfun
[b,...] = datasetfun(fun,A,...,'UniformOutput',false) allows
you to specify a function fun that returns values of different sizes or
types. datasetfun returns a cell array (or multiple cell arrays), where
the ith cell contains the value of fun applied to the ith dataset variable
of A. Setting 'UniformOutput' to true is equivalent to the default
behavior.
[b,...] = datasetfun(fun,A,...,'DatasetOutput',true)
specifies that the output(s) of fun are returned as variables in a dataset
array (or multiple dataset arrays). fun must return values with the
same number of rows each time it is called, but it may return values of
any type. The variables in the output dataset array(s) have the same
names as the variables in the input. Setting 'DatasetOutput' to false
(the default) specifies that the type of the output(s) from datasetfun is
determined by 'UniformOutput'.
[b,...] = datasetfun(fun,A,...,'DataVars',vars) allows you to
apply fun only to the dataset variables in A specified by vars. vars is
a positive integer, a vector of positive integers, a variable name, a cell
array containing one or more variable names, or a logical vector.
[b,...] = datasetfun(fun,A,...,'ObsNames',obsnames) specifies
observation names for the dataset output when 'DatasetOutput' is
true.
[b,...] = datasetfun(fun,A,...,'ErrorHandler',efun), where
efun is a function handle, specifies the MATLAB function to call if the
call to fun fails. The error-handling function is called with the following
input arguments:
• A structure with the fields identifier, message, and index,
respectively containing the identifier of the error that occurred, the
text of the error message, and the linear index into the input array(s)
at which the error occurred
• The set of input arguments at which the call to the function failed
The error-handling function should either re-throw an error, or return
the same number of outputs as fun. These outputs are then returned as
18-297
dataset.datasetfun
the outputs of datasetfun. If 'UniformOutput' is true, the outputs of
the error handler must also be scalars of the same type as the outputs
of fun. For example, the following code could be saved in a file as the
error-handling function:
function [A,B] = errorFunc(S,varargin)
warning(S.identifier,S.message);
A = NaN;
B = NaN;
If an error-handling function is not specified, the error from the call
to fun is rethrown.
Examples Compute statistics on selected variables in the hospital dataset array:
load hospital
stats = ...
datasetfun(@mean,hospital,...
'DataVars',{'Weight','BloodPressure'},...
'UniformOutput',false)
stats =
[154] [1x2 double]
stats{2}
ans =
122.7800 82.9600
Display the blood pressure variable:
datasetfun(@hist,hospital,...
'DataVars','BloodPressure',...
'UniformOutput',false);
title('{\bf Blood Pressure}')
legend('Systolic','Diastolic','Location','N')
18-298
dataset.datasetfun
See Also grpstats
18-299
daugment
Purpose D-optimal augmentation
Syntax dCE2 = daugment(dCE,mruns)
[dCE2,X] = daugment(dCE,mruns)
[dCE2,X] = daugment(dCE,mruns,model)
[dCE2,X] = daugment(...,param1,val1,param2,val2,...)
Description dCE2 = daugment(dCE,mruns) uses a coordinate-exchange algorithm
to D-optimally add mruns runs to an existing experimental design dCE
for a linear additive model.
[dCE2,X] = daugment(dCE,mruns) also returns the design matrix X
associated with the augmented design.
[dCE2,X] = daugment(dCE,mruns,model) uses the linear regression
model specified in model. model is one of the following strings:
• 'linear' — Constant and linear terms. This is the default.
• 'interaction' — Constant, linear, and interaction terms
• 'quadratic' — Constant, linear, interaction, and squared terms
• 'purequadratic' — Constant, linear, and squared terms
The order of the columns of X for a full quadratic model with n terms is:
1 The constant term
2 The linear terms in order 1, 2, ..., n
3 The interaction terms in order (1, 2), (1, 3), ..., (1, n), (2, 3), ..., (n–1, n)
4 The squared terms in order 1, 2, ..., n
Other models use a subset of these terms, in the same order.
Alternatively, model can be a matrix specifying polynomial terms of
arbitrary order. In this case, model should have one column for each
factor and one row for each term in the model. The entries in any row
18-300
daugment
of model are powers for the factors in the columns. For example, if a
model has factors X1, X2, and X3, then a row [0 1 2] in model specifies
the term (X1.^0).*(X2.^1).*(X3.^2). A row of all zeros in model
specifies a constant term, which can be omitted.
[dCE2,X] = daugment(...,param1,val1,param2,val2,...) specifies
additional parameter/value pairs for the design. Valid parameters and
their values are listed in the following table.
Parameter Value
'bounds' Lower and upper bounds for each factor, specified
as a 2-by-nfactors matrix, where nfactors is the
number of factors. Alternatively, this value can be
a cell array containing nfactors elements, each
element specifying the vector of allowable values for
the corresponding factor.
'categorical' Indices of categorical predictors.
'display' Either 'on' or 'off' to control display of the
iteration counter. The default is 'on'.
'excludefun' Handle to a function that excludes undesirable
runs. If the function is f, it must support the syntax
b = f(S), where S is a matrix of treatments with
nfactors columns, where nfactors is the number
of factors, and b is a vector of Boolean values with
the same number of rows as S. b(i) is true if the ith
row S should be excluded.
'init' Initial design as an mruns-by-nfactors matrix,
where nfactors is the number of factors. The
default is a randomly selected set of points.
'levels' Vector of number of levels for each factor.
18-301
daugment
Parameter Value
'maxiter' Maximum number of iterations. The default is 10.
'tries' Number of times to try to generate a design from
a new starting point. The algorithm uses random
points for each try, except possibly the first. The
default is 1.
Note The daugment function augments an existing design using a
coordinate-exchange algorithm; the 'start' parameter of the candexch
function provides the same functionality using a row-exchange
algorithm.
Examples The following eight-run design is adequate for estimating main effects
in a four-factor model:
dCEmain = cordexch(4,8)
dCEmain =
1 -1 -1 1
-1 -1 1 1
-1 1 -1 1
1 1 1 -1
1 1 1 1
-1 1 -1 -1
1 -1 -1 -1
-1 -1 1 -1
To estimate the six interaction terms in the model, augment the design
with eight additional runs:
dCEinteraction = daugment(dCEmain,8,'interaction')
dCEinteraction =
1 -1 -1 1
-1 -1 1 1
-1 1 -1 1
18-302
daugment
1 1 1 -1
1 1 1 1
-1 1 -1 -1
1 -1 -1 -1
-1 -1 1 -1
-1 1 1 1
-1 -1 -1 -1
1 -1 1 -1
1 1 -1 1
-1 1 1 -1
1 1 -1 -1
1 -1 1 1
1 1 1 -1
The augmented design is full factorial, with the original eight runs in
the first eight rows.
See Also dcovary, cordexch, candexch
18-303
dcovary
Purpose D-optimal design with fixed covariates
Syntax dCV = dcovary(nfactors,fixed)
[dCV,X] = dcovary(nfactors,fixed)
[dCV,X] = dcovary(nfactors,fixed,model)
[dCV,X] = daugment(...,param1,val1,param2,val2,...)
Description dCV = dcovary(nfactors,fixed) uses a coordinate-exchange
algorithm to generate a D-optimal design for a linear additive model
with nfactors factors, subject to the constraint that the model include
the fixed covariate factors in fixed. The number of runs in the design
is the number of rows in fixed. The design dCV augments fixed with
initial columns for treatments of the model terms.
[dCV,X] = dcovary(nfactors,fixed) also returns the design matrix
X associated with the design.
[dCV,X] = dcovary(nfactors,fixed,model) uses the linear
regression model specified in model. model is one of the following
strings:
• 'linear' — Constant and linear terms. This is the default.
• 'interaction' — Constant, linear, and interaction terms
• 'quadratic' — Constant, linear, interaction, and squared terms
• 'purequadratic' — Constant, linear, and squared terms
The order of the columns of X for a full quadratic model with n terms is:
1 The constant term
2 The linear terms in order 1, 2, ..., n
3 The interaction terms in order (1, 2), (1, 3), ..., (1, n), (2, 3), ..., (n–1, n)
4 The squared terms in order 1, 2, ..., n
Other models use a subset of these terms, in the same order.
18-304
dcovary
Alternatively, model can be a matrix specifying polynomial terms of
arbitrary order. In this case, model should have one column for each
factor and one row for each term in the model. The entries in any row
of model are powers for the factors in the columns. For example, if a
model has factors X1, X2, and X3, then a row [0 1 2] in model specifies
the term (X1.^0).*(X2.^1).*(X3.^2). A row of all zeros in model
specifies a constant term, which can be omitted.
[dCV,X] = daugment(...,param1,val1,param2,val2,...) specifies
additional parameter/value pairs for the design. Valid parameters and
their values are listed in the following table.
Parameter Value
'bounds' Lower and upper bounds for each factor, specified as
a 2-by-nfactors matrix. Alternatively, this value
can be a cell array containing nfactors elements,
each element specifying the vector of allowable
values for the corresponding factor.
'categorical' Indices of categorical predictors.
'display' Either 'on' or 'off' to control display of the
iteration counter. The default is 'on'.
'excludefun' Handle to a function that excludes undesirable
runs. If the function is f, it must support the syntax
b = f(S), where S is a matrix of treatments with
nfactors columns and b is a vector of Boolean
values with the same number of rows as S. b(i) is
true if the ith row S should be excluded.
'init' Initial design as an mruns-by-nfactors matrix. The
default is a randomly selected set of points.
'levels' Vector of number of levels for each factor.
18-305
dcovary
Parameter Value
'maxiter' Maximum number of iterations. The default is 10.
'tries' Number of times to try to generate a design from
a new starting point. The algorithm uses random
points for each try, except possibly the first. The
default is 1.
Examples Example 1
Suppose you want a design to estimate the parameters in a three-factor
linear additive model, with eight runs that necessarily occur at different
times. If the process experiences temporal linear drift, you may want
to include the run time as a variable in the model. Produce the design
as follows:
time = linspace(-1,1,8)';
[dCV1,X] = dcovary(3,time,'linear')
dCV1 =
-1.0000 1.0000 1.0000 -1.0000
1.0000 -1.0000 -1.0000 -0.7143
-1.0000 -1.0000 -1.0000 -0.4286
1.0000 -1.0000 1.0000 -0.1429
1.0000 1.0000 -1.0000 0.1429
-1.0000 1.0000 -1.0000 0.4286
1.0000 1.0000 1.0000 0.7143
-1.0000 -1.0000 1.0000 1.0000
X =
1.0000 -1.0000 1.0000 1.0000 -1.0000
1.0000 1.0000 -1.0000 -1.0000 -0.7143
1.0000 -1.0000 -1.0000 -1.0000 -0.4286
1.0000 1.0000 -1.0000 1.0000 -0.1429
1.0000 1.0000 1.0000 -1.0000 0.1429
1.0000 -1.0000 1.0000 -1.0000 0.4286
1.0000 1.0000 1.0000 1.0000 0.7143
1.0000 -1.0000 -1.0000 1.0000 1.0000
18-306
dcovary
The column vector time is a fixed factor, normalized to values between
±1. The number of rows in the fixed factor specifies the number of runs
in the design. The resulting design dCV gives factor settings for the
three controlled model factors at each time.
Example 2
The following example uses the dummyvar function to block an eight-run
experiment into 4 blocks of size 2 for estimating a linear additive model
with two factors:
fixed = dummyvar([1 1 2 2 3 3 4 4]);
dCV2 = dcovary(2,fixed(:,1:3),'linear')
dCV2 =
1 1 1 0 0
-1 -1 1 0 0
-1 1 0 1 0
1 -1 0 1 0
1 1 0 0 1
-1 -1 0 0 1
-1 1 0 0 0
1 -1 0 0 0
The first two columns of dCV2 contain the settings for the two factors;
the last three columns are dummy variable codings for the four blocks.
See Also daugment, cordexch, dummyvar
18-307
CompactTreeBagger.DefaultYfit property
Purpose Default value returned by predict
Description The DefaultYfit property controls what predicted value
CompactTreeBagger returns when no prediction is possible, for example
when the predict method needs to predict for an observation which has
only false values in the matrix supplied through 'useifort' argument.
For classification, you can set this property to either '' or
'MostPopular'. If you choose 'MostPopular' (default), the property
value becomes the name of the most probable class in the training data.
For regression, you can set this property to any numeric scalar. The
default is the mean of the response for the training data.
See Also predict, setDefaultYfit, TreeBagger.DefaultYfit.
18-308
TreeBagger.DefaultYfit property
Purpose Default value returned by predict and oobPredict
Description The DefaultYfit property controls what predicted value TreeBagger
returns when no prediction is possible, for example when the
oobPredict method needs to predict for an observation that is in-bag
for all trees in the ensemble.
For classification, you can set this property to either '' or
'MostPopular'. If you choose 'MostPopular' (default), the property
value becomes the name of the most probable class in the training data.
For regression, you can set this property to any numeric scalar. The
default is the mean of the response for the training data.
If you set this property to '' for classification or NaN for regression,
TreeBagger excludes the in-bag observations from computation of the
out-of-bagerror and margin.
See Also oobPredict, Predict, OOBIndices, CompactTreeBagger.DefaultYfit.
18-309
qrandstream.delete
Purpose Delete handle object
Syntax delete(h)
Description delete(h) deletes the handle object h, where h is a scalar handle. The
delete method deletes a handle object but does not clear the handle
from the workspace. A deleted handle is no longer valid.
See Also clear, isvalid, qrandstream
18-310
CompactTreeBagger.DeltaCritDecisionSplit property
Purpose Split criterion contributions for each predictor
Description The DeltaCritDecisionSplit property is a numeric array of size
1-by-Nvars of changes in the split criterion summed over splits on each
variable, averaged across the entire ensemble of grown trees.
See Also TreeBagger.DeltaCritDecisionSplit, classregtree.varimportance
18-311
TreeBagger.DeltaCritDecisionSplit property
Purpose Split criterion contributions for each predictor
Description The DeltaCritDecisionSplit property is a numeric array of size
1-by-Nvars of changes in the split criterion summed over splits on each
variable, averaged across the entire ensemble of grown trees.
See Also CompactTreeBagger.DeltaCritDecisionSplit,
classregtree.varimportance
18-312
dendrogram
Purpose Dendrogram plot
Syntax H = dendrogram(Z)
H = dendrogram(Z,p)
[H,T] = dendrogram(...)
[H,T,perm] = dendrogram(...)
[...] = dendrogram(...,'colorthreshold',t)
[...] = dendrogram(...,'orientation','orient')
[...] = dendrogram(...,'labels',S)
Description H = dendrogram(Z) generates a dendrogram plot of the hierarchical,
binary cluster tree represented by Z. Z is an (m-1)-by-3 matrix, generated
by the linkage function, where m is the number of objects in the
original data set. The output, H, is a vector of handles to the lines
in the dendrogram.
A dendrogram consists of many U-shaped lines connecting objects in a
hierarchical tree. The height of each U represents the distance between
the two objects being connected. If there were 30 or fewer data points
in the original dataset, each leaf in the dendrogram corresponds to one
data point. If there were more than 30 data points, the complete tree can
look crowded, and dendrogram collapses lower branches as necessary,
so that some leaves in the plot correspond to more than one data point.
H = dendrogram(Z,p) generates a dendrogram with no more than p
leaf nodes, by collapsing lower branches of the tree. To display the
complete tree, set p = 0.
[H,T] = dendrogram(...) generates a dendrogram and returns T, a
vector of length m that contains the leaf node number for each object in
the original data set. T is useful when p is less than the total number of
objects, so some leaf nodes in the display correspond to multiple objects.
For example, to find out which objects are contained in leaf node k of
the dendrogram, use find(T==k). When there are fewer than p objects
in the original data, all objects are displayed in the dendrogram. In this
case, T is the identity map, i.e., T = (1:m)', where each node contains
only a single object.
18-313
dendrogram
[H,T,perm] = dendrogram(...) generates a dendrogram and
returns the permutation vector of the node labels of the leaves of
the dendrogram. perm is ordered from left to right on a horizontal
dendrogram and bottom to top for a vertical dendrogram.
[...] = dendrogram(...,'colorthreshold',t) assigns a unique
color to each group of nodes in the dendrogram where the linkage is less
than the threshold t. t is a value in the interval [0,max(Z(:,3))].
Setting t to the string 'default' is the same as t = .7(max(Z(:,3))).
0 is the same as not specifying 'colorthreshold'. The value
max(Z(:,3)) treats the entire tree as one group and colors it all one
color.
[...] = dendrogram(...,'orientation','orient') orients the
dendrogram within the figure window. Acceptable values for ’orient’
are:
Value Description
'top' Top to bottom (default)
'bottom' Bottom to top
'left' Left to right
'right' Right to left
[...] = dendrogram(...,'labels',S) accepts a character array
or cell array of strings S with one label for each observation. Any
leaves in the tree containing a single observation are labeled with that
observation’s label.
Examples X = rand(100,2);
Y = pdist(X,'cityblock');
Z = linkage(Y,'average');
[H,T] = dendrogram(Z,'colorthreshold','default');
set(H,'LineWidth',2)
18-314
dendrogram
find(T==20)
ans =
20
49
62
65
73
96
This output indicates that leaf node 20 in the dendrogram contains the
original data points 20, 49, 62, 65, 73, and 96.
See Also cluster, clusterdata, cophenet, inconsistent, linkage, silhouette
18-315
dataset.Description property
Purpose String describing data set
Description Description is a string describing the data set. The default is an
empty string.
18-316
dfittool
Purpose Interactive distribution fitting
Syntax dfittool
dfittool(y)
dfittool(y,cens)
dfittool(y,cens,freq)
dfittool(y,cens,freq,dsname)
Description dfittool opens a graphical user interface for displaying fit distributions
to data. To fit distributions to your data and display them over plots
over plots of the empirical distributions, you can import data from the
workspace.
dfittool(y) displays the Distribution Fitting Tool and creates a data
set with data specified by the vector y.
dfittool(y,cens) uses the vector cens to specify whether the
observation y(j) is censored, (cens(j)==1) and/or observed, exactly
(cens(j)==0). If cens is omitted or empty, no y values are censored.
dfittool(y,cens,freq) uses the vector freq to specify the frequency
of each element of y. If freq is omitted or empty, all y values have a
frequency of 1.
dfittool(y,cens,freq,dsname) creates a data set with the name
dsname using the data vector y, censoring indicator cens, and frequency
vector freq.
For more information, see “Modeling Your Data Using the Distribution
Fitting GUI” on page 5-11.
See Also mle,randtool,disttool
18-317
qrandset.Dimensions property
Purpose Number of dimensions
Description Number of dimensions in the point set. The Dimensions property of
a point set contains a positive integer that indicates the number of
dimensions for which the points have values. For example, a point set
with Dimensions=5 produces points that each have five values.
Set this property by specifying the number of dimensions when
constructing a new point set. After construction, you cannot change the
value. The default number of dimensions is 2.
18-318
dataset.DimNames property
Purpose Two-element cell array of strings giving names of dimensions of data set
Description A two-element cell array of strings giving the names of the two
dimensions of the data set. The default is {'Observations'
'Variables'}.
18-319
categorical.disp
Purpose Display categorical array
Syntax disp(A)
Description disp(A) prints the categorical array A without printing the array
name. In all other ways it’s the same as leaving the semicolon off an
expression, except that empty arrays don’t display.
See Also categorical, display
18-320
classregtree.disp
Purpose Display classregtree object
Syntax display(t)
Description display(t) prints the classregtree object t.
See Also classregtree, view
18-321
cvpartition.disp
Purpose Display cvpartition object
Syntax disp(c)
Description disp(c) prints the cvpartition object c.
See Also cvpartition
18-322
dataset.disp
Purpose Display dataset array
Syntax disp(ds)
Description disp(ds) prints the dataset array ds, including variable names and
observation names (if present), without printing the dataset name. In
all other ways it’s the same as leaving the semicolon off an expression.
For numeric or categorical variables that are 2-D and have three or
fewer columns, disp prints the actual data using either short g, long
g, or bank format, depending on the current command line setting.
Otherwise, disp prints the size and type of each dataset element.
For character variables that are 2-D and 10 or fewer characters wide,
disp prints quoted strings. Otherwise, disp prints the size and type of
each dataset element.
For cell variables that are 2-D and have three or fewer columns,
disp prints the contents of each cell (or its size and type if too large).
Otherwise, disp prints the size of each dataset element.
For time series variables, disp prints columns for both the time and
the data. If the variable is 2-D and has three or fewer columns, disp
prints the actual data Otherwise, disp prints the size and type of each
dataset element.
For other types of variables, disp prints the size and type of each
dataset element.
See Also dataset | display | format
18-323
gmdistribution.disp
Purpose Display Gaussian mixture distribution object
Syntax disp(obj)
Description disp(obj) prints a text representation of the gmdistribution object,
obj, without printing the object name. In all other ways it’s the same as
leaving the semicolon off an expression.
See Also gmdistribution, display
18-324
NaiveBayes.disp
Purpose Display NaiveBayes classifier object
Syntax disp(nb)
Description disp(nb) prints a text representation of the NaiveBayes object nb,
without printing the object name. In all other ways it’s the same as
leaving the semicolon off an expression.
See Also NaiveBayes, display
18-325
piecewisedistribution.disp
Purpose Display piecewisedistribution object
Syntax disp(A)
Description disp(A) prints a text representation of the piecewisedistribution object
A, without printing the object name. In all other ways it’s the same as
leaving the semicolon off an expression.
See Also piecewisedistribution
18-326
qrandset.disp
Purpose Display qrandset object
Syntax disp(p)
Description disp(p) displays the properties of the quasi-random point set s,
without printing the variable name. disp prints out the number of
dimensions and points in the point-set, and follows this with the list of
all property values for the object.
See Also qrandset
18-327
qrandstream.disp
Purpose Display qrandstream object
Syntax disp(q)
Description disp(q) displays the quasi-random stream q, without printing the
variable name. disp prints the type and number of dimensions in the
stream, and follows it with the list of point set properties.
See Also qrandstream
18-328
categorical.display
Purpose Display categorical array
Syntax display(A)
Description display(A) prints the categorical array A. categorical callsdisplay
when a you do not use a semicolon to terminate a statement.
See Also categorical, disp
18-329
classregtree.display
Purpose Display classregtree object
Syntax display(t)
display(A)
Description display(t) prints the classregtree object t. classregtree
callsdisplay when a you do not use a semicolon to terminate a
statement.
display(A) prints the categorical array A. categorical callsdisplay
when a you do not use a semicolon to terminate a statement.
See Also classregtree, eval, prune, test
18-330
cvpartition.display
Purpose Display cvpartition object
Syntax display(c)
Description display(c) prints the cvpartition object c. cvpartition callsdisplay
when a you do not use a semicolon to terminate a statement.
See Also cvpartition
18-331
dataset.display
Purpose Display dataset array
Syntax display(ds)
Description display(ds) prints the dataset array ds, including variable names and
observation names (if present). dataset callsdisplay when a you do
not use a semicolon to terminate a statement
For numeric or categorical variables that are 2-D and have three or
fewer columns, display prints the actual data. Otherwise, display
prints the size and type of each dataset element.
For character variables that are 2-D and 10 or fewer characters wide,
display prints quoted strings. Otherwise, display prints the size and
type of each dataset element.
For cell variables that are 2-D and have three or fewer columns,
display prints the contents of each cell (or its size and type if too large).
Otherwise, display prints the size of each dataset element.
For time series variables, display prints columns for both the time and
the data. If the variable is 2-D and has three or fewer columns, display
prints the actual data. Otherwise, display prints the size and type of
each dataset element.
For other types of variables, display prints the size and type of each
dataset element.
See Also dataset, display, format
18-332
gmdistribution.display
Purpose Display Gaussian mixture distribution object
Syntax display(obj)
Description display(obj) prints a text representation of the gmdistribution
object obj. gmdistribution callsdisplay when a you do not use a
semicolon to terminate a statement.
See Also gmdistribution, disp
18-333
NaiveBayes.display
Purpose Display NaiveBayes classifier object
Syntax display(nb)
Description display(nb) prints a text representation of the NaiveBayes object
nb. NaiveBayes callsdisplay when a you do not use a semicolon to
terminate a statement.
See Also NaiveBayes, display
18-334
piecewisedistribution.display
Purpose Display piecewisedistribution object
Syntax display(A)
Description display(A) prints a text representation of the piecewisedistribution
object A, without printing the object name. piecewisedistribution
callsdisplay when a you do not use a semicolon to terminate a
statement.
See Also piecewisedistribution
18-335
ProbDist.DistName property
Purpose Read-only string containing probability distribution name of ProbDist
object
Description DistName is a read-only property of the ProbDist class. DistName is a
string containing the type of distribution used to create the object.
Values Possible values are:
• 'kernel'
• 'beta'
• 'binomial'
• 'birnbaumsaunders'
• 'exponential'
• 'extreme value'
• 'gamma'
• 'generalized extreme value'
• 'generalized pareto'
• 'inversegaussian'
• 'logistic'
• 'loglogistic'
• 'lognormal'
• 'nakagami'
• 'negative binomial'
• 'normal'
• 'poisson'
• 'rayleigh'
• 'rician'
18-336
ProbDist.DistName property
• 'tlocationscale'
• 'weibull'
Use this information to view and compare the type of distribution used
to create distribution objects.
18-337
NaiveBayes.Dist property
Purpose Distribution names
Description The Dist property is a string or a 1-by-NDims cell array of strings
indicating the types of distributions for all the features. If all the
features use the same type of distribution, Dist is a single string.
Otherwise Dist(j) indicates the distribution type used for the jth
feature.
The valid strings for this property are the following:
'normal' Normal distribution.
'kernel' Kernel smoothing density
estimate.
'mvmn' Multivariate multinomial
distribution.
'mn' Multinomial bag-of-tokens model.
18-338
gmdistribution.DistName property
Purpose Type of distribution
Description The string 'gaussian mixture distribution'.
18-339
disttool
Purpose Interactive density and distribution plots
Syntax disttool
Description disttool is a graphical interface for exploring the effects of changing
parameters on the plot of a cdf or pdf.
See Also randtool, dfittool
18-340
categorical.double
Purpose Convert categorical array to double array
Syntax B = double(A)
Description B = double(A) converts the categorical array A to a double array.
Each element of B contains the internal categorical level code for the
corresponding element of A.
See Also single
18-341
dataset.double
Purpose Convert dataset variables to double array
Syntax b = double(A)
b = double(a,vars)
Description b = double(A) returns the contents of the dataset A, converted to one
double array. The classes of the variables in the dataset must support
the conversion.
b = double(a,vars) returns the contents of the dataset variables
specified by vars. vars is a positive integer, a vector of positive
integers, a variable name, a cell array containing one or more variable
names, or a logical vector.
See Also dataset, single, replacedata
18-342
categorical.droplevels
Purpose Drop levels
Syntax B = droplevels(A)
B = droplevels(A,oldlevels)
Description B = droplevels(A) removes unused levels from the categorical array
A. B is a categorical array with the same size and values as A, but with a
list of potential levels that includes only those present in some element
of A.
B = droplevels(A,oldlevels) removes specified levels from the
categorical array A. oldlevels is a cell array of strings or a 2-D
character matrix specifying the levels to be removed.
droplevels removes levels, but does not remove elements. Elements
of B that correspond to elements of A having levels in oldlevels all
have an undefined level.
Examples Example 1
Drop unused age levels from the data in hospital.mat:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
AgeGroup = ordinal(hospital.Age,labels,[],edges);
AgeGroup = droplevels(AgeGroup);
getlabels(AgeGroup)
ans =
'20s' '30s' '40s' '50s'
Example 2
1 Load patient data from the CSV file hospital.dat and store the
information in a dataset array with observation names given by the
first column in the data (patient identification):
patients = dataset('file','hospital.dat',...
'delimiter',',',...
18-343
categorical.droplevels
'ReadObsNames',true);
2 Make the {0,1}-valued variable smoke nominal, and change the labels
to 'No' and 'Yes':
patients.smoke = nominal(patients.smoke,{'No','Yes'});
3 Add new levels to smoke as placeholders for more detailed histories
of smokers:
patients.smoke = addlevels(patients.smoke,...
{'0-5 Years','5-10 Years','LongTerm'});
4 Assuming the nonsmokers have never smoked, relabel the 'No' level:
patients.smoke = setlabels(patients.smoke,'Never','No');
5 Drop the undifferentiated 'Yes' level from smoke:
patients.smoke = droplevels(patients.smoke,'Yes');
Warning: OLDLEVELS contains categorical levels that
were present in A, caused some array elements to have
undefined levels.
Note that smokers now have an undefined level.
6 Set each smoker to one of the new levels, by observation name:
patients.smoke('YPL-320') = '5-10 Years';
See Also addlevels, getlabels, islevel, mergelevels, reorderlevels
18-344
dummyvar
Purpose Create dummy variables
Syntax D = dummyvar(group)
Description D = dummyvar(group) creates {0,1}-valued dummy variables for each
of the unique values in group. Columns of group represent categorical
predictor variables, with values indicating categorical levels. Rows of
group represent observations across variables. Each column of D is a
dummy variable for one categorical level of one of the variables in group.
group can be a numeric vector or categorical column vector, representing
levels within a single variable, or a numeric matrix or cell array
of categorical column vectors, representing levels within multiple
variables. If group is a numeric vector or matrix, values in any column
must be positive integers in the range from 1 to the number of levels for
the corresponding variable.
If group is n-by-p, D is n-by-S, where S is the sum of the number of
levels in each of the columns of group. The number of levels s in any
column of group is the maximum positive integer in the column or the
number of categorical levels. Levels are considered distinct if they
appear in different columns of group, even if they have the same value.
Columns of D are, from left to right, dummy variables created from
the first column of group, followed by dummy variables created from
the second column of group, etc.
dummyvar treats NaN values or undefined categorical levels in group as
missing data and returns NaN values in D.
Dummy variables are used in regression analysis and ANOVA to
indicate values of categorical predictors.
18-345
dummyvar
Note If a column of 1s is introduced in the matrix D, the resulting
matrix X = [ones(size(D,1),1) D] will be rank deficient. The
matrix D itself will be rank deficient if group has multiple columns.
This is because dummy variables produced from any column of group
always sum to a column of 1s. Regression and ANOVA calculations
often address this issue by eliminating one dummy variable (implicitly
setting the coefficients for dropped columns to zero) from each group of
dummy variables produced by a column of group.
Examples Suppose you are studying the effects of two machines and three
operators on a process. Use group to organize predictor data on
machine-operator combinations:
machine = [1 1 1 1 2 2 2 2]';
operator = [1 2 3 1 2 3 1 2]';
group = [machine operator]
group =
1 1
1 2
1 3
1 1
2 2
2 3
2 1
2 2
Use dummyvar to create dummy variables for a regression or ANOVA
calculation:
D = dummyvar(group)
D =
1 0 1 0 0
1 0 0 1 0
1 0 0 0 1
1 0 1 0 0
18-346
dummyvar
0 1 0 1 0
0 1 0 0 1
0 1 1 0 0
0 1 0 1 0
The first two columns of D represent observations of machine 1 and
machine 2, respectively; the remaining columns represent observations
of the three operators.
See Also “Grouped Data” on page 2-34
regress, anova1
18-347
dwtest
Purpose Durbin-Watson test
Syntax [P,DW] = dwtest(R,X)
[...] = dwtest(R,X,method)
[...] = dwtest(R,X,method,tail)
Description [P,DW] = dwtest(R,X) performs a Durbin-Watson test on the vector R
of residuals from a linear regression, where X is the design matrix from
that linear regression. P is the computed p value for the test, and DW
is the Durbin-Watson statistic. The Durbin-Watson test is used to test
if the residuals are uncorrelated, against the alternative that there is
autocorrelation among them.
[...] = dwtest(R,X,method) specifies the method to be used in
computing the p value. method can be either of the following:
• 'exact' — Calculates an exact p value using the PAN algorithm (the
default if the sample size is less than 400).
• 'approximate' — Calculates the p value using a normal
approximation (the default if the sample size is 400 or larger).
[...] = dwtest(R,X,method,tail) performs the test against one of
the following alternative hypotheses, specified by tail:
Tail Alternative Hypothesis
'both' Serial correlation is not 0.
'right' Serial correlation is greater than 0 (right-tailed test).
'left' Serial correlation is less than 0 (left-tailed test).
Examples Fit a straight line to the census data and note the autocorrelation in
the residuals:
load census
n = length(cdate);
X = [ones(n,1),cdate];
18-348
dwtest
[b,bint,r1] = regress(pop,X);
p1 = dwtest(r1,X)
plot(cdate,r1,'b-',cdate,zeros(n,1),'k:')
Adding a squared term reduces the autocorrelation but it is still
significantly different from zero:
X = [ones(n,1),cdate,cdate.^2];
[b,bint,r2] = regress(pop,X);
p2 = dwtest(r2,X)
line(cdate,r2,'color','r')
See Also regress
18-349
ecdf
Purpose Empirical cumulative distribution function
Syntax [f,x] = ecdf(y)
[f,x,flo,fup] = ecdf(y)
ecdf(...)
ecdf(ax,...)
[...] = ecdf(y,param1,val1,param2,val2,...)
Description [f,x] = ecdf(y) calculates the Kaplan-Meier estimate of the
cumulative distribution function (cdf), also known as the empirical cdf.
y is a vector of data values. f is a vector of values of the empirical cdf
evaluated at x.
[f,x,flo,fup] = ecdf(y) also returns lower and upper confidence
bounds for the cdf. These bounds are calculated using Greenwood’s
formula, and are not simultaneous confidence bounds.
ecdf(...) without output arguments produces a plot of the empirical
cdf.
ecdf(ax,...) plots into axes ax instead of gca.
[...] = ecdf(y,param1,val1,param2,val2,...) specifies
additional parameter/value pairs chosen from the following:
Parameter Value
'censoring' Boolean vector of the same size as x. Elements are
1 for observations that are right-censored and 0 for
observations that are observed exactly. Default is all
observations observed exactly.
'frequency' Vector of the same size as x containing nonnegative
integer counts. The jth element of this vector
gives the number of times the jth element of x was
observed. Default is 1 observation per element of x.
'alpha' Value between 0 and 1 for a confidence level of
100(1-alpha)%. Default is alpha=0.05 for 95%
confidence.
18-350
ecdf
Parameter Value
'function' Type of function returned as the f output argument,
chosen from 'cdf' (default), 'survivor', or
'cumulative hazard'.
'bounds' Either 'on' to include bounds, or 'off' (the default)
to omit them. Used only for plotting.
Examples Generate random failure times and random censoring times, and
compare the empirical cdf with the known true cdf:
y = exprnd(10,50,1); % Random failure times exponential(10)
d = exprnd(20,50,1); % Drop-out times exponential(20)
t = min(y,d); % Observe the minimum of these times
censored = (y>d); % Observe whether the subject failed
% Calculate and plot empirical cdf and confidence bounds
[f,x,flo,fup] = ecdf(t,'censoring',censored);
stairs(x,f,'LineWidth',2)
hold on
stairs(x,flo,'r:','LineWidth',2)
stairs(x,fup,'r:','LineWidth',2)
% Superimpose a plot of the known population cdf
xx = 0:.1:max(t);
yy = 1-exp(-xx/10);
plot(xx,yy,'g-','LineWidth',2)
legend('Empirical','LCB','UCB','Population',...
'Location','SE')
hold off
18-351
ecdf
References [1] Cox, D. R., and D. Oakes. Analysis of Survival Data. London:
Chapman & Hall, 1984.
See Also cdfplot, ecdfhist
18-352
ecdfhist
Purpose Empirical cumulative distribution function histogram
Syntax n = ecdfhist(f,x)
n = ecdfhist(f,x,m)
n = ecdfhist(f,x,c)
[n,c] = ecdfhist(...)
ecdfhist(...)
Description n = ecdfhist(f,x) takes a vector f of empirical cumulative
distribution function (cdf) values and a vector x of evaluation points,
and returns a vector n containing the heights of histogram bars for 10
equally spaced bins. The function computes the bar heights from the
increases in the empirical cdf, and normalizes them so that the area of
the histogram is equal to 1. In contrast, hist produces bars whose
heights represent bin counts.
n = ecdfhist(f,x,m), where m is a scalar, uses m bins.
n = ecdfhist(f,x,c), where c is a vector, uses bins with centers
specified by c.
[n,c] = ecdfhist(...) also returns the position of the bin centers
in c.
ecdfhist(...) without output arguments produces a histogram bar
plot of the results.
Examples The following code generates random failure times and random
censoring times, and compares the empirical pdf with the known true
pdf.
y = exprnd(10,50,1); % Random failure times
d = exprnd(20,50,1); % Drop-out times
t = min(y,d); % Observe the minimum of these times
censored = (y>d); % Observe whether the subject failed
% Calculate the empirical cdf and plot a histogram from it
[f,x] = ecdf(t,'censoring',censored);
ecdfhist(f,x)
18-353
ecdfhist
set(get(gca,'Children'),'FaceColor',[.8 .8 1])
hold on
% Superimpose a plot of the known population pdf
xx = 0:.1:max(t);
yy = exp(-xx/10)/10;
plot(xx,yy,'r-','LineWidth',2)
hold off
See Also ecdf, hist, histc
18-354
categorical.end
Purpose Last index in indexing expression for categorical array
Syntax end(A,k,n)
Description end(A,k,n) indexes expressions involving the categorical array A
when end is part of the k-th index out of n indices. For example, the
expression A(end-1,:) calls A’s end method with end(A,1,2).
See Also single
18-355
dataset.end
Purpose Last index in indexing expression for dataset array
Syntax end(A,k,n)
Description end(A,k,n) is called for indexing expressions involving the dataset A
when end is part of the k-th index out of n indices. For example, the
expression A(end-1,:) calls A’s end method with end(A,1,2).
See Also size
18-356
qrandset.end
Purpose Last index in indexing expression for point set
Syntax end(p,k,n)
Description end(p,k,n) is called for indexing expressions involving the point set
p when end is part of the k-th index out of n indices. For example, the
expression p(end-1,:) calls p’s end method with end(p,1,2).
See Also qrandset
18-357
evcdf
Purpose Extreme value cumulative distribution function
Syntax P = evcdf(X,mu,sigma)
[P,PLO,PUP] = evcdf(X,mu,sigma,pcov,alpha)
Description P = evcdf(X,mu,sigma) computes the cumulative distribution function
(cdf) for the type 1 extreme value distribution, with location parameter
mu and scale parameter sigma, at each of the values in X. X, mu, and
sigma can be vectors, matrices, or multidimensional arrays that all
have the same size. A scalar input is expanded to a constant array of
the same size as the other inputs. The default values for mu and sigma
are 0 and 1, respectively.
[P,PLO,PUP] = evcdf(X,mu,sigma,pcov,alpha) produces confidence
bounds for P when the input parameters mu and sigma are estimates.
pcov is a 2-by-2 covariance matrix of the estimated parameters. alpha
has a default value of 0.05, and specifies 100(1 - alpha)% confidence
bounds. PLO and PUP are arrays of the same size as P, containing the
lower and upper confidence bounds.
The function evcdf computes confidence bounds for P using a normal
approximation to the distribution of the estimate
X − ˆ
ˆ
and then transforming those bounds to the scale of the output P. The
computed bounds give approximately the desired confidence level when
you estimate mu, sigma, and pcov from large samples, but in smaller
samples other methods of computing the confidence bounds might be
more accurate.
The type 1 extreme value distribution is also known as the Gumbel
distribution. The version used here is suitable for modeling minima;
the mirror image of this distribution can be used to model maxima by
negating X. See “Extreme Value Distribution” on page B-19 for more
details. If x has a Weibull distribution, then X = log(x) has the type 1
extreme value distribution.
18-358
evcdf
See Also cdf, evpdf, evinv, evstat, evfit, evlike, evrnd
“Extreme Value Distribution” on page B-19
18-359
evfit
Purpose Extreme value parameter estimates
Syntax parmhat = evfit(data)
[parmhat,parmci] = evfit(data)
[parmhat,parmci] = evfit(data,alpha)
[...] = evfit(data,alpha,censoring)
[...] = evfit(data,alpha,censoring,freq)
[...] = evfit(data,alpha,censoring,freq,options)
Description parmhat = evfit(data) returns maximum likelihood estimates of
the parameters of the type 1 extreme value distribution given the
data in the vector data. parmhat(1) is the location parameter, , and
parmhat(2) is the scale parameter, σ.
[parmhat,parmci] = evfit(data) returns 95% confidence intervals
for the parameter estimates on the and σ parameters in the 2-by-2
matrix parmci. The first column of the matrix of the extreme value
fit contains the lower and upper confidence bounds for the parameter
, and the second column contains the confidence bounds for the
parameter σ.
[parmhat,parmci] = evfit(data,alpha) returns 100(1 - alpha)%
confidence intervals for the parameter estimates, where alpha is a value
in the range [0 1] specifying the width of the confidence intervals. By
default, alpha is 0.05, which corresponds to 95% confidence intervals.
[...] = evfit(data,alpha,censoring) accepts a Boolean vector,
censoring, of the same size as data, which is 1 for observations that
are right-censored and 0 for observations that are observed exactly.
[...] = evfit(data,alpha,censoring,freq) accepts a frequency
vector, freq of the same size as data. Typically, freq contains integer
frequencies for the corresponding elements in data, but can contain
any nonnegative values. Pass in [] for alpha, censoring, or freq to
use their default values.
[...] = evfit(data,alpha,censoring,freq,options) accepts
a structure, options, that specifies control parameters for the
iterative algorithm the function uses to compute maximum likelihood
18-360
evfit
estimates. You can create options using the function statset.
Enter statset('evfit') to see the names and default values of the
parameters that evfit accepts in the options structure. See the
reference page for statset for more information about these options.
The type 1 extreme value distribution is also known as the Gumbel
distribution. The version used here is suitable for modeling minima;
the mirror image of this distribution can be used to model maxima by
negating X. See “Extreme Value Distribution” on page B-19 for more
details. If x has a Weibull distribution, then X = log(x) has the type 1
extreme value distribution.
See Also mle, evlike, evpdf, evcdf, evinv, evstat, evrnd
“Extreme Value Distribution” on page B-19
18-361
evinv
Purpose Extreme value inverse cumulative distribution function
Syntax X = evinv(P,mu,sigma)
[X,XLO,XUP] = evinv(P,mu,sigma,pcov,alpha)
Description X = evinv(P,mu,sigma) returns the inverse cumulative distribution
function (cdf) for a type 1 extreme value distribution with location
parameter mu and scale parameter sigma, evaluated at the values in P.
P, mu, and sigma can be vectors, matrices, or multidimensional arrays
that all have the same size. A scalar input is expanded to a constant
array of the same size as the other inputs. The default values for mu and
sigma are 0 and 1, respectively.
[X,XLO,XUP] = evinv(P,mu,sigma,pcov,alpha) produces confidence
bounds for X when the input parameters mu and sigma are estimates.
pcov is the covariance matrix of the estimated parameters. alpha
is a scalar that specifies 100(1 – alpha)% confidence bounds for the
estimated parameters, and has a default value of 0.05. XLO and XUP are
arrays of the same size as X containing the lower and upper confidence
bounds.
The function evinv computes confidence bounds for P using a normal
approximation to the distribution of the estimate
ˆ + ˆ q
where q is the Pth quantile from an extreme value distribution with
parameters μ = 0 and σ = 1. The computed bounds give approximately
the desired confidence level when you estimate mu, sigma, and pcov
from large samples, but in smaller samples other methods of computing
the confidence bounds might be more accurate.
The type 1 extreme value distribution is also known as the Gumbel
distribution. The version used here is suitable for modeling minima;
the mirror image of this distribution can be used to model maxima by
negating X. See “Extreme Value Distribution” on page B-19 for more
details. If x has a Weibull distribution, then X = log(x) has the type 1
extreme value distribution.
18-362
evinv
See Also icdf, evcdf, evpdf, evstat, evfit, evlike, evrnd
18-363
qrandstream.eq
Purpose Test handle equality
Syntax h1 == h2
tf = eq(h1, h2)
Description h1 == h2 performs element-wise comparisons between handle arrays
h1 and h2. h1 and h2 must be of the same dimensions unless one is a
scalar. The result is a logical array of the same dimensions, where each
element is an element-wise equality result. If one of h1 or h2 is scalar,
scalar expansion is performed and the result will match the dimensions
of the array that is not scalar.
tf = eq(h1, h2) stores the result in a logical array of the same
dimensions.
See Also qrandstream, ge, gt, le, lt, ne
18-364
CompactTreeBagger.error
Purpose Error (misclassification probability or MSE)
Syntax err = error(B,X,Y)
err = error(B,X,Y,'param1',val1,'param2',val2,...)
Description err = error(B,X,Y) computes the misclassification probability (for
classification trees) or mean squared error (MSE, for regression trees)
for each tree, for predictors X given true response Y. For classification, Y
can be either a numeric vector, character matrix, cell array of strings,
categorical vector or logical vector. For regression, Y must be a numeric
vector. err is a vector with one error measure for each of the NTrees
trees in the ensemble B.
err = error(B,X,Y,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'mode' String indicating how the method computes errors.
If set to 'cumulative' (default), error computes
cumulative errors and err is a vector of length NTrees,
where the first element gives error from trees(1),
second element gives error fromtrees(1:2) etc, up
to trees(1:NTrees). If set to 'individual', err
is a vector of length NTrees, where each element is
an error from each tree in the ensemble. If set to
'ensemble', err is a scalar showing the cumulative
error for the entire ensemble.
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements in the
input vector, and a scalar for 'ensemble' mode. For
example, in the 'cumulative' mode, the first element
gives error from trees(1), the second element gives
error from trees(1:2) etc.
18-365
CompactTreeBagger.error
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. The method uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple non-weighted majority vote. You cannot use
this argument in the 'individual' mode.
'useifort' Logical matrix of size Nobs-by-NTrees indicating
which trees should be used to make predictions for
each observation. By default the method uses all trees
for all observations.
See Also TreeBagger.error
18-366
TreeBagger.error
Purpose Error (misclassification probability or MSE)
Syntax err = error(B,X,Y)
err = error(B,X,Y,'param1',val1,'param2',val2,...)
Description err = error(B,X,Y) computes the misclassification probability for
classification trees or mean squared error (MSE) for regression trees for
each tree, for predictors X given true response Y. For classification, Y
can be either a numeric vector, character matrix, cell array of strings,
categorical vector or logical vector. For regression, Y must be a numeric
vector. err is a vector with one error measure for each of the NTrees
trees in the ensemble B.
err = error(B,X,Y,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'mode' String indicating how the method computes errors.
If set to 'cumulative' (default), error computes
cumulative errors and err is a vector of length NTrees,
where the first element gives error from trees(1),
second element gives error fromtrees(1:2) etc, up
to trees(1:NTrees). If set to 'individual', err
is a vector of length NTrees, where each element is
an error from each tree in the ensemble. If set to
'ensemble', err is a scalar showing the cumulative
error for the entire ensemble.
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements in the
input vector, and a scalar for 'ensemble' mode. For
example, in the 'cumulative' mode, the first element
gives error from trees(1), the second element gives
error from trees(1:2) etc.
18-367
TreeBagger.error
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. The method uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple non-weighted majority vote. You cannot use
this argument in the 'individual' mode.
'useifort' Logical matrix of size Nobs-by-NTrees indicating
which trees should be used to make predictions for
each observation. By default the method uses all trees
for all observations.
See Also CompactTreeBagger.error
18-368
classregtree.eval
Purpose Predicted responses
Syntax yfit = eval(t,X)
yfit = eval(t,X,s)
[yfit,nodes] = eval(...)
[yfit,nodes,cnums] = eval(...)
[...] = t(X)
[...] = t(X,s)
Description yfit = eval(t,X) takes a classification or regression tree t and a
matrix X of predictors, and produces a vector yfit of predicted response
values. For a regression tree, yfit(i) is the fitted response value for
a point having the predictor values X(i,:). For a classification tree,
yfit(i) is the class into which the tree assigns the point with data
X(i,:).
yfit = eval(t,X,s) takes an additional vector s of pruning levels,
with 0 representing the full, unpruned tree. t must include a pruning
sequence as created by classregtree or by prune. If s has k elements
and X has n rows, the output yfit is an n-by-k matrix, with the jth
column containing the fitted values produced by the s(j) subtree. s
must be sorted in ascending order.
To compute fitted values for a tree that is not part of the optimal
pruning sequence, first use prune to prune the tree.
[yfit,nodes] = eval(...) also returns a vector nodes the same size
as yfit containing the node number assigned to each row of X. Use view
to display the node numbers for any node you select.
[yfit,nodes,cnums] = eval(...) is valid only for classification trees.
It returns a vector cnum containing the predicted class numbers.
NaN values in X are treated as missing. If eval encounters a missing
value when it attempts to evaluate the split rule at a branch node, it
cannot determine whether to proceed to the left or right child node.
Instead, it sets the corresponding fitted value equal to the fitted value
assigned to the branch node.
[...] = t(X) or [...] = t(X,s) also invoke eval.
18-369
classregtree.eval
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-370
classregtree.eval
Find assigned class names:
sfit = eval(t,meas);
Compute that proportion is correctly classified:
pct = mean(strcmp(sfit,species))
pct =
0.9800
18-371
classregtree.eval
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, prune, test, view
18-372
evlike
Purpose Extreme value negative log-likelihood
Syntax nlogL = evlike(params,data)
[nlogL,AVAR] = evlike(params,data)
[...] = evlike(params,data,censoring)
[...] = evlike(params,data,censoring,freq)
Description nlogL = evlike(params,data) returns the negative of the
log-likelihood for the type 1 extreme value distribution. params(1) is
the tail location parameter, mu, and params(2) is the scale parameter,
sigma. nlogL is a scalar.
[nlogL,AVAR] = evlike(params,data) returns the inverse of Fisher’s
information matrix, AVAR. If the input parameter values in params are
the maximum likelihood estimates, the diagonal elements of AVAR are
their asymptotic variances. AVAR is based on the observed Fisher’s
information, not the expected information.
[...] = evlike(params,data,censoring) accepts a Boolean
vector of the same size as data, which is 1 for observations that are
right-censored and 0 for observations that are observed exactly.
[...] = evlike(params,data,censoring,freq) accepts a frequency
vector of the same size as data. freq typically contains integer
frequencies for the corresponding elements in data, but can contain any
nonnegative values. Pass in [] for censoring to use its default value.
The type 1 extreme value distribution is also known as the Gumbel
distribution. The version used here is suitable for modeling minima;
the mirror image of this distribution can be used to model maxima by
negating data. See “Extreme Value Distribution” on page B-19 for more
details. If x has a Weibull distribution, then X = log(x) has the type 1
extreme value distribution.
See Also evfit, evpdf, evcdf, evinv, evstat, evrnd
“Extreme Value Distribution” on page B-19
18-373
evpdf
Purpose Extreme value probability density function
Syntax Y = evpdf(X,mu,sigma)
Description Y = evpdf(X,mu,sigma) returns the pdf of the type 1 extreme value
distribution with location parameter mu and scale parameter sigma,
evaluated at the values in X. X, mu, and sigma can be vectors, matrices,
or multidimensional arrays that all have the same size. A scalar input
is expanded to a constant array of the same size as the other inputs.
The default values for mu and sigma are 0 and 1, respectively.
The type 1 extreme value distribution is also known as the Gumbel
distribution. The version used here is suitable for modeling minima;
the mirror image of this distribution can be used to model maxima by
negating X. See “Extreme Value Distribution” on page B-19 for more
details. If x has a Weibull distribution, then X = log(x) has the type 1
extreme value distribution.
See Also pdf, evcdf, evinv, evstat, evfit, evlike, evrnd
“Extreme Value Distribution” on page B-19
18-374
evrnd
Purpose Extreme value random numbers
Syntax R = evrnd(mu,sigma)
R = evrnd(mu,sigma,v)
R = evrnd(mu,sigma,m,n)
Description R = evrnd(mu,sigma) generates random numbers from the extreme
value distribution with parameters specified by location parameter mu
and scale parameter sigma. mu and sigma can be vectors, matrices, or
multidimensional arrays that have the same size, which is also the size
of R. A scalar input for mu or sigma is expanded to a constant array with
the same dimensions as the other input.
R = evrnd(mu,sigma,v) generates an array R of size v containing
random numbers from the extreme value distribution with parameters
mu and sigma, where v is a row vector. If v is a 1-by-2 vector, R is
a matrix with v(1) rows and v(2) columns. If v is 1-by-n, R is an
n-dimensional array.
If mu and sigma are both scalars, R = evrnd(mu,sigma,m,n) returns an
m-by-n matrix.
The type 1 extreme value distribution is also known as the Gumbel
distribution. The version used here is suitable for modeling minima;
the mirror image of this distribution can be used to model maxima by
negating R. See “Extreme Value Distribution” on page B-19 for more
details. If x has a Weibull distribution, then X = log(x) has the type 1
extreme value distribution.
See Also random, evpdf, evcdf, evinv, evstat, evfit, evlike
“Extreme Value Distribution” on page B-19
18-375
evstat
Purpose Extreme value mean and variance
Syntax [M,V] = evstat(mu,sigma)
Description [M,V] = evstat(mu,sigma) returns the mean of and variance for
the type 1 extreme value distribution with location parameter mu and
scale parameter sigma. mu and sigma can be vectors, matrices, or
multidimensional arrays that all have the same size. A scalar input is
expanded to a constant array of the same size as the other input. The
default values for mu and sigma are 0 and 1, respectively.
The type 1 extreme value distribution is also known as the Gumbel
distribution. The version used here is suitable for modeling minima;
the mirror image of this distribution can be used to model maxima.
See “Extreme Value Distribution” on page B-19 for more details. If
x has a Weibull distribution, then X = log(x) has the type 1 extreme
value distribution.
See Also evpdf, evcdf, evinv, evfit, evlike, evrnd
“Extreme Value Distribution” on page B-19
18-376
expcdf
Purpose Exponential cumulative distribution function
Syntax P = expcdf(X,mu)
[P,PLO,PUP] = expcdf(X,mu,pcov,alpha)
Description P = expcdf(X,mu) computes the exponential cdf at each of the values
in X using the corresponding mean parameter mu. X and mu can be
vectors, matrices, or multidimensional arrays that all have the same
size. A scalar input is expanded to a constant array with the same
dimensions as the other input. The parameters in mu must be positive.
The exponential cdf is
x −t −x
1
p = F ( x |u) = ∫ e dt = 1 − e
0
The result, p, is the probability that a single observation from an
exponential distribution will fall in the interval [0 x].
[P,PLO,PUP] = expcdf(X,mu,pcov,alpha) produces confidence
bounds for P when the input mean parameter mu is an estimate. pcov
is the variance of the estimated mu. alpha specifies 100(1 - alpha)%
confidence bounds. The default value of alpha is 0.05. PLO and PUP are
arrays of the same size as P containing the lower and upper confidence
bounds. The bounds are based on a normal approximation for the
distribution of the log of the estimate of mu. If you estimate mu from a set
of data, you can get a more accurate set of bounds by applying expfit to
the data to get a confidence interval for mu, and then evaluating expinv
at the lower and upper endpoints of that interval.
Examples The following code shows that the median of the exponential
distribution is *log(2).
mu = 10:10:60;
p = expcdf(log(2)*mu,mu)
p =
0.5000 0.5000 0.5000 0.5000 0.5000 0.5000
18-377
expcdf
What is the probability that an exponential random variable is less
than or equal to the mean, µ?
mu = 1:6;
x = mu;
p = expcdf(x,mu)
p =
0.6321 0.6321 0.6321 0.6321 0.6321 0.6321
See Also cdf | exppdf | expinv | expstat | expfit | explike | exprnd
How To • “Exponential Distribution” on page B-16
18-378
expfit
Purpose Exponential parameter estimates
Syntax muhat = expfit(data)
[muhat,muci] = expfit(data)
[muhat,muci] = expfit(data,alpha)
[...] = expfit(data,alpha,censoring)
[...] = expfit(data,alpha,censoring,freq)
Description muhat = expfit(data) estimates the mean of an exponentially
distributed sample data. Each entry of muhat corresponds to the data
in a column of data.
[muhat,muci] = expfit(data) returns 95% confidence intervals for
the mean parameter estimates in matrix muci. The first row of muci
contains the lower bounds of the confidence intervals, and the second
row contains the upper bounds.
[muhat,muci] = expfit(data,alpha) returns 100(1 - alpha)%
confidence intervals for the parameter estimates, where alpha is a value
in the range [0 1] specifying the width of the confidence intervals. By
default, alpha is 0.05, which corresponds to 95% confidence intervals.
[...] = expfit(data,alpha,censoring) accepts a Boolean vector,
censoring, of the same size as data, which is 1 for observations that
are right-censored and 0 for observations that are observed exactly.
data must be a vector in order to pass in the argument censoring.
[...] = expfit(data,alpha,censoring,freq) accepts a frequency
vector, freq of the same size as data. Typically, freq contains integer
frequencies for the corresponding elements in data, but can contain
any nonnegative values. Pass in [] for alpha, censoring, or freq to
use their default values.
Examples The following estimates the mean mu of exponentially distributed data,
and returns a 95% confidence interval for the estimate:
mu = 3;
data = exprnd(mu,100,1); % Simulated data
18-379
expfit
[muhat,muci] = expfit(data)
muhat =
2.7511
muci =
2.2826
3.3813
See Also mle, explike, exppdf, expcdf, expinv, expstat, exprnd
18-380
ExhaustiveSearcher class
Superclasses NeighborSearcher
Purpose Nearest neighbors search using exhaustive search
Description An ExhaustiveSearcher object performs kNN (k-nearest neighbor)
search using exhaustive search. Search objects store information about
the data used, and the distance metric and parameters. The search
performance for this object, compared with the KDTreeSearcher object,
tends to be better for larger dimensions (10 or more) and worse for
smaller dimensions. For more information on search objects, see “What
Are Search Objects?” on page 12-20.
Construction NS = ExhaustiveSearcher(X,'Name',Value) constructs an
ExhaustiveSearcher object based on X, where rows of X correspond
to observations and columns correspond to variables, using one or
more optional name/value pairs. You can then use this object to find
neighbors in X nearest to the query points.
NS = createns(X,'NSMethod','exhaustive','Name',Value) creates
an ExhaustiveSearcher object based on X using createns, where rows
of X correspond to observations and columns correspond to variables,
using one or more optional name/value pairs. You can use this object to
find neighbors in X nearest to the query points.
Name/Value Pairs
Both the ExhaustiveSearcher and the createns functions accept one
or more of the following optional name/value pairs as input:
Distance
A string or function handle specifying the default distance metric
used when you call the knnsearch method.
• 'euclidean' — Euclidean distance (default).
• 'seuclidean' — Standardized Euclidean distance. Each
coordinate difference between rows in X and the query matrix is
scaled by dividing by the corresponding element of the standard
18-381
ExhaustiveSearcher class
deviation computed from X, S=nanstd(X). To specify another
value for S, use the Scale argument.
• 'cityblock' — City block distance.
• 'chebychev' — Chebychev distance (maximum coordinate
difference).
• 'minkowski' — Minkowski distance.
• 'mahalanobis' — Mahalanobis distance, computed using a
positive definite covariance matrix C. The default value of C is
nancov(X). To change the value of C, use the Cov parameter.
• 'cosine' — One minus the cosine of the included angle
between observations (treated as vectors).
• 'correlation' — One minus the sample linear correlation
between observations (treated as sequences of values).
• 'spearman' — One minus the sample Spearman’s rank
correlation between observations (treated as sequences of
values).
• 'hamming' — Hamming distance, percentage of coordinates
that differ.
• 'jaccard' — One minus the Jaccard coefficient, the percentage
of nonzero coordinates that differ.
• custom distance function — A distance function specified using
@ (for example, @distfun). A distance function must be of the
form function D2 = distfun(ZI, ZJ), taking as arguments
a 1-by-n vector ZI containing a single row from X or from the
query points Y, an m2-by-n matrix ZJ containing multiple rows
of X or Y, and returning an m2-by-1 vector of distances D2,
whose jth element is the distance between the observations
ZI and ZJ(j,:).
For more information on these distance metrics, see “Distance
Metrics” on page 12-14.
18-382
ExhaustiveSearcher class
P
A positive scalar indicating the exponent of Minkowski distance.
This parameter is only valid when Distance is 'minkowski'.
Default is 2.
Cov
A positive definite matrix indicating the covariance matrix when
computing the Mahalanobis distance. This parameter is only
valid when Distance is 'mahalanobis'. Default is nancov(X).
Scale
A vector S with the length equal to the number of columns in
X. Each coordinate of X and each query point is scaled by the
corresponding element of S when computing the standardized
Euclidean distance. This parameter is only valid when Distance
is 'seuclidean'. Default is nanstd(X).
Properties X
A matrix used to create the object
Distance
A string specifying a built-in distance metric or a function handle
that you provide when you create the object. This property is the
default distance metric used when you call the knnsearch method
to find nearest neighbors for future query points.
DistParameter
Specifies the additional parameter for the chosen distance metric.
The value is:
• If 'Distance' is 'minkowski': A positive scalar indicating the
exponent of the Minkowski distance.
• If 'Distance' is 'mahalanobis': A positive definite matrix
representing the covariance matrix used for computing the
Mahalanobis distance.
18-383
ExhaustiveSearcher class
• If 'Distance' is 'seuclidean': A vector representing the scale
value of the data when computing the 'seuclidean' distance.
• Otherwise: Empty.
Methods knnsearch Find k-nearest neighbors using
ExhaustiveSearcher object
Examples Create an ExhaustiveSearcher object using the constructor:
load fisheriris
x = meas(:,3:4);
NS = ExhaustiveSearcher(x,'distance','minkowski')
NS =
ExhaustiveSearcher
Properties:
X: [150x2 double]
Distance: 'minkowski'
DistParameter: 2
Create an ExhaustiveSearcher object using createns:
load fisheriris
x = meas(:,3:4);
NS = createns(x,'NsMethod','exhaustive',...
'distance','minkowski')
NS =
ExhaustiveSearcher
18-384
ExhaustiveSearcher class
Properties:
X: [150x2 double]
Distance: 'minkowski'
DistParameter: 2
For more in-depth examples using the knnsearch method, see the
method reference page or see “Example: Classifying Query Data Using
knnsearch” on page 12-22.
References [1] Friedman, J. H., Bentely, J. and Finkel, R. A. (1977). An Algorithm
for Finding Best Matches in Logarithmic Expected Time, ACM
Transactions on Mathematical Software 3, 209.
See Also createns | KDTreeSearcher | NeighborSearcher
How To • “k-Nearest Neighbor Search” on page 12-17
• “Distance Metrics” on page 12-14
18-385
expinv
Purpose Exponential inverse cumulative distribution function
Syntax X = expinv(P,mu)
[X,XLO,XUP] = expinv(X,mu,pcov,alpha)
Description X = expinv(P,mu) computes the inverse of the exponential cdf with
parameters specified by mean parameter mu for the corresponding
probabilities in P. P and mu can be vectors, matrices, or multidimensional
arrays that all have the same size. A scalar input is expanded to a
constant array with the same dimensions as the other input. The
parameters in mu must be positive and the values in P must lie on the
interval [0 1].
[X,XLO,XUP] = expinv(X,mu,pcov,alpha) produces confidence
bounds for X when the input mean parameter mu is an estimate. pcov
is the variance of the estimated mu. alpha specifies 100(1 - alpha)%
confidence bounds. The default value of alpha is 0.05. XLO and XUP are
arrays of the same size as X containing the lower and upper confidence
bounds. The bounds are based on a normal approximation for the
distribution of the log of the estimate of mu. If you estimate mu from a set
of data, you can get a more accurate set of bounds by applying expfit to
the data to get a confidence interval for mu, and then evaluating expinv
at the lower and upper end points of that interval.
The inverse of the exponential cdf is
x = F −1 ( p| ) = − ln(1 − p)
The result, x, is the value such that an observation from an exponential
distribution with parameter µ will fall in the range [0 x] with
probability p.
Examples Let the lifetime of light bulbs be exponentially distributed with µ = 700
hours. What is the median lifetime of a bulb?
expinv(0.50,700)
ans =
18-386
expinv
485.2030
Suppose you buy a box of “700 hour” light bulbs. If 700 hours is the
mean life of the bulbs, half of them will burn out in less than 500 hours.
See Also icdf | expcdf | exppdf | expstat | expfit | explike | exprnd
How To • “Exponential Distribution” on page B-16
18-387
explike
Purpose Exponential negative log-likelihood
Syntax nlogL = explike(param,data)
[nlogL,avar] = explike(param,data)
[...] = explike(param,data,censoring)
[...] = explike(param,data,censoring,freq)
Description nlogL = explike(param,data) returns the negative of the
log-likelihood for the exponential distribution. param is the mean
parameter, mu. nlogL is a scalar.
[nlogL,avar] = explike(param,data) returns the inverse of Fisher’s
information, avar, a scalar. If the input parameter value in param is
the maximum likelihood estimate, avar is its asymptotic variance.
avar is based on the observed Fisher’s information, not the expected
information.
[...] = explike(param,data,censoring) accepts a Boolean vector,
censoring, of the same size as data, which is 1 for observations that
are right-censored and 0 for observations that are observed exactly.
[...] = explike(param,data,censoring,freq) accepts a frequency
vector, freq, of the same size as data. The vector freq typically
contains integer frequencies for the corresponding elements in data,
but can contain any nonnegative values. Pass in [] for censoring to
use its default value.
See Also expcdf | exppdf | expstat | expfit | expinv | exprnd
How To • “Exponential Distribution” on page B-16
18-388
dataset.export
Purpose Write dataset array to file
Syntax export(DS,'file',filename)
export(DS)
export(DS,'file',filename,'Delimiter',delim)
export(DS,'XLSfile',filename)
export(DS,'XPTFile',filename)
export(DS,...,'WriteVarNames',false)
export(DS,...,'WriteObsNames',false)
Description export(DS,'file',filename) writes the dataset array DS to a
tab-delimited text file, including variable names and observation
names, if present. If the observation names exist, the name in the first
column of the first line of the file is the first dimension name for the
dataset (by default, 'Observations'). export overwrites any existing
file named filename.
export(DS) writes to a text file whose default name is the name of
the dataset array DS appended by '.txt'. If export cannot construct
the file name from the dataset array input, it writes to the file
'dataset.txt'. export overwrites any existing file.
export(DS,'file',filename,'Delimiter',delim) writes the dataset
array DS to a text file using the delimiter delim. delim must be one
of the following:
• ' ' or 'space'
• '\t' or 'tab'
• ',' or 'comma'
• ';' or 'semi'
• '|' or 'bar'
export(DS,'XLSfile',filename) writes the dataset array DS to an
Microsoft® Excel® spreadsheet file, including variable names and
observation names (if present). You may also specify the 'Sheet'
18-389
dataset.export
and 'Range' parameter name/value pairs, with parameter values as
accepted by the xlsread function.
export(DS,'XPTFile',filename) writes the dataset array DS to a
SAS XPORT format file. When writing to an XPORT format file variables
must be scalar valued. export saves observation names to a variable
called obsnames unless the WriteObsNames parameter described below
is false. The XPORT format restricts the length of variable names to
eight characters; longer variable names will be truncated.
export(DS,...,'WriteVarNames',false) does not write the variable
names to the text file. export(DS,...,'WriteVarNames',true) is
the default, writing the names as column headings in the first line of
the file.
export(DS,...,'WriteObsNames',false) does
not write the observation names to the text file.
export(DS,...,'WriteObsNames',true) is the default, writing the
names as the first column of the file.
In some cases, export creates a text file that does not represent A
exactly, as described below. If you use dataset to read the file back
into MATLAB, the new dataset array may not have exactly the same
contents as the original dataset array. Save A as a MAT-file if you need
to import it again as a dataset array.
export writes out numeric variables using long g format, and
categorical or character variables as unquoted strings. For
non-character variables with more than one column, export writes out
multiple delimiter-separated fields on each line, and constructs suitable
column headings for the first line of the file. export writes out both the
time and the data fields of timeseries variables, as separate columns.
export writes out variables that have more than two dimensions as a
single empty field in each line of the file. For cell-valued variables,
export writes out the contents of each cell only when the cell contains a
single row, and writes out a single empty field otherwise.
In some cases, export may create a file that cannot be read back into
MATLAB using dataset. Writing a dataset array that contains a
cell-valued variable whose cell contents are not scalars will result in a
18-390
dataset.export
mismatch in the file between the number of fields on each line and
the number of column headings on the first line. Writing a dataset
array that contains a cell-valued variable whose cell contents are not
all the same length will result in a different number of fields on each
line in the file.
Examples Move data between external text files and dataset arrays in the
MATLAB workspace:
A = dataset('file','sat2.dat','delimiter',',')
A =
Test Gender Score
'Verbal' 'Male' 470
'Verbal' 'Female' 530
'Quantitative' 'Male' 520
'Quantitative' 'Female' 480
export(A(A.Score > 500,:),'file','HighScores.txt')
B = dataset('file','HighScores.txt','delimiter','\t')
B =
Test Gender Score
'Verbal' 'Female' 530
'Quantitative' 'Male' 520
See Also dataset
18-391
exppdf
Purpose Exponential probability density function
Syntax Y = exppdf(X,mu)
Description Y = exppdf(X,mu) returns the pdf of the exponential distribution
with mean parameter mu, evaluated at the values in X. X and mu can
be vectors, matrices, or multidimensional arrays that have the same
size. A scalar input is expanded to a constant array with the same
dimensions as the other input. The parameters in mu must be positive.
The exponential pdf is
−x
1
y = f ( x| ) = e
The exponential pdf is the gamma pdf with its first parameter equal to 1.
The exponential distribution is appropriate for modeling waiting
times when the probability of waiting an additional period of time is
independent of how long you have already waited. For example, the
probability that a light bulb will burn out in its next minute of use is
relatively independent of how many minutes it has already burned.
Examples y = exppdf(5,1:5)
y =
0.0067 0.0410 0.0630 0.0716 0.0736
y = exppdf(1:5,1:5)
y =
0.3679 0.1839 0.1226 0.0920 0.0736
See Also pdf | expcdf | expinv | expstat | expfit | explike | exprnd
How To • “Exponential Distribution” on page B-16
18-392
exprnd
Purpose Exponential random numbers
Syntax R = exprnd(mu)
R = exprnd(mu,v)
R = exprnd(mu,m,n)
Description R = exprnd(mu) generates random numbers from the exponential
distribution with mean parameter mu. mu can be a vector, a matrix, or a
multidimensional array. The size of R is the size of mu.
R = exprnd(mu,v) generates an array R of size v containing random
numbers from the exponential distribution with mean mu, where v is a
row vector. If v is a 1-by-2 vector, R is a matrix with v(1) rows and v(2)
columns. If v is 1-by-n, R is an n-dimensional array.
R = exprnd(mu,m,n) generates random numbers from the exponential
distribution with mean parameter mu, where scalars m and n are the
row and column dimensions of R.
Examples n1 = exprnd(5:10)
n1 =
7.5943 18.3400 2.7113 3.0936 0.6078 9.5841
n2 = exprnd(5:10,[1 6])
n2 =
3.2752 1.1110 23.5530 23.4303 5.7190 3.9876
n3 = exprnd(5,2,3)
n3 =
24.3339 13.5271 1.8788
4.7932 4.3675 2.6468
See Also random | expcdf | exppdf | expstat | expfit | explike | expinv
How To • “Exponential Distribution” on page B-16
18-393
expstat
Purpose Exponential mean and variance
Syntax [m,v] = expstat(mu)
Description [m,v] = expstat(mu) returns the mean of and variance for the
exponential distribution with parameters mu. mu can be a vectors,
matrix, or multidimensional array. The mean of the exponential
distribution is µ, and the variance is µ2.
Examples [m,v] = expstat([1 10 100 1000])
m =
1 10 100 1000
v =
1 100 10000 1000000
See Also expinv | expcdf | exppdf | expstat | expfit | explike | exprnd
How To • “Exponential Distribution” on page B-16
18-394
factoran
Purpose Factor analysis
Syntax lambda = factoran(X,m)
[lambda,psi] = factoran(X,m)
[lambda,psi,T] = factoran(X,m)
[lambda,psi,T,stats] = factoran(X,m)
[lambda,psi,T,stats,F] = factoran(X,m)
[...] = factoran(...,param1,val1,param2,val2,...)
Definition factoran computes the maximum likelihood estimate (MLE) of the
factor loadings matrix Λ in the factor analysis model
x = + Λf + e
where x is a vector of observed variables, μ is a constant vector of
means, Λ is a constant d-by-m matrix of factor loadings, f is a vector
of independent, standardized common factors, and e is a vector of
independent specific factors. x, μ, and e are of length d. f is of length m.
Alternatively, the factor analysis model can be specified as
cov( x) = ΛΛ T + Ψ
where Ψ = cov(e) is a d-by-d diagonal matrix of specific variances.
Description lambda = factoran(X,m) returns the maximum likelihood estimate,
lambda, of the factor loadings matrix, in a common factor analysis
model with m common factors. X is an n-by-d matrix where each row
is an observation of d variables. The (i,j)th element of the d-by-m
matrix lambda is the coefficient, or loading, of the jth factor for the ith
variable. By default, factoran calls the function rotatefactors to
rotate the estimated factor loadings using the 'varimax' option.
[lambda,psi] = factoran(X,m) also returns maximum likelihood
estimates of the specific variances as a column vector psi of length d.
[lambda,psi,T] = factoran(X,m) also returns the m-by-m factor
loadings rotation matrix T.
18-395
factoran
[lambda,psi,T,stats] = factoran(X,m) also returns a structure
stats containing information relating to the null hypothesis, H0, that
the number of common factors is m. stats includes the following fields:
Field Description
loglike Maximized log-likelihood value
dfe Error degrees of freedom = ((d-m)^2 - (d+m))/2
chisq Approximate chi-squared statistic for the null hypothesis
p Right-tail significance level for the null hypothesis
factoran does not compute the chisq and p fields unless dfe is positive
and all the specific variance estimates in psi are positive (see “Heywood
Case” on page 18-402 below). If X is a covariance matrix, then you must
also specify the 'nobs' parameter if you want factoran to compute the
chisq and p fields.
[lambda,psi,T,stats,F] = factoran(X,m) also returns, in F,
predictions of the common factors, known as factor scores. F is an n-by-m
matrix where each row is a prediction of m common factors. If X is a
covariance matrix, factoran cannot compute F. factoran rotates F
using the same criterion as for lambda.
[...] = factoran(...,param1,val1,param2,val2,...) enables
you to specify optional parameter name/value pairs to control the model
fit and the outputs. The following are the valid parameter/value pairs.
Parameter Value
'xtype' Type of input in the matrix X. 'xtype' can be one
of:
'data' Raw data (default)
'covariance'Positive definite covariance or
correlation matrix
18-396
factoran
Parameter Value
'scores' Method for predicting factor scores. 'scores' is
ignored if X is not raw data.
'wls' Synonyms for a weighted
'Bartlett' least-squares estimate that
treats F as fixed (default)
'regression'Synonyms for a minimum mean
'Thomson' squared error prediction that is
equivalent to a ridge regression
'start' Starting point for the specific variances psi in
the maximum likelihood optimization. Can be
specified as:
'random' Chooses d uniformly distributed
values on the interval [0,1].
'Rsquared' Chooses the starting vector
as a scale factor times
diag(inv(corrcoef(X))) (default).
For examples, see Jöreskog [2].
Positive Performs the given number of
integer maximum likelihood fits, each
initialized as with 'random'.
factoran returns the fit with the
highest likelihood.
Matrix Performs one maximum likelihood
fit for each column of the specified
matrix. The ith optimization is
initialized with the values from the
ith column. The matrix must have
d rows.
18-397
factoran
Parameter Value
'rotate' Method used to rotate factor loadings and scores.
'rotate' can have the same values as the
'Method' parameter of rotatefactors. See
the reference page for rotatefactors for a full
description of the available methods.
'none' Performs no rotation.
'equamax' Special case of the orthomax
rotation. Use the 'normalize',
'reltol', and 'maxit' parameters
to control the details of the rotation.
'orthomax' Orthogonal rotation that maximizes
a criterion based on the variance of
the loadings.
Use the 'coeff', 'normalize',
'reltol', and 'maxit' parameters
to control the details of the rotation.
'parsimax' Special case of the orthomax rotation
(default). Use the 'normalize',
'reltol', and ’maxit’ parameters
to control the details of the rotation.
'pattern' Performs either an oblique rotation
(the default) or an orthogonal
rotation to best match a specified
pattern matrix. Use the 'type'
parameter to choose the type
of rotation. Use the 'target'
parameter to specify the pattern
matrix.
18-398
factoran
Parameter Value
'procrustes'Performs either an oblique (the
default) or an orthogonal rotation to
best match a specified target matrix
in the least squares sense.
Use the 'type' parameter to choose
the type of rotation. Use 'target'
to specify the target matrix.
'promax' Performs an oblique procrustes
rotation to a target matrix
determined by factoran as a
function of an orthomax solution.
Use the 'power' parameter to
specify the exponent for creating the
target matrix. Because 'promax'
uses 'orthomax' internally, you can
also specify the parameters that
apply to 'orthomax'.
'quartimax' Special case of the orthomax rotation
(default). Use the 'normalize',
'reltol', and ’maxit’ parameters
to control the details of the rotation.
'varimax' Special case of the orthomax rotation
(default). Use the 'normalize',
'reltol', and 'maxit' parameters
to control the details of the rotation.
18-399
factoran
Parameter Value
Function Function handle to rotation function
of the form
[B,T] =
myrotation(A,...)
where A is a d-by-m matrix of
unrotated factor loadings, B is a
d-by-m matrix of rotated loadings,
and T is the corresponding m-by-m
rotation matrix.
Use the factoran parameter
'userargs' to pass additional
arguments to this rotation function.
See “Example 4” on page 18-407.
'coeff' Coefficient, often denoted as γ, defining the
specific 'orthomax' criterion. Must be from 0 to
1. The value 0 corresponds to quartimax, and 1
corresponds to varimax. Default is 1.
'normalize' Flag indicating whether the loading matrix should
be row-normalized (1) or left unnormalized (0) for
'orthomax' or 'varimax' rotation. Default is 1.
'reltol' Relative convergence tolerance for 'orthomax' or
'varimax' rotation. Default is sqrt(eps).
'maxit' Iteration limit for 'orthomax' or 'varimax'
rotation. Default is 250.
'target' Target factor loading matrix for 'procrustes'
rotation. Required for 'procrustes' rotation. No
default value.
'type' Type of 'procrustes' rotation. Can be 'oblique'
(default) or 'orthogonal'.
18-400
factoran
Parameter Value
'power' Exponent for creating the target matrix in the
'promax' rotation. Must be ≥ 1. Default is 4.
'userargs' Denotes the beginning of additional input values
for a user-defined rotation function. factoran
appends all subsequent values, in order and
without processing, to the rotation function
argument list, following the unrotated factor
loadings matrix A. See “Example 4” on page 18-407.
'nobs' If X is a covariance or correlation matrix, indicates
the number of observations that were used in its
estimation. This allows calculation of significance
for the null hypothesis even when the original data
are not available. There is no default. 'nobs' is
ignored if X is raw data.
'delta' Lower bound for the specific variances psi during
the maximum likelihood optimization. Default is
0.005.
'optimopts' Structure that specifies control parameters for
the iterative algorithm the function uses to
compute maximum likelihood estimates. Create
this structure with the function statset. Enter
statset('factoran') to see the names and
default values of the parameters that factoran
accepts in the options structure. See the reference
page for statset for more information about these
options.
Remarks Observed Data Variables
The variables in the observed data matrix X must be linearly
independent, i.e., cov(X) must have full rank, for maximum likelihood
estimation to succeed. factoran reduces both raw data and a covariance
matrix to a correlation matrix before performing the fit.
18-401
factoran
factoran standardizes the observed data X to zero mean and unit
variance before estimating the loadings lambda. This does not affect the
model fit, because MLEs in this model are invariant to scale. However,
lambda and psi are returned in terms of the standardized variables,
i.e., lambda*lambda'+diag(psi) is an estimate of the correlation
matrix of the original data X (although not after an oblique rotation).
See “Example 1” on page 18-402 and “Example 3” on page 18-404.
Heywood Case
If elements of psi are equal to the value of the 'delta' parameter
(i.e., they are essentially zero), the fit is known as a Heywood case, and
interpretation of the resulting estimates is problematic. In particular,
there can be multiple local maxima of the likelihood, each with different
estimates of the loadings and the specific variances. Heywood cases
can indicate overfitting (i.e., m is too large), but can also be the result
of underfitting.
Rotation of Factor Loadings and Scores
Unless you explicitly specify no rotation using the 'rotate' parameter,
factoran rotates the estimated factor loadings, lambda, and the factor
scores, F. The output matrix T is used to rotate the loadings, i.e.,
lambda = lambda0*T, where lambda0 is the initial (unrotated) MLE of
the loadings. T is an orthogonal matrix for orthogonal rotations, and
the identity matrix for no rotation. The inverse of T is known as the
primary axis rotation matrix, while T itself is related to the reference
axis rotation matrix. For orthogonal rotations, the two are identical.
factoran computes factor scores that have been rotated by inv(T'),
i.e., F = F0 * inv(T'), where F0 contains the unrotated predictions.
The estimated covariance of F is inv(T'*T), which, for orthogonal or no
rotation, is the identity matrix. Rotation of factor loadings and scores
is an attempt to create a more easily interpretable structure in the
loadings matrix after maximum likelihood estimation.
Examples Example 1
Load the carbig data, and fit the default model with two factors.
18-402
factoran
load carbig
X = [Acceleration Displacement Horsepower MPG Weight];
X = X(all(~isnan(X),2),:);
[Lambda,Psi,T,stats,F] = factoran(X,2,...
'scores','regression');
inv(T'*T) % Estimated correlation matrix of F, == eye(2)
Lambda*Lambda'+diag(Psi) % Estimated correlation matrix
Lambda*inv(T) % Unrotate the loadings
F*T' % Unrotate the factor scores
biplot(Lambda,... % Create biplot of two factors
'LineWidth',2,...
'MarkerSize',20)
18-403
factoran
Example 2
Although the estimates are the same, the use of a covariance matrix
rather than raw data doesn’t let you request scores or significance level:
[Lambda,Psi,T] = factoran(cov(X),2,'xtype','cov')
[Lambda,Psi,T] = factoran(corrcoef(X),2,'xtype','cov')
Example 3
Use promax rotation:
[Lambda,Psi,T,stats,F] = factoran(X,2,'rotate','promax',...
'powerpm',4);
18-404
factoran
inv(T'*T) % Est'd corr of F,
% no longer eye(2)
Lambda*inv(T'*T)*Lambda'+diag(Psi) % Est'd corr of X
Plot the unrotated variables with oblique axes superimposed.
invT = inv(T)
Lambda0 = Lambda*invT
line([-invT(1,1) invT(1,1) NaN -invT(2,1) invT(2,1)], ...
[-invT(1,2) invT(1,2) NaN -invT(2,2) invT(2,2)], ...
'Color','r','linewidth',2)
hold on
biplot(Lambda0,...
'LineWidth',2,...
'MarkerSize',20)
xlabel('Loadings for unrotated Factor 1')
ylabel('Loadings for unrotated Factor 2')
18-405
factoran
Plot the rotated variables against the oblique axes:
biplot(Lambda,'LineWidth',2,'MarkerSize',20)
18-406
factoran
Example 4
Syntax for passing additional arguments to a user-defined rotation
function:
[Lambda,Psi,T] = ...
factoran(X,2,'rotate',@myrotation,'userargs',1,'two');
References [1] Harman, H. H. Modern Factor Analysis. 3rd Ed. Chicago:
University of Chicago Press, 1976.
[2] Jöreskog, K. G. “Some Contributions to Maximum Likelihood Factor
Analysis.” Psychometrika. Vol. 32, Issue 4, 1967, pp. 443–482.
18-407
factoran
[3] Lawley, D. N., and A. E. Maxwell. Factor Analysis as a Statistical
Method. 2nd Ed. New York: American Elsevier Publishing Co., 1971.
See Also biplot, princomp, procrustes, pcacov, rotatefactors, statset
18-408
TreeBagger.FBoot property
Purpose Fraction of in-bag observations
Description The FBoot property is the fraction of observations to be randomly
selected with replacement for each bootstrap replica. The size of each
replica is given by n*FBoot, where n is the number of observations in
the training set. The default value is 1.
18-409
fcdf
Purpose F cumulative distribution function
Syntax P = fcdf(X,V1,V2)
Description P = fcdf(X,V1,V2) computes the F cdf at each of the values in X using
the corresponding numerator degrees of freedom V1 and denominator
degrees of freedom V2. X, V1, and V2 can be vectors, matrices, or
multidimensional arrays that are all the same size. A scalar input is
expanded to a constant matrix with the same dimensions as the other
inputs. V1 and V2 parameters must contain real positive values.
The F cdf is
⎡ ( + 2 ) ⎤ 1 1 −2
Γ⎢ 1 ⎥ ⎛ ⎞ 2 2
p = F ( x | 1 , 2 ) = ∫
x ⎣ 2 ⎦ 1 t
0 ⎛ 1 ⎞ ⎛ 2 ⎞
⎜ ⎟ 1 + 2
dt
Γ⎜ ⎟Γ⎜ ⎟ ⎝ 2 ⎠ ⎡ ⎛ ⎞ ⎤ 2
⎝ 2⎠ ⎝ 2 ⎠ 1
⎢1 + ⎜ ⎟ t ⎥
⎣ ⎝ 2 ⎠ ⎦
The result, p, is the probability that a single observation from an F
distribution with parameters ν1 and ν2 will fall in the interval [0 x].
Examples The following illustrates a useful mathematical identity for the F
distribution:
nu1 = 1:5;
nu2 = 6:10;
x = 2:6;
F1 = fcdf(x,nu1,nu2)
F1 =
0.7930 0.8854 0.9481 0.9788 0.9919
F2 = 1 - fcdf(1./x,nu2,nu1)
F2 =
0.7930 0.8854 0.9481 0.9788 0.9919
18-410
fcdf
See Also cdf, fpdf, finv, fstat, frnd
“F Distribution” on page B-25
18-411
ff2n
Purpose Two-level full factorial design
Syntax dFF2 = ff2n(n)
Description dFF2 = ff2n(n) gives factor settings dFF2 for a two-level full factorial
design with n factors. dFF2 is m-by-n, where m is the number of
treatments in the full-factorial design. Each row of dFF2 corresponds
to a single treatment. Each column contains the settings for a single
factor, with values of 0 and 1 for the two levels.
Examples dFF2 = ff2n(3)
dFF2 =
0 0 0
0 0 1
0 1 0
0 1 1
1 0 0
1 0 1
1 1 0
1 1 1
See Also fullfact
18-412
TreeBagger.fillProximities
Purpose Proximity matrix for training data
Syntax B = fillProximities(B)
B = fillProximities(B,'param1',val1,'param2',val2,...
Description B = fillProximities(B) computes a proximity matrix for the training
data and stores it in the Properties field of B.
B = fillProximities(B,'param1',val1,'param2',val2,...)
specifies optional parameter name/value pairs:
'trees' Either 'all' or a vector of indices of the trees in
the ensemble to be used in computing the proximity
matrix. Default is 'all'.
'nprint' Number of training cycles (grown trees) after which
TreeBagger displays a diagnostic message showing
training progress. Default is no diagnostic messages.
See Also CompactTreeBagger.outlierMeasure, CompactTreeBagger.proximity
18-413
qrandstream.findobj
Purpose Find objects matching specified conditions
Syntax hm = findobj(h, 'conditions')
Description The findobj method of the handle class follows the same syntax as
the MATLAB findobj command, except that the first argument must
be an array of handles to objects.
hm = findobj(h, 'conditions') searches the handle object array
h and returns an array of handle objects matching the specified
conditions. Only the public members of the objects of h are considered
when evaluating the conditions.
See Also findobj, qrandstream
18-414
qrandstream.findprop
Purpose Find property of MATLAB handle object
Syntax p = findprop(h,'propname')
Description p = findprop(h,'propname') finds and returns the meta.property
object associated with property name propname of scalar handle object
h. propname must be a string. It can be the name of a property defined
by the class of h or a dynamic property added to scalar object h.
If no property named propname exists for object h, an empty
meta.property array is returned.
See Also dynamicprops, findobj, meta.property, qrandstream
18-415
finv
Purpose F inverse cumulative distribution function
Syntax X = finv(P,V1,V2)
Description X = finv(P,V1,V2) computes the inverse of the F cdf with numerator
degrees of freedom V1 and denominator degrees of freedom V2 for the
corresponding probabilities in P. P, V1, and V2 can be vectors, matrices,
or multidimensional arrays that all have the same size. A scalar input
is expanded to a constant array with the same dimensions as the other
inputs.
V1 and V2 parameters must contain real positive values, and the values
in P must lie on the interval [0 1].
The F inverse function is defined in terms of the F cdf as
x = F −1 ( p| 1 , 2 ) = { x : F ( x | 1 , 2 ) = p}
where
⎡ ( + 2 ) ⎤ 1 1 −2
x Γ⎢ 1 ⎥ ⎛ ⎞ 2 2
p = F ( x | 1 , 2 ) = ∫ ⎣ 2 ⎦ 1 t
⎜ ⎟ 1 + 2
dt
⎛ ⎞ ⎛ ⎞
0 Γ⎜ 1 ⎟Γ⎜ 2 ⎟ ⎝ 2 ⎠ ⎡ ⎛ 1 ⎞ ⎤ 2
⎝ 2⎠ ⎝ 2 ⎠ ⎢1 + ⎜ ⎟ t ⎥
⎣ ⎝ 2 ⎠ ⎦
Examples Find a value that should exceed 95% of the samples from an F
distribution with 5 degrees of freedom in the numerator and 10 degrees
of freedom in the denominator.
x = finv(0.95,5,10)
x =
3.3258
You would observe values greater than 3.3258 only 5% of the time by
chance.
18-416
finv
See Also icdf, fcdf, fpdf, fstat, frnd
“F Distribution” on page B-25
18-417
gmdistribution.fit
Purpose Gaussian mixture parameter estimates
Syntax obj = gmdistribution.fit(X,k)
obj = gmdistribution.fit(...,param1,val1,param2,val2,...)
Description obj = gmdistribution.fit(X,k) uses an Expectation Maximization
(EM) algorithm to construct an object obj of the gmdistribution
class containing maximum likelihood estimates of the parameters in
a Gaussian mixture model with k components for data in the n-by-d
matrix X, where n is the number of observations and d is the dimension
of the data.
gmdistribution treats NaN values as missing data. Rows of X with NaN
values are excluded from the fit.
obj = gmdistribution.fit(...,param1,val1,param2,val2,...)
provides control over the iterative EM algorithm. Parameters and
values are listed below.
Parameter Value
'Start' Method used to choose initial component parameters.
One of the following:
• 'randSample' — To select k observations from X at
random as initial component means. The mixing
proportions are uniform. The initial covariance
matrices for all components are diagonal, where
the element j on the diagonal is the variance of
X(:,j). This is the default.
• S — A structure array with fields mu, Sigma,
and PComponents. See gmdistribution for
descriptions of values.
• s — A vector of length n containing an initial guess
of the component index for each point.
18-418
gmdistribution.fit
Parameter Value
'Replicates' A positive integer giving the number of times to
repeat the EM algorithm, each time with a new set of
parameters. The solution with the largest likelihood
is returned. A value larger than 1 requires the
'randSample' start method. The default is 1.
'CovType' 'diagonal' if the covariance matrices are restricted
to be diagonal; 'full' otherwise. The default is
'full'.
'SharedCov' Logical true if all the covariance matrices are
restricted to be the same (pooled estimate); logical
false otherwise.
'Regularize' A nonnegative regularization number added to
the diagonal of covariance matrices to make them
positive-definite. The default is 0.
'Options' Options structure for the iterative EM algorithm, as
created by statset. gmdistribution.fit uses the
parameters 'Display' with a default value of 'off',
'MaxIter' with a default value of 100, and 'TolFun'
with a default value of 1e6.
In some cases, gmdistribution may converge to a solution where one
or more of the components has an ill-conditioned or singular covariance
matrix.
The following issues may result in an ill-conditioned covariance matrix:
• The number of dimension of your data is relatively high and there
are not enough observations.
• Some of the features (variables) of your data are highly correlated.
• Some or all the features are discrete.
• You tried to fit the data to too many components.
18-419
gmdistribution.fit
In general, you can avoid getting ill-conditioned covariance matrices by
using one of the following precautions:
• Pre-process your data to remove correlated features.
• Set 'SharedCov' to true to use an equal covariance matrix for every
component.
• Set 'CovType' to 'diagonal'.
• Use 'Regularize' to add a very small positive number to the
diagonal of every covariance matrix.
• Try another set of initial values.
In other cases gmdistribution may pass through an intermediate step
where one or more of the components has an ill-conditioned covariance
matrix. Trying another set of initial values may avoid this issue without
altering your data or model.
Examples Generate data from a mixture of two bivariate Gaussian distributions
using the mvnrnd function:
MU1 = [1 2];
SIGMA1 = [2 0; 0 .5];
MU2 = [-3 -5];
SIGMA2 = [1 0; 0 1];
X = [mvnrnd(MU1,SIGMA1,1000);mvnrnd(MU2,SIGMA2,1000)];
scatter(X(:,1),X(:,2),10,'.')
hold on
18-420
gmdistribution.fit
Next, fit a two-component Gaussian mixture model:
options = statset('Display','final');
obj = gmdistribution.fit(X,2,'Options',options);
10 iterations, log-likelihood = -7046.78
h = ezcontour(@(x,y)pdf(obj,[x y]),[-8 6],[-8 6]);
18-421
gmdistribution.fit
Among the properties of the fit are the parameter estimates:
ComponentMeans = obj.mu
ComponentMeans =
0.9391 2.0322
-2.9823 -4.9737
ComponentCovariances = obj.Sigma
ComponentCovariances(:,:,1) =
1.7786 -0.0528
-0.0528 0.5312
18-422
gmdistribution.fit
ComponentCovariances(:,:,2) =
1.0491 -0.0150
-0.0150 0.9816
MixtureProportions = obj.PComponents
MixtureProportions =
0.5000 0.5000
The Akaike information is minimized by the two-component model:
AIC = zeros(1,4);
obj = cell(1,4);
for k = 1:4
obj{k} = gmdistribution.fit(X,k);
AIC(k)= obj{k}.AIC;
end
[minAIC,numComponents] = min(AIC);
numComponents
numComponents =
2
model = obj{2}
model =
Gaussian mixture distribution
with 2 components in 2 dimensions
Component 1:
Mixing proportion: 0.500000
Mean: 0.9391 2.0322
Component 2:
Mixing proportion: 0.500000
Mean: -2.9823 -4.9737
Both the Akaike and Bayes information are negative log-likelihoods for
the data with penalty terms for the number of estimated parameters.
They are often used to determine an appropriate number of components
for a model when the number of components is unspecified.
18-423
gmdistribution.fit
References [1] McLachlan, G., and D. Peel. Finite Mixture Models. Hoboken, NJ:
John Wiley & Sons, Inc., 2000.
See Also gmdistribution, cluster
18-424
NaiveBayes.fit
Purpose Create Naive Bayes classifier object by fitting training data
Syntax nb = NaiveBayes.fit(training, class)
nb = NaiveBayes.fit(..., 'param1',val1, 'param2',val2, ...)
Description nb = NaiveBayes.fit(training, class) builds a NaiveBayes
classifier object nb. training is an N-by-D numeric matrix of training
data. Rows of training correspond to observations; columns correspond
to features. class is a classing variable for training (see “Grouped
Data” on page 2-34) taking K distinct levels. Each element of class
defines which class the corresponding row of training belongs to.
training and class must have the same number of rows.
nb = NaiveBayes.fit(..., 'param1',val1, 'param2',val2, ...)
specifies one or more of the following name/value pairs:
• 'Distribution' – a string or a 1-by-D cell vector of strings,
specifying which distributions fit uses to model the data. If the
value is a string, fit models all the features using one type of
distribution. fit can also model different features using different
types of distributions. If the value is a cell vector, its jth element
specifies the distribution fit uses for the jth feature. The available
types of distributions are:
'normal' Normal (Gaussian) distribution.
(default)
'kernel' Kernel smoothing density estimate.
18-425
NaiveBayes.fit
'mvmn' Multivariate multinomial distribution for
discrete data. fit assumes each individual
feature follows a multinomial model within a
class. The parameters for a feature include
the probabilities of all possible values that the
corresponding feature can take.
'mn' Multinomial distribution for classifying the
count-based data such as the bag-of-tokens
model. In the bag-of-tokens model, the value of
the jth feature is the number of occurrences of
the jth token in this observation, so it must be
a non-negative integer. When 'mn' is used, fit
considers each observation as multiple trials of
a multinomial distribution, and considers each
occurrence of a token as one trial. The number
of categories (bins) in this multinomial model is
the number of distinct tokens, i.e., the number of
columns of training.
• 'Prior' – The prior probabilities for the classes, specified as one of
the following:
'empirical' fit estimates the prior probabilities from the
(default) relative frequencies of the classes in training.
'uniform' The prior probabilities are equal for all classes.
18-426
NaiveBayes.fit
vector A numeric vector of length K specifying the
prior probabilities in the class order of class.
structure A structure S containing class levels and their
prior probabilities. S must have two fields:
• S.prob: A numeric vector of prior
probabilities.
- S.class: A vector of the same type as class,
containing unique class levels indicating
the class for the corresponding element of
prob. S.class must contain all the K levels
in class. It can also contain classes that
do not appear in class. This can be useful
if training is a subset of a larger training
set. fit ignores any classes that appear in
S.class but not in class.
If the prior probabilities don’t sum to one, fit will normalize them.
• 'KSWidth' – The bandwidth of the kernel smoothing window. The
default is to select a default bandwidth automatically for each
combination of feature and class, using a value that is optimal for
a Gaussian distribution. You can specify the value as one of the
following:
scalar Width for all features in all classes.
row 1-by-D vector where the jth element is the bandwidth for
vector the jth feature in all classes.
column K-by-1 vector where the ith element specifies the
vector bandwidth for all features in the ith class. K represents
the number of class levels.
18-427
NaiveBayes.fit
matrix K-by-D matrix M where M(i,j) specifies the bandwidth
for the jth feature in the ith class.
structure A structure S containing class levels and their
bandwidths. S must have two fields:
• S.width – A numeric array of bandwidths specified as
a row vector, or a matrix with D columns.
- S.class – A vector of the same type as class,
containing unique class levels indicating the class for
the corresponding row of width.
• 'KSSupport' – The regions where the density can be applied. It can
be a string, a two-element vector as shown below, or a 1-by-D cell
array of these values:
'unbounded' The density can extend over the whole real
(default) line.
'positive' The density is restricted to positive values.
[L,U] A two-element vector specifying the finite
lower bound L and upper bound U for the
support of the density.
• 'KSType' – The type of kernel smoother to use. It can be a string or
a 1-by-D cell array of strings. Each string can be 'normal' (default),
'box', 'triangle', or 'epanechnikov'.
See Also “Naive Bayes Classification” on page 12-6, “Grouped Data” on page 2-34
18-428
fitdist
Purpose Fit probability distribution to data
Syntax PD = fitdist(X, DistName)
[PDCA, GN, GL] = fitdist(X, DistName, 'By', GroupVar)
... = fitdist(..., param1, val1, param2, val2, ...)
Description PD = fitdist(X, DistName) fits the probability distribution specified
by DistName to the data in the column vector X, and returns PD, an
object representing the fitted distribution.
[PDCA, GN, GL] = fitdist(X, DistName, 'By', GroupVar) takes a
grouping variable, GroupVar, fits the specified distribution to the data in
X from each group, and returns PDCA, a cell array of the fitted probability
distribution objects. GroupVar can also be a cell array of multiple
grouping variables. GN is a cell array of group labels. GL is a cell array
of grouping variable levels, with one column for each grouping variable.
... = fitdist(..., param1, val1, param2, val2, ...)
specifies optional parameter name/value pairs, as described in the
Parameter/Values table. Parameter and value names are case
insensitive.
Input X A column vector of data.
Arguments
Note Any NaN values in X are ignored by the
fitting calculations. Additionally, any NaN
values in the censoring vector or frequency
vector will cause the corresponding values in
X to be ignored by the fitting calculations.
DistName A string specifying a distribution. Choices are:
• 'kernel' — To fit a nonparametric
kernel-smoothing distribution.
18-429
fitdist
• Any of the following to fit a parametric
distribution:
- 'beta'
- 'binomial'
- 'birnbaumsaunders'
- 'exponential'
- 'extreme value' or 'ev'
- 'gamma'
- 'generalized extreme value' or
'gev'
- 'generalized pareto' or 'gp'
- 'inversegaussian'
- 'logistic'
- 'loglogistic'
- 'lognormal'
- 'nakagami'
- 'negative binomial' or 'nbin'
- 'normal'
- 'poisson'
- 'rayleigh'
- 'rician'
- 'tlocationscale'
- 'weibull' or 'wbl'
18-430
fitdist
For more information on these parametric
distributions, see Appendix B, “Distribution
Reference”.
GroupVar A grouping variable or a cell array of multiple
grouping variables. For more information on
grouping variables, see “Grouped Data” on
page 2-34.
Parameter Values
'censoring'A Boolean vector the same size as X, containing 1s when
the corresponding elements in X are right-censored
observations and 0s when the corresponding elements
are exact observations. Default is a vector of 0s.
Note Any NaN values in this censoring vector are
ignored by the fitting calculations. Additionally, any
NaN values in X or the frequency vector will cause
the corresponding values in the censoring vector to be
ignored by the fitting calculations.
'frequency'A vector the same size as X, containing nonnegative
integers specifying the frequencies for the corresponding
elements in X. Default is a vector of 1s.
Note Any NaN values in this frequency vector are
ignored by the fitting calculations. Additionally, any
NaN values in X or the censoring vector will cause
the corresponding values in the frequency vector to be
ignored by the fitting calculations.
18-431
fitdist
Parameter Values
'options' A structure created by the statset function to specify
control parameters for the iterative fitting algorithm.
'n' For 'binomial' distributions only, a positive integer
specifying the N parameter (number of trials).
'theta' For 'generalized pareto' distributions only, value
specifying the theta (threshold) parameter for the
generalized Pareto distribution. Default is 0.
'kernel' For 'kernel' distributions only, a string specifying the
type of kernel smoother to use. Choices are:
• 'normal' (default)
• 'box'
• 'triangle'
• 'epanechnikov'
'support' For 'kernel' distributions only, any of the following to
specify the support:
• 'unbounded' — Default. If the density can extend
over the whole real line.
• 'positive' — To restrict it to positive values.
• A two-element vector giving finite lower and upper
limits for the support of the density.
'width' For 'kernel' distributions only, a value specifying the
bandwidth of the kernel smoothing window. The default
is optimal for estimating normal densities, but you may
want to choose a smaller value to reveal features such as
multiple modes.
18-432
fitdist
Output PD An object in either the ProbDistUnivKernel
Arguments class or the ProbDistUnivParam class, which
are derived from the ProbDist class.
PDCA A cell array of the fitted probability
distribution objects.
GN A cell array of group labels.
GL A cell array of grouping variable levels, with
one column for each grouping variable.
Examples Creating a ProbDistUnivKernel Object
1 Load a MAT-file, included with the Statistics Toolbox software, which
contains MPG, a column vector of data.
load carsmall
2 Create a ProbDistUnivKernel object by fitting a nonparametric
kernel-smoothing distribution to the data:
ksd = fitdist(MPG,'kernel')
ksd =
kernel distribution
Kernel = normal
Bandwidth = 4.11428
Support = unbounded
Creating a ProbDistUnivParam Object
1 Load a MAT-file, included with the Statistics Toolbox software, which
contains MPG, a column vector of data, and Origin, a cell array of
seven grouping variables.
18-433
fitdist
load carsmall
2 Create a cell array of ProbDistUnivParam objects by fitting a
parametric distribution, namely a Weibull distribution, to the data,
and also grouping the data. Since there is only one car from Italy,
fitdist will return an error, since you cannot fit a distribution to a
single observation.
wd = fitdist(MPG,'weibull','by',Origin)
Algorithm The fitdist function fits most distributions using maximum likelihood.
Two exceptions are the normal and lognormal distributions with
uncensored data. For the uncensored normal distribution, the estimated
value of the sigma parameter is the square root of the unbiased
estimate of the variance. For the uncensored lognormal distribution,
the estimated value of the sigma parameter is the square root of the
unbiased estimate of the variance of the log of the data.
References [1] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous
Univariate Distributions. Vol. 1, Hoboken, NJ: Wiley-Interscience,
1993.
[2] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous
Univariate Distributions. Vol. 2, Hoboken, NJ: Wiley-Interscience,
1994.
[3] Bowman, A. W., and A. Azzalini. Applied Smoothing Techniques for
Data Analysis. New York: Oxford University Press, 1997.
Alternatives dfittool — Opens a graphical user interface for displaying fit
distributions to data, or for fitting distributions to your data and
displaying them over plots of the empirical distributions, by importing
data from the workspace.
See Also disttool
randtool
18-434
fitdist
statset — Function that creates a structure that specifies control
parameters for the iterative fitting algorithm
ProbDist class
ProbDistUnivKernel class
ProbDistUnivParam class
Appendix B, “Distribution Reference” — For more information on
parametric distributions
“Grouped Data” on page 2-34 — For more information on grouping
variables
18-435
categorical.flipdim
Purpose Flip categorical array along specified dimension
Syntax B = flipdim(A,dim)
Description B = flipdim(A,dim) returns the categorical array A with dimension
dim flipped.
See Also fliplr, flipud, permute, rot90
18-436
categorical.fliplr
Purpose Flip categorical matrix in left/right direction
Syntax B = fliplr(A)
Description B = fliplr(A) returns the 2-D categorical matrix A with rows
preserved and columns flipped in the left/right direction.
See Also flipdim, flipud, permute, rot90
18-437
categorical.flipud
Purpose Flip categorical matrix in up/down direction
Syntax B = flipud(A)
Description B = flipud(A) returns the 2-D categorical matrix A with rows
preserved and columns flipped in the up/down direction.
See Also flipdim, fliplr, permute, rot90
18-438
fpdf
Purpose F probability density function
Syntax Y = fpdf(X,V1,V2)
Description Y = fpdf(X,V1,V2) computes the F pdf at each of the values in X using
the corresponding numerator degrees of freedom V1 and denominator
degrees of freedom V2. X, V1, and V2 can be vectors, matrices, or
multidimensional arrays that all have the same size. A scalar input is
expanded to a constant array with the same dimensions as the other
inputs. V1 and V2 parameters must contain real positive values, and
the values in X must lie on the interval [0 ∞).
The probability density function for the F distribution is
⎡ ( + 2 ) ⎤ 1 1 −2
Γ⎢ 1 ⎥ ⎛ ⎞ 2 2
y = f ( x | 1 , 2 ) = ⎣ ⎦ 1
2 x
⎜ ⎟ 1 + 2
⎛ ⎞ ⎛ ⎞
Γ⎜ 1 ⎟Γ⎜ 2 ⎟ ⎝ 2 ⎠ ⎡ ⎛ ⎞ ⎤ 2
⎝ 2⎠ ⎝ 2 ⎠ 1
⎢1 + ⎜ ⎟ x ⎥
⎣ ⎝ 2⎠ ⎦
Examples y = fpdf(1:6,2,2)
y =
0.2500 0.1111 0.0625 0.0400 0.0278 0.0204
z = fpdf(3,5:10,5:10)
z =
0.0689 0.0659 0.0620 0.0577 0.0532 0.0487
See Also pdf, fcdf, finv, fstat, frnd
“F Distribution” on page B-25
18-439
fracfact
Purpose Fractional factorial design
Syntax dfF = fracfact(generators)
[dfF,confounding] = fracfact(generators)
Description dfF = fracfact(generators) gives factor settings dfF for a two-level
Box-Hunter-Hunter fractional factorial design specified by the
generators in generators. generators is a string consisting of
words formed from the letters a-z, separated by spaces. For example,
generators = 'a b c ab ac'. Alternatively, generators is a cell
array of strings with one word per cell, as returned by fracfactgen.
Single-character words indicate basic factors, for which the design
includes all full-factorial treatments. Multiple-character words indicate
factors whose levels are limited by the design to products of the levels
of component basic factors. dfF is m-by-n, where m is the number
of treatments in the design and n is the number factors specified by
generators.
[dfF,confounding] = fracfact(generators) also returns a cell
array confounding that shows the confounding pattern among the
main effects and the two-factor interactions.
Examples Suppose you wish to determine the effects of four two-level factors, for
which there may be two-way interactions. A full-factorial design would
require 24 = 16 runs. The fracfactgen function finds generators for a
resolution IV (separating main effects) fractional-factorial design that
requires only 23 = 8 runs:
generators = fracfactgen('a b c d',3,4)
generators =
'a'
'b'
'c'
'abc'
The more economical design and the corresponding confounding pattern
are returned by fracfact:
18-440
fracfact
[dfF,confounding] = fracfact(generators)
dfF =
-1 -1 -1 -1
-1 -1 1 1
-1 1 -1 1
-1 1 1 -1
1 -1 -1 1
1 -1 1 -1
1 1 -1 -1
1 1 1 1
confounding =
'Term' 'Generator' 'Confounding'
'X1' 'a' 'X1'
'X2' 'b' 'X2'
'X3' 'c' 'X3'
'X4' 'abc' 'X4'
'X1*X2' 'ab' 'X1*X2 + X3*X4'
'X1*X3' 'ac' 'X1*X3 + X2*X4'
'X1*X4' 'bc' 'X1*X4 + X2*X3'
'X2*X3' 'bc' 'X1*X4 + X2*X3'
'X2*X4' 'ac' 'X1*X3 + X2*X4'
'X3*X4' 'ab' 'X1*X2 + X3*X4'
The confounding pattern shows, for example, that the two-way
interaction between X1 and X2 is confounded by the two-way interaction
between X3 and X4.
References [1] Box, G. E. P., W. G. Hunter, and J. S. Hunter. Statistics for
Experimenters. Hoboken, NJ: Wiley-Interscience, 1978.
See Also fracfactgen, hadamard
18-441
fracfactgen
Purpose Fractional factorial design generators
Syntax generators = fracfactgen(terms)
generators = fracfactgen(terms,k)
generators = fracfactgen(terms,k,R)
generators = fracfactgen(terms,k,R,basic)
Description generators = fracfactgen(terms) uses the Franklin-Bailey
algorithm to find generators for the smallest two-level
fractional-factorial design for estimating linear model terms specified
by terms. terms is a string consisting of words formed from the
letters a-z, separated by spaces. For example, terms = 'a b c ab
ac'. Single-character words indicate main effects to be estimated;
multiple-character words indicate interactions. Alternatively, terms
is an m-by-n matrix of 0s and 1s where m is the number of model
terms to be estimated and n is the number of factors. For example,
if terms contains rows [0 1 0 0] and [1 0 0 1], then the factor b
and the interaction between factors a and d are included in the model.
generators is a cell array of strings with one generator per cell. Pass
generators to fracfact to produce the fractional-factorial design and
corresponding confounding pattern.
generators = fracfactgen(terms,k) returns generators for a
two-level fractional-factorial design with 2k-runs, if possible. If k is [],
fracfactgen finds the smallest design.
generators = fracfactgen(terms,k,R) finds a design with resolution
R, if possible. The default resolution is 3.
A design of resolution R is one in which no n-factor interaction is
confounded with any other effect containing less than R – n factors.
Thus a resolution III design does not confound main effects with one
another but may confound them with two-way interactions, while a
resolution IV design does not confound either main effects or two-way
interactions but may confound two-way interactions with each other.
If fracfactgen is unable to find a design at the requested resolution, it
tries to find a lower-resolution design sufficient to calibrate the model.
18-442
fracfactgen
If it is successful, it returns the generators for the lower-resolution
design along with a warning. If it fails, it returns an error.
generators = fracfactgen(terms,k,R,basic) also accepts a vector
basic specifying the indices of factors that are to be treated as basic.
These factors receive full-factorial treatments in the design. The default
includes factors that are part of the highest-order interaction in terms.
Examples Suppose you wish to determine the effects of four two-level factors, for
which there may be two-way interactions. A full-factorial design would
require 24 = 16 runs. The fracfactgen function finds generators for a
resolution IV (separating main effects) fractional-factorial design that
requires only 23 = 8 runs:
generators = fracfactgen('a b c d',3,4)
generators =
'a'
'b'
'c'
'abc'
The more economical design and the corresponding confounding pattern
are returned by fracfact:
[dfF,confounding] = fracfact(generators)
dfF =
-1 -1 -1 -1
-1 -1 1 1
-1 1 -1 1
-1 1 1 -1
1 -1 -1 1
1 -1 1 -1
1 1 -1 -1
1 1 1 1
confounding =
'Term' 'Generator' 'Confounding'
'X1' 'a' 'X1'
'X2' 'b' 'X2'
18-443
fracfactgen
'X3' 'c' 'X3'
'X4' 'abc' 'X4'
'X1*X2' 'ab' 'X1*X2 + X3*X4'
'X1*X3' 'ac' 'X1*X3 + X2*X4'
'X1*X4' 'bc' 'X1*X4 + X2*X3'
'X2*X3' 'bc' 'X1*X4 + X2*X3'
'X2*X4' 'ac' 'X1*X3 + X2*X4'
'X3*X4' 'ab' 'X1*X2 + X3*X4'
The confounding pattern shows, for example, that the two-way
interaction between X1 and X2 is confounded by the two-way interaction
between X3 and X4.
References [1] Box, G. E. P., W. G. Hunter, and J. S. Hunter. Statistics for
Experimenters. Hoboken, NJ: Wiley-Interscience, 1978.
See Also fracfact, hadamard
18-444
friedman
Purpose Friedman’s test
Syntax p = friedman(X,reps)
p = friedman(X,reps,displayopt)
[p,table] = friedman(...)
[p,table,stats] = friedman(...)
Description p = friedman(X,reps) performs the nonparametric Friedman’s test
to compare column effects in a two-way layout. Friedman’s test is
similar to classical balanced two-way ANOVA, but it tests only for
column effects after adjusting for possible row effects. It does not test
for row effects or interaction effects. Friedman’s test is appropriate
when columns represent treatments that are under study, and rows
represent nuisance effects (blocks) that need to be taken into account
but are not of any interest.
The different columns of X represent changes in a factor A. The different
rows represent changes in a blocking factor B. If there is more than one
observation for each combination of factors, input reps indicates the
number of replicates in each “cell,” which must be constant.
The matrix below illustrates the format for a set-up where column
factor A has three levels, row factor B has two levels, and there are two
replicates (reps=2). The subscripts indicate row, column, and replicate,
respectively.
⎡ x111 x121 x131 ⎤
⎢x x122 x132 ⎥⎥
⎢ 112
⎢ x211 x221 x231 ⎥
⎢ ⎥
⎣ x212 x222 x232 ⎦
Friedman’s test assumes a model of the form
xijk = + i + j + ijk
18-445
friedman
where μ is an overall location parameter, i represents the column
effect, j represents the row effect, and ijk represents the error. This
test ranks the data within each level of B, and tests for a difference
across levels of A. The p that friedman returns is the p value for the
null hypothesis that i = 0 . If the p value is near zero, this casts doubt
on the null hypothesis. A sufficiently small p value suggests that at
least one column-sample median is significantly different than the
others; i.e., there is a main effect due to factor A. The choice of a critical
p value to determine whether a result is “statistically significant” is left
to the researcher. It is common to declare a result significant if the
p value is less than 0.05 or 0.01.
friedman also displays a figure showing an ANOVA table, which divides
the variability of the ranks into two or three parts:
• The variability due to the differences among the column effects
• The variability due to the interaction between rows and columns (if
reps is greater than its default value of 1)
• The remaining variability not explained by any systematic source
The ANOVA table has six columns:
• The first shows the source of the variability.
• The second shows the Sum of Squares (SS) due to each source.
• The third shows the degrees of freedom (df) associated with each
source.
• The fourth shows the Mean Squares (MS), which is the ratio SS/df.
• The fifth shows Friedman’s chi-square statistic.
• The sixth shows the p value for the chi-square statistic.
18-446
friedman
p = friedman(X,reps,displayopt) enables the ANOVA table display
when displayopt is 'on' (default) and suppresses the display when
displayopt is 'off'.
[p,table] = friedman(...) returns the ANOVA table (including
column and row labels) in cell array table. (You can copy a text version
of the ANOVA table to the clipboard by selecting Copy Text from the
Edit menu.
[p,table,stats] = friedman(...) returns a stats structure that
you can use to perform a follow-up multiple comparison test. The
friedman test evaluates the hypothesis that the column effects are
all the same against the alternative that they are not all the same.
Sometimes it is preferable to perform a test to determine which pairs of
column effects are significantly different, and which are not. You can
use the multcompare function to perform such tests by supplying the
stats structure as input.
Assumptions
Friedman’s test makes the following assumptions about the data in X:
• All data come from populations having the same continuous
distribution, apart from possibly different locations due to column
and row effects.
• All observations are mutually independent.
The classical two-way ANOVA replaces the first assumption with the
stronger assumption that data come from normal distributions.
Examples Let’s repeat the example from the anova2 function, this time applying
Friedman’s test. Recall that the data below come from a study of
popcorn brands and popper type (Hogg 1987). The columns of the
matrix popcorn are brands (Gourmet, National, and Generic). The
rows are popper type (Oil and Air). The study popped a batch of each
brand three times with each popper. The values are the yield in cups
of popped popcorn.
18-447
friedman
load popcorn
popcorn
popcorn =
5.5000 4.5000 3.5000
5.5000 4.5000 4.0000
6.0000 4.0000 3.0000
6.5000 5.0000 4.0000
7.0000 5.5000 5.0000
7.0000 5.0000 4.5000
p = friedman(popcorn,3)
p =
0.0010
The small p value of 0.001 indicates the popcorn brand affects the yield
of popcorn. This is consistent with the results from anova2.
References [1] Hogg, R. V., and J. Ledolter. Engineering Statistics. New York:
MacMillan, 1987.
[2] Hollander, M., and D. A. Wolfe. Nonparametric Statistical Methods.
Hoboken, NJ: John Wiley & Sons, Inc., 1999.
See Also anova2, multcompare, kruskalwallis
18-448
frnd
Purpose F random numbers
Syntax R = frnd(V1,V2)
R = frnd(V1,V2,v)
R = frnd(V1,V2,m,n)
Description R = frnd(V1,V2) generates random numbers from the F distribution
with numerator degrees of freedom V1 and denominator degrees of
freedom V2. V1 and V2 can be vectors, matrices, or multidimensional
arrays that all have the same size. A scalar input for V1 or V2 is
expanded to a constant array with the same dimensions as the other
input.
R = frnd(V1,V2,v) generates random numbers from the F distribution
with parameters V1 and V2, where v is a row vector. If v is a 1-by-2
vector, R is a matrix with v(1) rows and v(2) columns. If v is 1-by-n, R
is an n-dimensional array.
R = frnd(V1,V2,m,n) generates random numbers from the F
distribution with parameters V1 and V2, where scalars m and n are the
row and column dimensions of R.
Examples n1 = frnd(1:6,1:6)
n1 =
0.0022 0.3121 3.0528 0.3189 0.2715 0.9539
n2 = frnd(2,2,[2 3])
n2 =
0.3186 0.9727 3.0268
0.2052 148.5816 0.2191
n3 = frnd([1 2 3;4 5 6],1,2,3)
n3 =
0.6233 0.2322 31.5458
2.5848 0.2121 4.4955
See Also random, fpdf, fcdf, finv, fstat
18-449
frnd
“F Distribution” on page B-25
18-450
fstat
Purpose F mean and variance
Syntax [M,V] = fstat(V1,V2)
Description [M,V] = fstat(V1,V2) returns the mean of and variance for the F
distribution with numerator degrees of freedom V1 and denominator
degrees of freedom V2. V1 and V2 can be vectors, matrices, or
multidimensional arrays that all have the same size, which is also the
size of M and V. A scalar input for V1 or V2 is expanded to a constant
arrays with the same dimensions as the other input.
The mean of the F distribution for values of ν2 greater than 2 is
2
2 − 2
The variance of the F distribution for values of ν2 greater than 4 is
2 22 ( 1 + 2 − 2)
1 ( 2 − 2)2 ( 2 − 4)
The mean of the F distribution is undefined if ν2 is less than 3. The
variance is undefined for ν2 less than 5.
Examples fstat returns NaN when the mean and variance are undefined.
[m,v] = fstat(1:5,1:5)
m =
NaN NaN 3.0000 2.0000 1.6667
v =
NaN NaN NaN NaN 8.8889
See Also fpdf, fcdf, finv, frnd
“F Distribution” on page B-25
18-451
fsurfht
Purpose Interactive contour plot
Syntax fsurfht(fun,xlims,ylims)
fsurfht(fun,xlims,ylims,p1,p2,p3,p4,p5)
Description fsurfht(fun,xlims,ylims) is an interactive contour plot of the
function specified by the text variable fun. The x-axis limits are
specified by xlims in the form [xmin xmax], and the y-axis limits are
specified by ylims in the form [ymin ymax].
fsurfht(fun,xlims,ylims,p1,p2,p3,p4,p5) allows for five optional
parameters that you can supply to the function fun.
The intersection of the vertical and horizontal reference lines on the plot
defines the current x value and y value. You can drag these reference
lines and watch the calculated z-values (at the top of the plot) update
simultaneously. Alternatively, you can type the x value and y value into
editable text fields on the x-axis and y-axis.
Examples Plot the Gaussian likelihood function for the gas.mat data.
load gas
Create a function containing the following commands, and name it
gauslike.m.
function z = gauslike(mu,sigma,p1)
n = length(p1);
z = ones(size(mu));
for i = 1:n
z = z .* (normpdf(p1(i),mu,sigma));
end
The gauslike function calls normpdf, treating the data sample as fixed
and the parameters µ and σ as variables. Assume that the gas prices
are normally distributed, and plot the likelihood surface of the sample.
fsurfht('gauslike',[112 118],[3 5],price1)
18-452
fsurfht
The sample mean is the x value at the maximum, but the sample
standard deviation is not the y value at the maximum.
mumax = mean(price1)
mumax =
115.1500
sigmamax = std(price1)*sqrt(19/20)
sigmamax =
3.7719
18-453
fsurfht
18-454
fullfact
Purpose Full factorial design
Syntax dFF = fullfact(levels)
Description dFF = fullfact(levels) gives factor settings dFF for a full factorial
design with n factors, where the number of levels for each factor is given
by the vector levels of length n. dFF is m-by-n, where m is the number
of treatments in the full-factorial design. Each row of dFF corresponds
to a single treatment. Each column contains the settings for a single
factor, with integer values from one to the number of levels.
Examples The following generates an eight-run full-factorial design with two
levels in the first factor and four levels in the second factor:
dFF = fullfact([2 4])
dFF =
1 1
2 1
1 2
2 2
1 3
2 3
1 4
2 4
See Also ff2n
18-455
gagerr
Purpose Gage repeatability and reproducibility study
Syntax gagerr(y,{part,operator})
gagerr(y,GROUP)
gagerr(y,part)
gagerr(...,param1,val1,param2,val2,...)
[TABLE, stats] = gagerr(...)
Description gagerr(y,{part,operator}) performs a gage repeatability and
reproducibility study on measurements in y collected by operator on
part. y is a column vector containing the measurements on different
parts. part and operator are categorical variables, numeric vectors,
character matrices, or cell arrays of strings. The number of elements in
part and operator should be the same as in y.
gagerr prints a table in the command window in which the
decomposition of variance, standard deviation, study var (5.15 x
standard deviation) are listed with respective percentages for different
sources. Summary statistics are printed below the table giving the
number of distinct categories (NDC) and the percentage of Gage R&R
of total variations (PRR).
gagerr also plots a bar graph showing the percentage of different
components of variations. Gage R&R, repeatability, reproducibility, and
part-to-part variations are plotted as four vertical bars. Variance and
study var are plotted as two groups.
To determine the capability of a measurement system using NDC, use
the following guidelines:
• If NDC > 5, the measurement system is capable.
• If NDC < 2, the measurement system is not capable.
• Otherwise, the measurement system may be acceptable.
To determine the capability of a measurement system using PRR, use
the following guidelines:
18-456
gagerr
• If PRR < 10%, the measurement system is capable.
• If PRR > 30%, the measurement system is not capable.
• Otherwise, the measurement system may be acceptable.
gagerr(y,GROUP) performs a gage R&R study on measurements in
y with part and operator represented in GROUP. GROUP is a numeric
matrix whose first and second columns specify different parts and
operators, respectively. The number of rows in GROUP should be the
same as the number of elements in y. (See “Grouped Data” on page
2-34.)
gagerr(y,part) performs a gage R&R study on measurements in y
without operator information. The assumption is that all variability is
contributed by part.
gagerr(...,param1,val1,param2,val2,...) performs a gage R&R
study using one or more of the following parameter name/value pairs:
• 'spec' — A two-element vector that defines the lower and upper
limit of the process, respectively. In this case, summary statistics
printed in the command window include Precision-to-Tolerance
Ratio (PTR). Also, the bar graph includes an additional group, the
percentage of tolerance.
To determine the capability of a measurement system using PTR, use
the following guidelines:
- If PTR < 0.1, the measurement system is capable.
- If PTR > 0.3, the measurement system is not capable.
- Otherwise, the measurement system may be acceptable.
• 'printtable' — A string with a value 'on' or 'off' that indicates
whether the tabular output should be printed in the command
window or not. The default value is 'on'.
• 'printgraph' — A string with a value 'on' or 'off' that indicates
whether the bar graph should be plotted or not. The default value
is 'on'.
18-457
gagerr
• 'randomoperator' — A logical value, true or false, that indicates
whether the effect of operator is random or not. The default value
is true.
• 'model' — The model to use, specified by one of:
- 'linear' — Main effects only (default)
- 'interaction' — Main effects plus two-factor interactions
- 'nested' — Nest operator in part
The default value is 'linear'.
[TABLE, stats] = gagerr(...) returns a 6-by-5 matrix TABLE and
a structure stats. The columns of TABLE, from left to right, represent
variance, percentage of variance, standard deviations, study var,
and percentage of study var. The rows of TABLE, from top to bottom,
represent different sources of variations: gage R&R, repeatability,
reproducibility, operator, operator and part interactions, and part.
stats is a structure containing summary statistics for the performance
of the measurement system. The fields of stats are:
• ndc — Number of distinct categories
• prr — Percentage of gage R&R of total variations
• ptr — Precision-to-tolerance ratio. The value is NaN if the parameter
'spec' is not given.
Examples Conduct a gage R&R study for a simulated measurement system using
a mixed ANOVA model without interactions:
y = randn(100,1); % measurements
part = ceil(3*rand(100,1)); % parts
operator = ceil(4*rand(100,1)); % operators
gagerr(y,{part, operator},'randomoperator',true) % analysis
Source Variance % Variance sigma 5.15*sigma % 5.15*sigma
===================================================================================
18-458
gagerr
Gage R&R 0.77 100.00 0.88 4.51 100.00
Repeatability 0.76 99.08 0.87 4.49 99.54
Reproducibility 0.01 0.92 0.08 0.43 9.61
Operator 0.01 0.92 0.08 0.43 9.61
Part 0.00 0.00 0.00 0.00 0.00
Total 0.77 100.00 0.88 4.51
-----------------------------------------------------------------------------------
Number of distinct categories (NDC): 0
% of Gage R&R of total variations (PRR): 100.00
Note: The last column of the above table does not have to sum to 100%
18-459
gagerr
See Also “Grouped Data” on page 2-34
18-460
gamcdf
Purpose Gamma cumulative distribution function
Syntax gamcdf(X,A,B)
[P,PLO,PUP] = gamcdf(X,A,B,pcov,alpha)
Description gamcdf(X,A,B) computes the gamma cdf at each of the values in X
using the corresponding shape parameters in A and scale parameters
in B. X, A, and B can be vectors, matrices, or multidimensional arrays
that all have the same size. A scalar input is expanded to a constant
array with the same dimensions as the other inputs. The parameters in
A and B must be positive.
The gamma cdf is
x −t
1 a −1 b
p = F ( x | a, b) = ∫
ba Γ(a) 0
t e dt
The result, p, is the probability that a single observation from a gamma
distribution with parameters a and b will fall in the interval [0 x].
[P,PLO,PUP] = gamcdf(X,A,B,pcov,alpha) produces confidence
bounds for P when the input parameters A and B are estimates.
pcov is a 2-by-2 matrix containing the covariance matrix of the
estimated parameters. alpha has a default value of 0.05, and specifies
100(1-alpha)% confidence bounds. PLO and PUP are arrays of the same
size as P containing the lower and upper confidence bounds.
gammainc is the gamma distribution with b fixed at 1.
Examples a = 1:6;
b = 5:10;
prob = gamcdf(a.*b,a,b)
prob =
0.6321 0.5940 0.5768 0.5665 0.5595 0.5543
18-461
gamcdf
The mean of the gamma distribution is the product of the parameters,
ab. In this example, the mean approaches the median as it increases
(i.e., the distribution becomes more symmetric).
See Also cdf, gampdf, gaminv, gamstat, gamfit, gamlike, gamrnd, gamma
“Gamma Distribution” on page B-27
18-462
gamfit
Purpose Gamma parameter estimates
Syntax phat = gamfit(data)
[phat,pci] = gamfit(data)
[phat,pci] = gamfit(data,alpha)
[...] = gamfit(data,alpha,censoring,freq,options)
Description phat = gamfit(data) returns the maximum likelihood estimates
(MLEs) for the parameters of the gamma distribution given the data
in vector data.
[phat,pci] = gamfit(data) returns MLEs and 95% percent
confidence intervals. The first row of pci is the lower bound of the
confidence intervals; the last row is the upper bound.
[phat,pci] = gamfit(data,alpha) returns 100(1 - alpha)%
confidence intervals. For example, alpha = 0.01 yields 99% confidence
intervals.
[...] = gamfit(data,alpha,censoring) accepts a Boolean vector of
the same size as data that is 1 for observations that are right-censored
and 0 for observations that are observed exactly.
[...] = gamfit(data,alpha,censoring,freq) accepts a frequency
vector of the same size as data. freq typically contains integer
frequencies for the corresponding elements in data, but may contain
any nonnegative values.
[...] = gamfit(data,alpha,censoring,freq,options) accepts a
structure, options, that specifies control parameters for the iterative
algorithm the function uses to compute maximum likelihood estimates.
The gamma fit function accepts an options structure which can be
created using the function statset. Enter statset('gamfit') to see
the names and default values of the parameters that gamfit accepts
in the options structure.
Examples Fit a gamma distribution to random data generated from a specified
gamma distribution:
18-463
gamfit
a = 2; b = 4;
data = gamrnd(a,b,100,1);
[p,ci] = gamfit(data)
p =
2.1990 3.7426
ci =
1.6840 2.8298
2.7141 4.6554
References [1] Hahn, Gerald J., and S. S. Shapiro. Statistical Models in
Engineering. Hoboken, NJ: John Wiley & Sons, Inc., 1994, p. 88.
See Also mle, gamlike, gampdf, gamcdf, gaminv, gamstat, gamrnd
“Gamma Distribution” on page B-27
18-464
gaminv
Purpose Gamma inverse cumulative distribution function
Syntax X = gaminv(P,A,B)
[X,XLO,XUP] = gamcdf(P,A,B,pcov,alpha)
Description X = gaminv(P,A,B) computes the inverse of the gamma cdf
with shape parameters in A and scale parameters in B for the
corresponding probabilities in P. P, A, and B can be vectors, matrices, or
multidimensional arrays that all have the same size. A scalar input is
expanded to a constant array with the same dimensions as the other
inputs. The parameters in A and B must all be positive, and the values
in P must lie on the interval [0 1].
The gamma inverse function in terms of the gamma cdf is
x = F −1 ( p| a, b) = { x : F ( x | a, b) = p}
where
x −t
1 a −1 b
p = F ( x | a, b) = ∫
ba Γ(a) 0
t e dt
[X,XLO,XUP] = gamcdf(P,A,B,pcov,alpha) produces confidence
bounds for P when the input parameters A and B are estimates.
pcov is a 2-by-2 matrix containing the covariance matrix of the
estimated parameters. alpha has a default value of 0.05, and specifies
100(1-alpha)% confidence bounds. PLO and PUP are arrays of the same
size as P containing the lower and upper confidence bounds.
Algorithm There is no known analytical solution to the integral equation above.
gaminv uses an iterative approach (Newton’s method) to converge on
the solution.
Examples This example shows the relationship between the gamma cdf and its
inverse function.
18-465
gaminv
a = 1:5;
b = 6:10;
x = gaminv(gamcdf(1:5,a,b),a,b)
x =
1.0000 2.0000 3.0000 4.0000 5.0000
See Also icdf, gamcdf, gampdf, gamstat, gamfit, gamlike, gamrnd
“Gamma Distribution” on page B-27
18-466
gamlike
Purpose Gamma negative log-likelihood
Syntax nlogL = gamlike(params,data)
[nlogL,AVAR] = gamlike(params,data)
Description nlogL = gamlike(params,data) returns the negative of the gamma
log-likelihood of the parameters, params, given data. params(1)=A,
shape parameters, and params(2)=B, scale parameters.
[nlogL,AVAR] = gamlike(params,data) also returns AVAR, which is
the asymptotic variance-covariance matrix of the parameter estimates
when the values in params are the maximum likelihood estimates. AVAR
is the inverse of Fisher’s information matrix. The diagonal elements of
AVAR are the asymptotic variances of their respective parameters.
[...] = gamlike(params,data,censoring) accepts a Boolean
vector of the same size as data that is 1 for observations that are
right-censored and 0 for observations that are observed exactly.
[...] = gamfit(params,data,censoring,freq) accepts a frequency
vector of the same size as data. freq typically contains integer
frequencies for the corresponding elements in data, but may contain
any non-negative values.
gamlike is a utility function for maximum likelihood estimation of
the gamma distribution. Since gamlike returns the negative gamma
log-likelihood function, minimizing gamlike using fminsearch is the
same as maximizing the likelihood.
Examples Compute the negative log-likelihood of parameter estimates computed
by the gamfit function:
a = 2; b = 3;
r = gamrnd(a,b,100,1);
[nlogL,AVAR] = gamlike(gamfit(r),r)
nlogL =
267.5648
AVAR =
18-467
gamlike
0.0788 -0.1104
-0.1104 0.1955
See Also gamfit, gampdf, gamcdf, gaminv, gamstat, gamrnd
“Gamma Distribution” on page B-27
18-468
gampdf
Purpose Gamma probability density function
Syntax Y = gampdf(X,A,B)
Description Y = gampdf(X,A,B) computes the gamma pdf at each of the values in X
using the corresponding shape parameters in A and scale parameters in
B. X, A, and B can be vectors, matrices, or multidimensional arrays that
all have the same size. A scalar input is expanded to a constant array
with the same dimensions as the other inputs. The parameters in A and
B must all be positive, and the values in X must lie on the interval [0 ∞).
The gamma pdf is
−x
1 a −1
y = f ( x | a, b) = x eb
ba Γ(a)
The gamma probability density function is useful in reliability models of
lifetimes. The gamma distribution is more flexible than the exponential
distribution in that the probability of a product surviving an additional
period may depend on its current age. The exponential and χ2 functions
are special cases of the gamma function.
Examples The exponential distribution is a special case of the gamma distribution.
mu = 1:5;
y = gampdf(1,1,mu)
y =
0.3679 0.3033 0.2388 0.1947 0.1637
y1 = exppdf(1,mu)
y1 =
0.3679 0.3033 0.2388 0.1947 0.1637
See Also pdf, gamcdf, gaminv, gamstat, gamfit, gamlike, gamrnd
“Gamma Distribution” on page B-27
18-469
gamrnd
Purpose Gamma random numbers
Syntax R = gamrnd(A,B)
R = gamrnd(A,B,v)
R = gamrnd(A,B,m,n)
Description R = gamrnd(A,B) generates random numbers from the gamma
distribution with shape parameters in A and scale parameters in B. A
and B can be vectors, matrices, or multidimensional arrays that all have
the same size. A scalar input for A or B is expanded to a constant array
with the same dimensions as the other input.
R = gamrnd(A,B,v) generates random numbers from the gamma
distribution with parameters A and B, where v is a row vector. If v is a
1-by-2 vector, R is a matrix with v(1) rows and v(2) columns. If v is
1-by-n, R is an n-dimensional array.
R = gamrnd(A,B,m,n) generates gamma random numbers with
parameters A and B, where scalars m and n are the row and column
dimensions of R.
Examples n1 = gamrnd(1:5,6:10)
n1 =
9.1132 12.8431 24.8025 38.5960 106.4164
n2 = gamrnd(5,10,[1 5])
n2 =
30.9486 33.5667 33.6837 55.2014 46.8265
n3 = gamrnd(2:6,3,1,5)
n3 =
12.8715 11.3068 3.0982 15.6012 21.6739
See Also randg, random, gampdf, gamcdf, gaminv, gamstat, gamfit, gamlike
“Gamma Distribution” on page B-27
18-470
gamstat
Purpose Gamma mean and variance
Syntax [M,V] = gamstat(A,B)
Description [M,V] = gamstat(A,B) returns the mean of and variance for the
gamma distribution with shape parameters in A and scale parameters
in B. A and B can be vectors, matrices, or multidimensional arrays that
have the same size, which is also the size of M and V. A scalar input
for A or B is expanded to a constant array with the same dimensions
as the other input.
The mean of the gamma distribution with parameters a and b is ab.
The variance is ab2.
Examples [m,v] = gamstat(1:5,1:5)
m =
1 4 9 16 25
v =
1 8 27 64 125
[m,v] = gamstat(1:5,1./(1:5))
m =
1 1 1 1 1
v =
1.0000 0.5000 0.3333 0.2500 0.2000
See Also gampdf, gamcdf, gaminv, gamfit, gamlike, gamrnd
“Gamma Distribution” on page B-27
18-471
qrandstream.ge
Purpose Greater than or equal relation for handles
Syntax h1 >= h2
Description h1 >= h2 performs element-wise comparisons between handle arrays
h1 and h2. h1 and h2 must be of the same dimensions unless one is a
scalar. The result is a logical array of the same dimensions, where each
element is an element-wise >= result.
If one of h1 or h2 is scalar, scalar expansion is performed and the result
will match the dimensions of the array that is not scalar.
tf = ge(h1, h2) stores the result in a logical array of the same
dimensions.
See Also qrandstream, eq, gt, le, lt, ne
18-472
geocdf
Purpose Geometric cumulative distribution function
Syntax Y = geocdf(X,P)
Description Y = geocdf(X,P) computes the geometric cdf at each of the values in
X using the corresponding probabilities in P. X and P can be vectors,
matrices, or multidimensional arrays that all have the same size. A
scalar input is expanded to a constant array with the same dimensions
as the other input. The parameters in P must lie on the interval [0 1].
The geometric cdf is
floor ( x)
y = F ( x | p) = ∑ pqi
i =0
where q = 1 − p .
The result, y, is the probability of observing up to x trials before a
success, when the probability of success in any given trial is p.
Examples Suppose you toss a fair coin repeatedly. If the coin lands face up (heads),
that is a success. What is the probability of observing three or fewer
tails before getting a heads?
p = geocdf(3,0.5)
p =
0.9375
See Also cdf, geopdf, geoinv, geostat, geornd, mle
“Geometric Distribution” on page B-41
18-473
geoinv
Purpose Geometric inverse cumulative distribution function
Syntax X = geoinv(Y,P)
Description X = geoinv(Y,P) returns the smallest positive integer X such that the
geometric cdf evaluated at X is equal to or exceeds Y. You can think of
Y as the probability of observing X successes in a row in independent
trials where P is the probability of success in each trial.
Y and P can be vectors, matrices, or multidimensional arrays that all
have the same size. A scalar input for P or Y is expanded to a constant
array with the same dimensions as the other input. The values in P
and Y must lie on the interval [0 1].
Examples The probability of correctly guessing the result of 10 coin tosses in a row
is less than 0.001 (unless the coin is not fair).
psychic = geoinv(0.999,0.5)
psychic =
9
The example below shows the inverse method for generating random
numbers from the geometric distribution.
rndgeo = geoinv(rand(2,5),0.5)
rndgeo =
0 1 3 1 0
0 1 0 2 0
See Also icdf, geocdf, geopdf, geostat, geornd
“Geometric Distribution” on page B-41
18-474
geomean
Purpose Geometric mean
Syntax m = geomean(x)
geomean(X,dim)
Description m = geomean(x) calculates the geometric mean of a sample. For
vectors, geomean(x) is the geometric mean of the elements in x. For
matrices, geomean(X) is a row vector containing the geometric means of
each column. For N-dimensional arrays, geomean operates along the
first nonsingleton dimension of X.
geomean(X,dim) takes the geometric mean along the dimension dim
of X.
The geometric mean is
1
⎡ n ⎤n
m = ⎢∏ xi ⎥
⎢⎣ i=1 ⎥⎦
Examples The arithmetic mean is greater than or equal to the geometric mean.
x = exprnd(1,10,6);
geometric = geomean(x)
geometric =
0.7466 0.6061 0.6038 0.2569 0.7539 0.3478
average = mean(x)
average =
1.3509 1.1583 0.9741 0.5319 1.0088 0.8122
See Also mean, median, harmmean, trimmean
“Geometric Distribution” on page B-41
18-475
geopdf
Purpose Geometric probability density function
Syntax Y = geopdf(X,P)
Description Y = geopdf(X,P) computes the geometric pdf at each of the values in
X using the corresponding probabilities in P. X and P can be vectors,
matrices, or multidimensional arrays that all have the same size. A
scalar input is expanded to a constant array with the same dimensions
as the other input. The parameters in P must lie on the interval [0 1].
The geometric pdf is
y = f ( x | p) = pq x I(0,1,...) ( x)
where q = 1 − p .
Examples Suppose you toss a fair coin repeatedly. If the coin lands face up
(heads), that is a success. What is the probability of observing exactly
three tails before getting a heads?
p = geopdf(3,0.5)
p =
0.0625
See Also pdf, geocdf, geoinv, geostat, geornd
“Geometric Distribution” on page B-41
18-476
geornd
Purpose Geometric random numbers
Syntax R = geornd(P)
R = geornd(P,v)
R = geornd(P,m,n)
Description R = geornd(P) generates geometric random numbers with probability
parameter P. P can be a vector, a matrix, or a multidimensional array.
The size of R is the size of P. The geometric distribution is useful when
you want to model the number of successive failures preceding a success,
where the probability of success in any given trial is the constant P.
R = geornd(P,v) generates geometric random numbers with
probability parameter P, where v is a row vector. If v is a 1-by-2 vector,
R is a matrix with v(1) rows and v(2) columns. If v is 1-by-n, R is an
n-dimensional array.
R = geornd(P,m,n) generates geometric random numbers with
probability parameter P, where scalars m and n are the row and column
dimensions of R.
The parameters in P must lie on the interval [0 1].
Examples r1 = geornd(1 ./ 2.^(1:6))
r1 =
2 10 2 5 2 60
r2 = geornd(0.01,[1 5])
r2 =
65 18 334 291 63
r3 = geornd(0.5,1,6)
r3 =
0 7 1 3 1 0
See Also random, geopdf, geocdf, geoinv, geostat
“Geometric Distribution” on page B-41
18-477
geostat
Purpose Geometric mean and variance
Syntax [M,V] = geostat(P)
Description [M,V] = geostat(P) returns the mean of and variance for the
geometric distribution with corresponding probabilities in P.
The mean of the geometric distribution with parameter p is q/p, where q
= 1-p. The variance is q/p2.
Examples [m,v] = geostat(1./(1:6))
m =
0 1.0000 2.0000 3.0000 4.0000 5.0000
v =
0 2.0000 6.0000 12.0000 20.0000 30.0000
See Also geopdf, geocdf, geoinv, geornd
“Geometric Distribution” on page B-41
18-478
dataset.get
Purpose Access dataset array properties
Syntax get(A)
s = get(A)
p = get(A,PropertyName)
p = get(A,{PropertyName1,PropertyName2,...})
Description get(A) displays a list of property/value pairs for the dataset array A.
s = get(A) returns the values in a scalar structure s with field names
given by the properties.
p = get(A,PropertyName) returns the value of the property specified
by the string PropertyName.
p = get(A,{PropertyName1,PropertyName2,...}) allows multiple
property names to be specified and returns their values in a cell array.
Examples Create a dataset array from Fisher’s iris data and access the
information:
load fisheriris
NumObs = size(meas,1);
NameObs = strcat({'Obs'},num2str((1:NumObs)','%-d'));
iris = dataset({nominal(species),'species'},...
{meas,'SL','SW','PL','PW'},...
'ObsNames',NameObs);
get(iris)
Description: ''
Units: {}
DimNames: {'Observations' 'Variables'}
UserData: []
ObsNames: {150x1 cell}
VarNames: {'species' 'SL' 'SW' 'PL' 'PW'}
ON = get(iris,'ObsNames');
ON(1:3)
18-479
dataset.get
ans =
'Obs1'
'Obs2'
'Obs3'
See Also set, summary
18-480
categorical.getlabels
Purpose Access categorical array labels
Syntax labels = getlabels(A)
Description labels = getlabels(A) returns the labels of the levels in the
categorical array A as a cell array of strings labels. For ordinal A, the
labels are returned in the order of the levels.
Examples Example 1
Display levels in a nominal and an ordinal array:
standings = nominal({'Leafs','Canadiens','Bruins'});
getlabels(standings)
ans =
'Bruins' 'Canadiens' 'Leafs'
standings = ordinal(1:3,{'Leafs','Canadiens','Bruins'});
getlabels(standings)
ans =
'Leafs' 'Canadiens' 'Bruins'
Example 2
Display age groups containing data in hospital.mat:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
AgeGroup = ordinal(hospital.Age,labels,[],edges);
AgeGroup = droplevels(AgeGroup);
getlabels(AgeGroup)
ans =
'20s' '30s' '40s' '50s'
See Also getlevels, setlabels
18-481
categorical.getlevels
Purpose Get categorical array levels
Syntax S = getlevels(A)
Description S = getlevels(A) returns the levels for the categorical array A. S is
a vector with the same type as A.
See Also getlabels
18-482
gevcdf
Purpose Generalized extreme value cumulative distribution function
Syntax P = gevcdf(X,K,sigma,mu)
Description P = gevcdf(X,K,sigma,mu) returns the cdf of the generalized extreme
value (GEV) distribution with shape parameter K, scale parameter
sigma, and location parameter, mu, evaluated at the values in X. The
size of P is the common size of the input arguments. A scalar input
functions as a constant matrix of the same size as the other inputs.
Default values for K, sigma, and mu are 0, 1, and 0, respectively.
When K < 0, the GEV is the type III extreme value distribution. When
K > 0, the GEV distribution is the type II, or Frechet, extreme value
distribution. If w has a Weibull distribution as computed by the wblcdf
function, then -w has a type III extreme value distribution and 1/w has
a type II extreme value distribution. In the limit as K approaches 0, the
GEV is the mirror image of the type I extreme value distribution as
computed by the evcdf function.
The mean of the GEV distribution is not finite when K ≥ 1, and the
variance is not finite when K ≥ 1/2. The GEV distribution has positive
density only for values of X such that K*(X-mu)/sigma > -1.
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also cdf, gevpdf, gevinv, gevstat, gevfit, gevlike, gevrnd
“Generalized Extreme Value Distribution” on page B-32
18-483
gevfit
Purpose Generalized extreme value parameter estimates
Syntax parmhat = gevfit(X)
[parmhat,parmci] = gevfit(X)
[parmhat,parmci] = gevfit(X,alpha)
[...] = gevfit(X,alpha,options)
Description parmhat = gevfit(X) returns maximum likelihood estimates of the
parameters for the generalized extreme value (GEV) distribution given
the data in X. parmhat(1) is the shape parameter, K, parmhat(2) is the
scale parameter, sigma, and parmhat(3) is the location parameter, mu.
[parmhat,parmci] = gevfit(X) returns 95% confidence intervals for
the parameter estimates.
[parmhat,parmci] = gevfit(X,alpha) returns 100(1-alpha)%
confidence intervals for the parameter estimates.
[...] = gevfit(X,alpha,options) specifies control parameters for
the iterative algorithm used to compute ML estimates. This argument
can be created by a call to statset. See statset('gevfit') for
parameter names and default values. Pass in [] for alpha to use the
default values.
When K < 0, the GEV is the type III extreme value distribution. When
K > 0, the GEV distribution is the type II, or Frechet, extreme value
distribution. If w has a Weibull distribution as computed by the wblfit
function, then -w has a type III extreme value distribution and 1/w has
a type II extreme value distribution. In the limit as K approaches 0, the
GEV is the mirror image of the type I extreme value distribution as
computed by the evfit function.
The mean of the GEV distribution is not finite when K ≥ 1, and the
variance is not finite when K ≥ 1/2. The GEV distribution is defined
for K*(X-mu)/sigma > -1.
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
18-484
gevfit
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also mle, gevlike, gevpdf, gevcdf, gevinv, gevstat, gevrnd
“Generalized Extreme Value Distribution” on page B-32
18-485
gevinv
Purpose Generalized extreme value inverse cumulative distribution function
Syntax X = gevinv(P,K,sigma,mu)
Description X = gevinv(P,K,sigma,mu) returns the inverse cdf of the generalized
extreme value (GEV) distribution with shape parameter K, scale
parameter sigma, and location parameter mu, evaluated at the values
in P. The size of X is the common size of the input arguments. A scalar
input functions as a constant matrix of the same size as the other inputs.
Default values for K, sigma, and mu are 0, 1, and 0, respectively.
When K < 0, the GEV is the type III extreme value distribution. When
K > 0, the GEV distribution is the type II, or Frechet, extreme value
distribution. If w has a Weibull distribution as computed by the wblinv
function, then -w has a type III extreme value distribution and 1/w has
a type II extreme value distribution. In the limit as K approaches 0, the
GEV is the mirror image of the type I extreme value distribution as
computed by the evinv function.
The mean of the GEV distribution is not finite when K ≥ 1, and the
variance is not finite when K ≥ 1/2. The GEV distribution has positive
density only for values of X such that K*(X-mu)/sigma > -1.
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also icdf, gevcdf, gevpdf, gevstat, gevfit, gevlike, gevrnd
“Generalized Extreme Value Distribution” on page B-32
18-486
gevlike
Purpose Generalized extreme value negative log-likelihood
Syntax nlogL = gevlike(params,data)
[nlogL,ACOV] = gevlike(params,data)
Description nlogL = gevlike(params,data) returns the negative of the
log-likelihood nlogL for the generalized extreme value (GEV)
distribution, evaluated at parameters params. params(1) is the shape
parameter, K, params(2) is the scale parameter, mu, and params(3)
is the location parameter, sigma.
[nlogL,ACOV] = gevlike(params,data) returns the inverse of
Fisher’s information matrix, ACOV. If the input parameter values in
params are the maximum likelihood estimates, the diagonal elements
of ACOV are their asymptotic variances. ACOV is based on the observed
Fisher’s information, not the expected information.
When K < 0, the GEV is the type III extreme value distribution. When
K > 0, the GEV distribution is the type II, or Frechet, extreme value
distribution. If w has a Weibull distribution as computed by the wbllike
function, then -w has a type III extreme value distribution and 1/w has
a type II extreme value distribution. In the limit as K approaches 0, the
GEV is the mirror image of the type I extreme value distribution as
computed by the evlike function.
The mean of the GEV distribution is not finite when K ≥ 1, and the
variance is not finite when K ≥ 1/2. The GEV distribution has positive
density only for values of X such that K*(X-mu)/sigma > -1.
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah.Extreme Value Distributions: Theory and
Applications. London: Imperial College Press, 2000.
See Also gevfit, gevpdf, gevcdf, gevinv, gevstat, gevrnd
“Generalized Extreme Value Distribution” on page B-32
18-487
gevpdf
Purpose Generalized extreme value probability density function
Syntax Y = gevpdf(X,K,sigma,mu)
Description Y = gevpdf(X,K,sigma,mu) returns the pdf of the generalized extreme
value (GEV) distribution with shape parameter K, scale parameter
sigma, and location parameter, mu, evaluated at the values in X. The
size of Y is the common size of the input arguments. A scalar input
functions as a constant matrix of the same size as the other inputs.
Default values for K, sigma, and mu are 0, 1, and 0, respectively.
When K < 0, the GEV is the type III extreme value distribution. When
K > 0, the GEV distribution is the type II, or Frechet, extreme value
distribution. If w has a Weibull distribution as computed by the wblpdf
function, then -w has a type III extreme value distribution and 1/w has
a type II extreme value distribution. In the limit as K approaches 0, the
GEV is the mirror image of the type I extreme value distribution as
computed by the evcdf function.
The mean of the GEV distribution is not finite when K ≥ 1, and the
variance is not finite when K ≥ 1/2. The GEV distribution has positive
density only for values of X such that K*(X-mu)/sigma > -1.
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also pdf, gevcdf, gevinv, gevstat, gevfit, gevlike, gevrnd
“Generalized Extreme Value Distribution” on page B-32
18-488
gevrnd
Purpose Generalized extreme value random numbers
Syntax R = gevrnd(K,sigma,mu)
R = gevrnd(K,sigma,mu,m,n,p)
R = gevrnd(K,sigma,mu,[m,n,p])
Description R = gevrnd(K,sigma,mu) returns an array of random numbers chosen
from the generalized extreme value (GEV) distribution with shape
parameter K, scale parameter sigma, and location parameter, mu. The
size of R is the common size of the input arguments if all are arrays.
If any parameter is a scalar, the size of R is the size of the other
parameters.
R = gevrnd(K,sigma,mu,m,n,p) or
R = gevrnd(K,sigma,mu,[m,n,p]) returns an m-by-n-by-p array.
When K < 0, the GEV is the type III extreme value distribution. When
K > 0, the GEV distribution is the type II, or Frechet, extreme value
distribution. If w has a Weibull distribution as computed by the wblrnd
function, then -w has a type III extreme value distribution and 1/w has
a type II extreme value distribution. In the limit as K approaches 0, the
GEV is the mirror image of the type I extreme value distribution as
computed by the evrnd function.
The mean of the GEV distribution is not finite when K ≥ 1, and the
variance is not finite when K ≥ 1/2. The GEV distribution has positive
density only for values of X such that K*(X-mu)/sigma > -1.
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also random, gevpdf, gevcdf, gevinv, gevstat, gevfit, gevlike
“Generalized Extreme Value Distribution” on page B-32
18-489
gevstat
Purpose Generalized extreme value mean and variance
Syntax [M,V] = gevstat(K,sigma,mu)
Description [M,V] = gevstat(K,sigma,mu) returns the mean of and variance for
the generalized extreme value (GEV) distribution with shape parameter
K, scale parameter sigma, and location parameter, mu. The sizes of M and
V are the common size of the input arguments. A scalar input functions
as a constant matrix of the same size as the other inputs.
Default values for K, sigma, and mu are 0, 1, and 0, respectively.
When K < 0, the GEV is the type III extreme value distribution. When
K > 0, the GEV distribution is the type II, or Frechet, extreme value
distribution. If w has a Weibull distribution as computed by the wblstat
function, then -w has a type III extreme value distribution and 1/w has
a type II extreme value distribution. In the limit as K approaches 0, the
GEV is the mirror image of the type I extreme value distribution as
computed by the evstat function.
The mean of the GEV distribution is not finite when K ≥ 1, and the
variance is not finite when K ≥ 1/2. The GEV distribution has positive
density only for values of X such that K*(X-mu)/sigma > -1.
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also gevpdf, gevcdf, gevinv, gevfit, gevlike, gevrnd
“Generalized Extreme Value Distribution” on page B-32
18-490
gline
Purpose Interactively add line to plot
Syntax gline(h)
gline
hline = gline(...)
Description gline(h) allows you to draw a line segment in the figure with handle h
by clicking the pointer at the two endpoints. A rubber-band line tracks
the pointer movement.
gline with no input arguments defaults to h = gcf and draws in the
current figure.
hline = gline(...) returns the handle hline to the line.
Examples Use gline to connect two points in a plot:
x = 1:10;
y = x + randn(1,10);
scatter(x,y,25,'b','*')
lsline
mu = mean(y);
hold on
plot([1 10],[mu mu],'ro')
hline = gline; % Connect circles
set(hline,'Color','r')
18-491
gline
See Also refline, refcurve, lsline
18-492
glmfit
Purpose Generalized linear model regression
Syntax b = glmfit(X,y,distr)
b = glmfit(X,y,distr,param1,val1,param2,val2,...)
[b,dev] = glmfit(...)
[b,dev,stats] = glmfit(...)
Description b = glmfit(X,y,distr) returns a p-by-1 vector b of coefficient
estimates for a generalized linear regression of the responses in y on the
predictors in X, using the distribution distr. X is an n-by-p matrix of p
predictors at each of n observations. distr can be any of the following
strings: 'binomial', 'gamma', 'inverse gaussian', 'normal' (the
default), and 'poisson'.
In most cases, y is an n-by-1 vector of observed responses. For the
binomial distribution, y can be a binary vector indicating success or
failure at each observation, or a two column matrix with the first
column indicating the number of successes for each observation and the
second column indicating the number of trials for each observation.
This syntax uses the canonical link (see below) to relate the distribution
to the predictors.
Note By default, glmfit adds a first column of 1s to X, corresponding
to a constant term in the model. Do not enter a column of 1s directly
into X. You can change the default behavior of glmfit using the
'constant' parameter, below.
glmfit treats NaNs in either X or y as missing values, and ignores them.
b = glmfit(X,y,distr,param1,val1,param2,val2,...) additionally
allows you to specify optional parameter name/value pairs to control the
model fit. Acceptable parameters are as follows:
18-493
glmfit
Parameter Value Description
'link' 'identity', default µ = Xb
for the distribution
'normal'
'log', default for log(µ) = Xb
the distribution
'poisson'
'logit', default log(µ/(1 – µ)) = Xb
for the distribution
'binomial'
'probit' norminv(µ) = Xb
'comploglog' log( -log(1 – µ)) = Xb
'reciprocal' 1/µ = Xb
'loglog', default log( -log(µ)) = Xb
for the distribution
'gamma'
p (a number), default µp = Xb
for the distribution
'inverse
gaussian' (with
p = -2)
cell array of the User-specified link function
form {FL FD FI},
containing three
function handles,
created using @, that
define the link (FL),
the derivative of the
link (FD), and the
inverse link (FI).
18-494
glmfit
Parameter Value Description
'estdisp' 'on' Estimates a dispersion
parameter for the binomial
or Poisson distribution
'off' (Default for Uses the theoretical value of
binomial or Poisson 1.0 for those distributions
distribution)
'offset' Vector Used as an additional
predictor variable, but with a
coefficient value fixed at 1.0
'weights' Vector of prior
weights, such as
the inverses of the
relative variance of
each observation
'constant' 'on' (default) Includes a constant term in
the model. The coefficient of
the constant term is the first
element of b.
'off' Omit the constant term
[b,dev] = glmfit(...)returns dev, the deviance of the fit at the
solution vector. The deviance is a generalization of the residual sum of
squares. It is possible to perform an analysis of deviance to compare
several models, each a subset of the other, and to test whether the model
with more terms is significantly better than the model with fewer terms.
[b,dev,stats] = glmfit(...) returns dev and stats.
stats is a structure with the following fields:
• beta — Coefficient estimates b
• dfe — Degrees of freedom for error
• s — Theoretical or estimated dispersion parameter
18-495
glmfit
• sfit — Estimated dispersion parameter
• se — Vector of standard errors of the coefficient estimates b
• coeffcorr — Correlation matrix for b
• covb — Estimated covariance matrix for B
• t — t statistics for b
• p — p-values for b
• resid — Vector of residuals
• residp — Vector of Pearson residuals
• residd — Vector of deviance residuals
• resida — Vector of Anscombe residuals
If you estimate a dispersion parameter for the binomial or Poisson
distribution, then stats.s is set equal to stats.sfit. Also, the
elements of stats.se differ by the factor stats.s from their theoretical
values.
Example Fit a probit regression model for y on x. Each y(i) is the number
of successes in n(i) trials.
x = [2100 2300 2500 2700 2900 3100 ...
3300 3500 3700 3900 4100 4300]';
n = [48 42 31 34 31 21 23 23 21 16 17 21]';
y = [1 2 0 3 8 8 14 17 19 15 17 21]';
b = glmfit(x,[y n],'binomial','link','probit');
yfit = glmval(b, x,'probit','size', n);
plot(x, y./n,'o',x,yfit./n,'-','LineWidth',2)
18-496
glmfit
References [1] Dobson, A. J. An Introduction to Generalized Linear Models. New
York: Chapman & Hall, 1990.
[2] McCullagh, P., and J. A. Nelder. Generalized Linear Models. New
York: Chapman & Hall, 1990.
[3] Collett, D. Modeling Binary Data. New York: Chapman & Hall,
2002.
See Also glmval, regress, regstats
18-497
glmval
Purpose Generalized linear model values
Syntax yhat = glmval(b,X,link)
[yhat,dylo,dyhi] = glmval(b,X,link,stats)
[...] = glmval(...,param1,val1,param2,val2,...)
Description yhat = glmval(b,X,link) computes predicted values for the
generalized linear model with link function link and predictors X.
Distinct predictor variables should appear in different columns of X. b
is a vector of coefficient estimates as returned by the glmfit function.
link can be any of the strings used as values for the link parameter in
the glmfit function.
Note By default, glmval adds a first column of 1s to X, corresponding
to a constant term in the model. Do not enter a column of 1s directly
into X. You can change the default behavior of glmval using the
'constant' parameter, below.
[yhat,dylo,dyhi] = glmval(b,X,link,stats) also computes 95%
confidence bounds for the predicted values. When the stats structure
output of the glmfit function is specified, dylo and dyhi are also
returned. dylo and dyhi define a lower confidence bound of yhat-dylo,
and an upper confidence bound of yhat+dyhi. Confidence bounds
are nonsimultaneous, and apply to the fitted curve, not to a new
observation.
[...] = glmval(...,param1,val1,param2,val2,...) specifies
optional parameter name/value pairs to control the predicted values.
Acceptable parameters are:
18-498
glmval
Parameter Value
'confidence' — the confidence A scalar between 0 and 1
level for the confidence bounds
'size' — the size parameter (N) A scalar, or a vector with one
for a binomial model value for each row of X
'offset' — used as an additional A vector
predictor variable, but with a
coefficient value fixed at 1.0
'constant' • 'on' — Includes a constant
term in the model. The
coefficient of the constant term
is the first element of b.
• 'off' — Omit the constant
term
Examples Fit a probit regression model for y on x. Each y(i) is the number
of successes in n(i) trials.
x = [2100 2300 2500 2700 2900 3100 ...
3300 3500 3700 3900 4100 4300]';
n = [48 42 31 34 31 21 23 23 21 16 17 21]';
y = [1 2 0 3 8 8 14 17 19 15 17 21]';
b = glmfit(x,[y n],'binomial','link','probit');
yfit = glmval(b,x,'probit','size',n);
plot(x, y./n,'o',x,yfit./n,'-','LineWidth',2)
18-499
glmval
References [1] Dobson, A. J. An Introduction to Generalized Linear Models. New
York: Chapman & Hall, 1990.
[2] McCullagh, P., and J. A. Nelder. Generalized Linear Models. New
York: Chapman & Hall, 1990.
[3] Collett, D. Modeling Binary Data. New York: Chapman & Hall,
2002.
See Also glmfit
18-500
glyphplot
Purpose Glyph plot
Syntax glyphplot(X)
glyphplot(X,'glyph','face')
glyphplot(X,'glyph','face','features',f)
glyphplot(X,...,'grid',[rows,cols])
glyphplot(X,...,'grid',[rows,cols],'page',p)
glyphplot(X,...,'centers',C)
glyphplot,...,'centers',C,'radius',r)
glyphplot(X,...,'obslabels',labels)
glyphplot(X,...,'standardize',method)
glyphplot(X,...,prop1,val1,...)
h = glyphplot(X,...)
Description glyphplot(X) creates a star plot from the multivariate data in the
n-by-p matrix X. Rows of X correspond to observations, columns
to variables. A star plot represents each observation as a “star”
whose ith spoke is proportional in length to the ith coordinate of that
observation. glyphplot standardizes X by shifting and scaling each
column separately onto the interval [0,1] before making the plot, and
centers the glyphs on a rectangular grid that is as close to square as
possible. glyphplot treats NaNs in X as missing values, and does not
plot the corresponding rows of X. glyphplot(X,'glyph','star') is a
synonym for glyphplot(X).
glyphplot(X,'glyph','face') creates a face plot from X. A face plot
represents each observation as a “face,” whose ith facial feature is
drawn with a characteristic proportional to the ith coordinate of that
observation. The features are described in “Face Features” on page
18-503Face Features.
glyphplot(X,'glyph','face','features',f) creates a face plot
where the ith element of the index vector f defines which facial feature
will represent the ith column of X. f must contain integers from 0 to
17, where 0 indicate that the corresponding column of X should not be
plotted. See “Face Features” on page 18-503 for more information.
18-501
glyphplot
glyphplot(X,...,'grid',[rows,cols]) organizes the glyphs into a
rows-by-cols grid.
glyphplot(X,...,'grid',[rows,cols],'page',p) organizes the
glyph into one or more pages of a rows-by-cols grid, and displays the
page p. If p is a vector, glyphplot displays multiple pages in succession.
If p is 'all', glyphplot displays all pages. If p is 'scroll', glyphplot
displays a single plot with a scrollbar.
glyphplot(X,...,'centers',C) creates a plot with each glyph
centered at the locations in the n-by-2 matrix C.
glyphplot,...,'centers',C,'radius',r) creates a plot with glyphs
positioned using C, and scale the glyphs so the largest has radius r.
glyphplot(X,...,'obslabels',labels) labels each glyph with the
text in the character array or cell array of strings labels. By default,
the glyphs are labelled 1:N. Use '' for blank labels.
glyphplot(X,...,'standardize',method) standardizes X before
making the plot. Choices for method are
• 'column' — Maps each column of X separately onto the interval
[0,1]. This is the default.
• 'matrix' — Maps the entire matrix X onto the interval [0,1].
• 'PCA' — Transforms X to its principal component scores, in order of
decreasing eigenvalue, and maps each one onto the interval [0,1].
• 'off' — No standardization. Negative values in X may make a star
plot uninterpretable.
glyphplot(X,...,prop1,val1,...) sets properties to the specified
property values for all line graphics objects created by glyphplot.
h = glyphplot(X,...) returns a matrix of handles to the graphics
objects created by glyphplot. For a star plot, h(:,1) and h(:,2)
contain handles to the line objects for each star’s perimeter and spokes,
respectively. For a face plot, h(:,1) and h(:,2) contain object handles
18-502
glyphplot
to the lines making up each face and to the pupils, respectively. h(:,3)
contains handles to the text objects for the labels, if present.
Face Features
The following table describes the correspondence between the columns
of the vector f, the value of the 'Features' input parameter, and the
facial features of the glyph plot. If X has fewer than 17 columns, unused
features are displayed at their default value.
Column Facial Feature
1 Size of face
2 Forehead/jaw relative arc length
3 Shape of forehead
4 Shape of jaw
5 Width between eyes
6 Vertical position of eyes
7 Height of eyes
8 Width of eyes (this also affects eyebrow width)
9 Angle of eyes (this also affects eyebrow angle)
10 Vertical position of eyebrows
11 Width of eyebrows (relative to eyes)
12 Angle of eyebrows (relative to eyes)
13 Direction of pupils
14 Length of nose
15 Vertical position of mouth
16 Shape of mouth
17 Mouth arc length
18-503
glyphplot
Examples load carsmall
X = [Acceleration Displacement Horsepower MPG Weight];
glyphplot(X,'standardize','column',...
'obslabels',Model,...
'grid',[2 2],...
'page','scroll');
glyphplot(X,'glyph','face',...
'obslabels',Model,...
'grid',[2 3],...
'page',9);
18-504
glyphplot
See Also andrewsplot, parallelcoords
18-505
gmdistribution class
Purpose Gaussian mixture models
Description An object of the gmdistribution class defines a Gaussian mixture
distribution, which is a multivariate distribution that consists of a
mixture of one or more multivariate Gaussian distribution components.
The number of components for a given gmdistribution object is fixed.
Each multivariate Gaussian component is defined by its mean and
covariance, and the mixture is defined by a vector of mixing proportions.
Construction To create a Gaussian mixture distribution by specifying the distribution
parameters, use the gmdistribution constructor. To fit a Gaussian
mixture distribution model to data, use gmdistribution.fit.
fit Gaussian mixture parameter
estimates
gmdistribution Construct Gaussian mixture
distribution
Methods cdf Cumulative distribution function
for Gaussian mixture distribution
cluster Construct clusters from Gaussian
mixture distribution
disp Display Gaussian mixture
distribution object
display Display Gaussian mixture
distribution object
fit Gaussian mixture parameter
estimates
mahal Mahalanobis distance to
component means
18-506
gmdistribution class
pdf Probability density function for
Gaussian mixture distribution
posterior Posterior probabilities of
components
random Random numbers from Gaussian
mixture distribution
subsasgn Subscripted reference for
Gaussian mixture distribution
object
subsref Subscripted reference for
Gaussian mixture distribution
object
Properties All objects of the class have the properties listed in the following table.
CovType Type of covariance matrices
DistName Type of distribution
Mu Input matrix of means MU
NComponents Number k of mixture components
NDimensions Dimension d of multivariate
Gaussian distributions
PComponents Input vector of mixing proportions
SharedCov true if all covariance matrices
are restricted to be the same
Sigma Input array of covariances
Objects constructed with fit have the additional properties listed in
the following table.
18-507
gmdistribution class
AIC Akaike Information Criterion
BIC Bayes Information Criterion
Converged Determine if algorithm converged
Iters Number of iterations
NlogL Negative of log-likelihood
RegV Value of 'Regularize' parameter
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
References McLachlan, G., and D. Peel, Finite Mixture Models, John Wiley & Sons,
New York, 2000.
See Also “Normal Distribution” on page B-83
18-508
gmdistribution
Purpose Construct Gaussian mixture distribution
Syntax obj = gmdistribution(mu,sigma,p)
Description obj = gmdistribution(mu,sigma,p) constructs an object obj of the
gmdistribution class defining a Gaussian mixture distribution.
mu is a k-by-d matrix specifying the d-dimensional mean of each of the
k components.
sigma specifies the covariance of each component. The size of sigma is:
• d-by-d-by-k if there are no restrictions on the form of the covariance.
In this case, sigma(:,:,I) is the covariance of component I.
• 1-by-d-by-k if the covariance matrices are restricted to be diagonal,
but not restricted to be same across components. In this case,
sigma(:,:,I) contains the diagonal elements of the covariance of
component I.
• d-by-d matrix if the covariance matrices are restricted to be the same
across components, but not restricted to be diagonal. In this case,
sigma is the pooled estimate of covariance.
• 1-by-d if the covariance matrices are restricted to be diagonal and the
same across components. In this case, sigma contains the diagonal
elements of the pooled estimate of covariance.
p is an optional 1-by-k vector specifying the mixing proportions of each
component. If p does not sum to 1, gmdistribution normalizes it. The
default is equal proportions.
Examples Create a gmdistribution object defining a two-component mixture of
bivariate Gaussian distributions:
mu = [1 2;-3 -5];
sigma = cat(3,[2 0;0 .5],[1 0;0 1]);
p = ones(1,2)/2;
obj = gmdistribution(mu,sigma,p);
18-509
gmdistribution
ezsurf(@(x,y)pdf(obj,[x y]),[-10 10],[-10 10])
References [1] McLachlan, G., and D. Peel. Finite Mixture Models. Hoboken, NJ:
John Wiley & Sons, Inc., 2000.
See Also fit, pdf, cdf, random, cluster, posterior, mahal
18-510
gname
Purpose Add case names to plot
Syntax gname(cases)
gname
h = gname(cases,line_handle)
Description gname(cases) displays a figure window and waits for you to press
a mouse button or a keyboard key. The input argument cases is a
character array or a cell array of strings, in which each row of the
character array or each element of the cell array contains the case
name of a point. Moving the mouse over the graph displays a pair of
cross-hairs. If you position the cross-hairs near a point with the mouse
and click once, the graph displays the label corresponding to that point.
Alternatively, you can click and drag the mouse to create a rectangle
around several points. When you release the mouse button, the graph
displays the labels for all points in the rectangle. Right-click a point to
remove its label. When you are done labelling points, press the Enter
or Escape key to stop labeling.
gname with no arguments labels each case with its case number.
cases typically contains unique case names for each point, and is a cell
array of strings or a character matrix with each row representing a
name. cases can also be any grouping variable, which gname converts
to labels.
h = gname(cases,line_handle) returns a vector of handles to the text
objects on the plot. Use the scalar line_handle to identify the correct
line if there is more than one line object on the plot.
You can use gname to label plots created by the plot, scatter,
gscatter, plotmatrix, and gplotmatrix functions.
Examples This example uses the city ratings data sets to find out which cities are
the best and worst for education and the arts.
load cities
education = ratings(:,6);
arts = ratings(:,7);
18-511
gname
plot(education,arts,'+')
gname(names)
Click the point at the top of the graph to display its label, “New York.”
See Also gtext, gscatter, gplotmatrix
18-512
gpcdf
Purpose Generalized Pareto cumulative distribution function
Syntax P = gpcdf(X,K,sigma,theta)
Description P = gpcdf(X,K,sigma,theta) returns the cdf of the generalized
Pareto (GP) distribution with the tail index (shape) parameter K,
scale parameter sigma, and threshold (location) parameter, theta,
evaluated at the values in X. The size of P is the common size of the
input arguments. A scalar input functions as a constant matrix of the
same size as the other inputs.
Default values for K, sigma, and theta are 0, 1, and 0, respectively.
When K = 0 and theta = 0, the GP is equivalent to the exponential
distribution. When K > 0 and theta = sigma/K, the GP is equivalent
to the Pareto distribution. The mean of the GP is not finite when K ≥
1, and the variance is not finite when K ≥ 1/2. When K ≥ 0, the GP has
positive density for
X > theta, or, when
X −θ 1
K < 0, 0 ≤ ≤− .
σ K
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also cdf, gppdf, gpinv, gpstat, gpfit, gplike, gprnd
“Generalized Pareto Distribution” on page B-37
18-513
gpfit
Purpose Generalized Pareto parameter estimates
Syntax parmhat = gpfit(X)
[parmhat,parmci] = gpfit(X)
[parmhat,parmci] = gpfit(X,alpha)
[...] = gpfit(X,alpha,options)
Description parmhat = gpfit(X) returns maximum likelihood estimates of the
parameters for the two-parameter generalized Pareto (GP) distribution
given the data in X. parmhat(1) is the tail index (shape) parameter,
K and parmhat(2) is the scale parameter, sigma. gpfit does not fit
a threshold (location) parameter.
[parmhat,parmci] = gpfit(X) returns 95% confidence intervals for
the parameter estimates.
[parmhat,parmci] = gpfit(X,alpha) returns 100(1-alpha)%
confidence intervals for the parameter estimates.
[...] = gpfit(X,alpha,options) specifies control parameters for
the iterative algorithm used to compute ML estimates. This argument
can be created by a call to statset. See statset('gpfit') for
parameter names and default values.
Other functions for the generalized Pareto, such as gpcdf allow a
threshold parameter, theta. However, gpfit does not estimate theta.
It is assumed to be known, and subtracted from X before calling gpfit.
When K = 0 and theta = 0, the GP is equivalent to the exponential
distribution. When K > 0 and theta = sigma/K, the GP is equivalent
to the Pareto distribution. The mean of the GP is not finite when K ≥
1, and the variance is not finite when K ≥ 1/2. When K ≥ 0, the GP has
positive density for
X > theta, or, when
X −θ 1
0≤ ≤−
σ K
18-514
gpfit
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also mle, gplike, gppdf, gpcdf, gpinv, gpstat, gprnd
“Generalized Pareto Distribution” on page B-37
18-515
gpinv
Purpose Generalized Pareto inverse cumulative distribution function
Syntax X = gpinv(P,K,sigma,theta)
Description X = gpinv(P,K,sigma,theta) returns the inverse cdf for a generalized
Pareto (GP) distribution with tail index (shape) parameter K, scale
parameter sigma, and threshold (location) parameter theta, evaluated
at the values in P. The size of X is the common size of the input
arguments. A scalar input functions as a constant matrix of the same
size as the other inputs.
Default values for K, sigma, and theta are 0, 1, and 0, respectively.
When K = 0 and theta = 0, the GP is equivalent to the exponential
distribution. When K > 0 and theta = sigma/K, the GP is equivalent
to the Pareto distribution. The mean of the GP is not finite when K ≥
1, and the variance is not finite when K ≥ 1/2. When K ≥ 0, the GP has
positive density for
X > theta, or, when
X −θ 1
K < 0, 0 ≤ ≤− .
σ K
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also icdf, gpcdf, gppdf, gpstat, gpfit, gplike, gprnd
“Generalized Pareto Distribution” on page B-37
18-516
gplike
Purpose Generalized Pareto negative log-likelihood
Syntax nlogL = gplike(params,data)
[nlogL,ACOV] = gplike(params,data)
Description nlogL = gplike(params,data) returns the negative of the
log-likelihood nlogL for the two-parameter generalized Pareto (GP)
distribution, evaluated at parameters params. params(1) is the tail
index (shape) parameter, K, params(2) is the scale parameter, sigma,
and params(3) is the threshold (location) parameter, mu.
[nlogL,ACOV] = gplike(params,data) returns the inverse of Fisher’s
information matrix, ACOV. If the input parameter values in params are
the maximum likelihood estimates, the diagonal elements of ACOV are
their asymptotic variances. ACOV is based on the observed Fisher’s
information, not the expected information.
When K = 0 and theta = 0, the GP is equivalent to the exponential
distribution. When K > 0 and theta = sigma/K, the GP is equivalent
to the Pareto distribution. The mean of the GP is not finite when K ≥
1, and the variance is not finite when K ≥ 1/2. When K ≥ 0, the GP has
positive density for
X > theta, or, when
X −θ 1
K < 0, 0 ≤ ≤− .
σ K
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also gpfit, gppdf, gpcdf, gpinv, gpstat, gprnd
“Generalized Pareto Distribution” on page B-37
18-517
gppdf
Purpose Generalized Pareto probability density function
Syntax P = gppdf(X,K,sigma,theta)
Description P = gppdf(X,K,sigma,theta) returns the pdf of the generalized
Pareto (GP) distribution with the tail index (shape) parameter K,
scale parameter sigma, and threshold (location) parameter, theta,
evaluated at the values in X. The size of P is the common size of the
input arguments. A scalar input functions as a constant matrix of the
same size as the other inputs.
Default values for K, sigma, and theta are 0, 1, and 0, respectively.
When K = 0 and theta = 0, the GP is equivalent to the exponential
distribution. When K > 0 and theta = sigma/K, the GP is equivalent
to the Pareto distribution. The mean of the GP is not finite when K ≥
1, and the variance is not finite when K ≥ 1/2. When K ≥ 0, the GP has
positive density for
X > theta, or, when
X −θ 1
K < 0, 0 ≤ ≤− .
σ K
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also pdf, gpcdf, gpinv, gpstat, gpfit, gplike, gprnd
“Generalized Pareto Distribution” on page B-37
18-518
gplotmatrix
Purpose Matrix of scatter plots by group
Syntax gplotmatrix(x,y,group)
gplotmatrix(x,y,group,clr,sym,siz)
gplotmatrix(x,y,group,clr,sym,siz,doleg)
gplotmatrix(x,y,group,clr,sym,siz,doleg,dispopt)
gplotmatrix(x,y,group,clr,sym,siz,doleg,dispopt,xnam,ynam)
[h,ax,bigax] = gplotmatrix(...)
Description gplotmatrix(x,y,group) creates a matrix of scatter plots. Each
individual set of axes in the resulting figure contains a scatter plot of a
column of x against a column of y. All plots are grouped by the grouping
variable group. (See “Grouped Data” on page 2-34.)
x and y are matrices with the same number of rows. If x has p columns
and y has q columns, the figure contains a p-by-q matrix of scatter plots.
If you omit y or specify it as the empty matrix, [], gplotmatrix creates
a square matrix of scatter plots of columns of x against each other.
group is a grouping variable that can be a categorical variable, vector,
string array, or cell array of strings. group must have the same number
of rows as x and y. Points with the same value of group are placed
in the same group, and appear on the graph with the same marker
and color. Alternatively, group can be a cell array containing several
grouping variables (such as {g1 g2 g3}); in that case, observations are
in the same group if they have common values of all grouping variables.
gplotmatrix(x,y,group,clr,sym,siz) specifies the color, marker
type, and size for each group. clr is a string array of colors recognized
by the plot function. The default for clr is 'bgrcmyk'. sym is a string
array of symbols recognized by the plot command, with the default
value '.'. siz is a vector of sizes, with the default determined by
the DefaultLineMarkerSize property. If you do not specify enough
values for all groups, gplotmatrix cycles through the specified values
as needed.
gplotmatrix(x,y,group,clr,sym,siz,doleg) controls whether a
legend is displayed on the graph (doleg is 'on', the default) or not
(doleg is 'off').
18-519
gplotmatrix
gplotmatrix(x,y,group,clr,sym,siz,doleg,dispopt) controls what
appears along the diagonal of a plot matrix of y versus x. Allowable
values are 'none', to leave the diagonals blank, 'hist', to plot
histograms, or 'variable', to write the variable names. gplotmatrix
displays histograms along the diagonal only when there is only one
variable (i.e., gplotmatrix(x,[],[],[],[],[],'hist').
gplotmatrix(x,y,group,clr,sym,siz,doleg,dispopt,xnam,ynam)
specifies the names of the columns in the x and y arrays. These names
are used to label the x- and y-axes. xnam and ynam must be character
arrays or cell arrays of strings, with one name for each column of x
and y, respectively.
[h,ax,bigax] = gplotmatrix(...) returns three arrays of handles.
h is an array of handles to the lines on the graphs. The array’s third
dimension corresponds to groups in G. ax is a matrix of handles to the
axes of the individual plots. If dispopt is 'hist', ax contains one extra
row of handles to invisible axes in which the histograms are plotted.
bigax is a handle to big (invisible) axes framing the entire plot matrix.
bigax is fixed to point to the current axes, so a subsequent title,
xlabel, or ylabel command will produce labels that are centered with
respect to the entire plot matrix.
Examples Load the cities data. The ratings array has ratings of the cities in
nine categories (category names are in the array categories). group
is a code whose value is 2 for the largest cities. You can make scatter
plots of the first three categories against the other four, grouped by
the city size code:
load discrim
gplotmatrix(ratings(:,1:2),ratings(:,[4 7]),group)
The output figure (not shown) has an array of graphs with each city
group represented by a different color. The graphs are a little easier to
read if you specify colors and plotting symbols, label the axes with the
rating categories, and move the legend off the graphs:
gplotmatrix(ratings(:,1:2),ratings(:,[4 7]),group,...
18-520
gplotmatrix
'br','.o',[],'on','',categories(1:2,:),...
categories([4 7],:))
See Also “Grouped Data” on page 2-34
grpstats, gscatter, plotmatrix
18-521
gprnd
Purpose Generalized Pareto random numbers
Syntax R = gprnd(K,sigma,theta)
R = gprnd(K,sigma,theta,m,n,p)
R = gprnd(K,sigma,theta,[m,n,p])
Description R = gprnd(K,sigma,theta) returns an array of random numbers
chosen from the generalized Pareto (GP) distribution with tail index
(shape) parameter K, scale parameter sigma, and threshold (location)
parameter, theta. The size of R is the common size of the input
arguments if all are arrays. If any parameter is a scalar, the size of R is
the size of the other parameters.
Default values for K, sigma, and theta are 0, 1, and 0, respectively.
R = gprnd(K,sigma,theta,m,n,p) or R =
gprnd(K,sigma,theta,[m,n,p]) returns an m-by-n-by-p array.
When K = 0 and theta = 0, the GP is equivalent to the exponential
distribution. When K > 0 and theta = sigma/K, the GP is equivalent
to the Pareto distribution. The mean of the GP is not finite when K ≥
1, and the variance is not finite when K ≥ 1/2. When K ≥ 0, the GP has
positive density for
X > theta, or, when
X −θ 1
0≤ ≤−
σ K
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also random, gppdf, gpcdf, gpinv, gpstat, gpfit, gplike
18-522
gpstat
Purpose Generalized Pareto mean and variance
Syntax [M,V] = gpstat(K,sigma,theta)
Description [M,V] = gpstat(K,sigma,theta) returns the mean of and variance
for the generalized Pareto (GP) distribution with the tail index
(shape) parameter K, scale parameter sigma, and threshold (location)
parameter, theta.
The default value for theta is 0.
When K = 0 and theta = 0, the GP is equivalent to the exponential
distribution. When K > 0 and theta = sigma/K, the GP is equivalent
to the Pareto distribution. The mean of the GP is not finite when K ≥
1, and the variance is not finite when K ≥ 1/2. When K ≥ 0, the GP has
positive density for X > theta, or when
X −θ 1
K < 0, 0 ≤ ≤− .
σ K
References [1] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
[2] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory
and Applications. London: Imperial College Press, 2000.
See Also gppdf, gpcdf, gpinv, gpfit, gplike, gprnd
18-523
TreeBagger.growTrees
Purpose Train additional trees and add to ensemble
Syntax B = growTrees(B,ntrees)
B = growTrees(B,ntrees,’param1’,val1,’param2’,val2,...)
Description B = growTrees(B,ntrees) grows ntrees new trees and appends them
to those trees already stored in the ensemble B.
B = growTrees(B,ntrees,’param1’,val1,’param2’,val2,...)
pecifies optional parameter name/value pairs:
'nprint' Specifies that a diagnostic
message showing training
progress should display after
every value training cycles
(grown trees). Default is no
diagnostic messages.
'options' A struct that specifies options
that govern computation when
growing the ensemble of decision
trees. One option requests that
the computation of decision trees
on multiple bootstrap replicates
uses multiple processors, if the
Parallel Computing Toolbox is
available. Two options specify the
random number streams to use
in selecting bootstrap replicates.
You can create this argument
with a call to statset. You can
retrieve values of the individual
fields with a call to statget.
Applicable statset parameters
are:
• 'UseParallel' — If 'always'
and if a matlabpool of the
18-524
TreeBagger.growTrees
Parallel Computing Toolbox
is open, compute decision
trees drawn on separate
boostrap replicates in parallel.
If the Parallel Computing
Toolbox is not installed, or
a matlabpool is not open,
computation occurs in serial
mode. Default is 'never', or
serial computation.
• 'UseSubstreams' — If
'always' select each bootstrap
replicate using a separate
Substream of the random
number generator (aka
Stream). This option is
available only with RandStream
types that support Substreams.
Default is 'never', do not use
a different Substream to
compute each bootstrap
replicate.
• 'Streams' — An object of the
RandStream class, or a cell
array of RandStream objects.
Default is an empty cell array.
If you do not supply a value for
this parameter, TreeBagger
uses the default RandStream on
each MATLAB executable in
selecting bootstrap replicates.
Otherwise, TreeBagger selects
bootstrap replicates using the
supplied RandStream object(s).
If you select 'UseSubstreams',
the Streams parameter, if
18-525
TreeBagger.growTrees
present, must be a scalar
RandStream object. If you do
not select 'UseSubstreams',
then the Streams parameter,
if present, must match the
number of processors used for
the computation. For serial
computation, the Streams
parameter must be a scalar.
If computation is distributed
('UseParallel' is 'always'
and a matlabpool is open),
then the Streams parameter
must be a cell array of the
same length as the matlabpool
size. In this case, each element
of the cell array supplies the
random number generator for
bootstrap sampling on one of
the parallel workers.
See Also classregtree
18-526
grp2idx
Purpose Create index vector from grouping variable
Syntax [G,GN]=grp2idx(S)
[indices,names] = grp2idx(group)
[G,GN,GL] = grp2idx(S)
Description [G,GN]=grp2idx(S) creates an index vector G from the grouping
variable S. S can be a categorical, numeric, or logical vector; a cell vector
of strings; or a character matrix with each row representing a group
label. The result G is a vector taking integer values from 1 up to the
number K of distinct groups. GN is a cell array of strings representing
group labels. If S is a character matrix, GN(G,:) reproduces S, otherwise
GN(G) reproduces S (aside from any differences in type).
The order of indices depends on the grouping variable:
• For numeric and logical grouping variables, the order is the sorted
order of group.
• For categorical grouping variables, the order is the order of
getlabels(group).
• For string grouping variables, the order is the order of first
appearance in group.
[indices,names] = grp2idx(group) also returns a cell array of
group names, so that names(indices) reproduces group, apart from
differences in type.
[G,GN,GL] = grp2idx(S) returns a column vector GL representing the
group levels. The set of groups and their order in GL and GN are the
same, except that GL has the same type as S. If S is a character matrix,
GL(G,:) reproduces S, otherwise GL(G) reproduces S.
grp2idx treats NaNs (numeric or logical), empty strings (char or cell
array of strings), or values (categorical) in S as missing values and
returns NaNs in the corresponding rows of G. GN and GL don’t include
entries for missing values.
18-527
grp2idx
Examples Load the data in hospital.mat and create a categorical grouping
variable:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
AgeGroup = ordinal(hospital.Age,labels,[],edges);
ages = hospital.Age(1:5)
ages =
38
43
38
40
49
group = AgeGroup(1:5)
group =
30s
40s
30s
40s
40s
indices = grp2idx(group)
indices =
4
5
4
5
5
See Also “Grouped Data” on page 2-34
crosstab, getlabels, grpstats, gscatter
18-528
grpstats
Purpose Summary statistics by group
Syntax means = grpstats(X)
means = grpstats(X,group)
grpstats(X,group,alpha)
dsstats = grpstats(ds,groupvars)
[A,B,...] = grpstats(X,group,whichstats)
[...] = grpstats(...,whichstats,'Param1',VAL1,'Param2',VAL2,
...)
Description means = grpstats(X) computes the mean of the entire sample without
grouping, where X is a matrix of observations.
means = grpstats(X,group) returns the means of each column of X by
group. The array, group defines the grouping such that two elements
of X are in the same group if their corresponding group values are
the same. (See “Grouped Data” on page 2-34.) The grouping variable
group can be a categorical variable, vector, string array, or cell array
of strings. It can also be a cell array containing several grouping
variables (such as {g1 g2 g3}) to group the values in X by each unique
combination of grouping variable values.
grpstats(X,group,alpha) displays a plot of the means versus index
with 100(1-alpha)% confidence intervals around each mean.
dsstats = grpstats(ds,groupvars), when ds is a dataset array,
returns a dataset dsstats that contains the mean, computed by group,
for variables in ds. groupvars specifies the grouping variables in ds
that define the groups, and is a positive integer, a vector of positive
integers, the nam of a dataset variable, a cell array containing one or
more dataset variable names, or a logical vector. A grouping variable
may be a vector of categorical, logical, or numeric values, a character
array of strings, or a cell vector of strings. dsstats contains those
grouping variables, plus one variable giving the number of observations
in ds for each group, as well as one variable for each of the remaining
dataset variables in ds. These variables must be numeric or logical.
dsstats contains one observation for each group of observations in ds.
18-529
grpstats
groupvars can be [] or omitted to compute the mean of each variable
across the entire dataset without grouping.
grpstats treats NaNs as missing values, and removes them.
grpstats ignores empty group names.
[A,B,...] = grpstats(X,group,whichstats) returns the statistics
specified in whichstats. The input whichstats can be a single function
handle or name, or a cell array containing multiple function handles or
names. The number of outputs (A,B, ...) must match the number function
handles and names in whichstats. Acceptable names are as follows:
• 'mean' — mean
• 'sem' — standard error of the mean
• 'numel' — count, or number of non-NaN elements
• 'gname' — group name
• 'std' — standard deviation
• 'var' — variance
• 'min' — minimum
• 'max' — maximum
• 'range' — maximum - minimum
• 'meanci' — 95% confidence interval for the mean
• 'predci' — 95% prediction interval for a new observation
Each function included in whichstats must accept a column vector of
data and compute a descriptive statistic for it. For example, @median
and @skewness are suitable functions to apply to a numeric input. A
function must return the same size output each time grpstats calls
it, even if the input for some groups is empty. The function typically
returns a scalar value, but may return an nvals-by-1 column vector
if the descriptive statistic is not a scalar (a confidence interval, for
example). The size of each output A, B, ... is ngroups-by-ncols-by-nvals,
18-530
grpstats
where ngroups is the number of groups, ncols is the number of columns
in the data X, and nvals is the number of values returned by the function
for data from a single group in one column of X. If X is a vector of data,
then the size of each output A, B, .... is ngroups-by-nvals.
A function included in whichstats may also be written to accept a
matrix of data and compute a descriptive statistic for each column. The
function should return either a row vector, or an nvals-by-ncols matrix
if the descriptive statistic is not a scalar.
For the case when data are contained in a numeric matrix X, a function
specified in whichstats may also be written to accept a matrix of
data and ompute a descriptive statistic for each column. The function
should return either a row vector, or an nvals-by-ncols matrix if the
descriptive statistic is not a scalar.
[...] =
grpstats(...,whichstats,'Param1',VAL1,'Param2',VAL2,...)
specifies additional parameter name/value pairs chosen
from the following:
'Alpha' A value from 0 to 1 that
specifies the confidence level as
100(1-alpha)% for the 'meanci'
and 'predci' options. Default is
0.05.
'DataVars' The names of the variables in
ds to which the functions in
whichstats should be applied.
dsstats contains one summary
statistic variable for each of
these data variables. datavars
is a positive integer, a vector
of positive integers, a variable
name, a cell array containing
one or more variable names, or a
logical vector.
18-531
grpstats
'VarNames' The names of the variables in
dsstats. By default, grpstats
uses the names from ds for
the grouping variable names,
and constructs names for the
summary statistic variables
based on the function name and
the data variable names from ds.
dsstats contains ngroupvars + 1 + ndatavars*nfuns variables,
where ngroupvars is the number of variables specified in groupvars,
ndatavars is the number of variables specified in datavars, and nfuns
is the number of summary statistics specified in whichstats.
Examples load carsmall
[m,p,g] = grpstats(Weight,Model_Year,...
{'mean','predci','gname'})
n = length(m)
errorbar((1:n)',m,p(:,2)-m)
set(gca,'xtick',1:n,'xticklabel',g)
title('95% prediction intervals for mean weight by year')
18-532
grpstats
See Also “Grouped Data” on page 2-34
gscatter, grp2idx, dataset.grpstats
18-533
dataset.grpstats
Purpose Summary statistics by group for dataset arrays
Syntax B = grpstats(A,groupvars)
B = grpstats(A,groupvars,whichstats)
B = grpstats(A,groupvars,whichstats,...,'DataVars',vars)
B = grpstats(A,groupvars,whichstats,...,'VarNames',names)
Description B = grpstats(A,groupvars) returns a dataset array B that contains
the means, computed by group, for variables in the dataset array A.
The optional input groupvars specifies the variables in A that define
the groups. groupvars can be a positive integer, a vector of positive
integers, a variable name, a cell array containing one or more variable
names, or a logical vector. groupvars can also be [] or omitted to
compute the means of the variables in A without grouping. Grouping
variables can be vectors of categorical, logical, or numeric values, a
character array of strings, or a cell vector of strings. (See “Grouped
Data” on page 2-34.)
B contains the grouping variables, plus a variable giving the number
of observations in A for each group, plus a variable for each of the
remaining variables in A. B contains one observation for each group
of observations in A.
grpstats treats NaNs as missing values, and removes them.
B = grpstats(A,groupvars,whichstats) returns a dataset array B
with variables for each of the statistics specified in whichstats, applied
to each of the nongrouping variables in A. whichstats can be a single
function handle or name, or a cell array containing multiple function
handles or names. The names can be chosen from among the following:
• 'mean' — mean
• 'sem' — standard error of the mean
• 'numel' — count, or number of non-NaN elements
• 'gname' — group name
• 'std' — standard deviation
18-534
dataset.grpstats
• 'var' — variance
• 'meanci' — 95% confidence interval for the mean
• 'predci' — 95% prediction interval for a new observation
Each function included in whichstats must accept a subset of the rows
of a dataset variable, and compute column-wise descriptive statistics for
it. A function should typically return a value that has one row but is
otherwise the same size as its input data. For example, @median and
@skewness are suitable functions to apply to a numeric dataset variable.
A summary statistic function may also return values with more
than one row, provided the return values have the same number of
rows each time grpstats applies the function to different subsets of
data from a given dataset variable. For a dataset variable that is
nobs-by-m-by-... if a summary statistic function returns values that are
nvals-by-m-by-... then the corresponding summary statistic variable
in B is ngroups-by-m-by-...-by-nvals, where ngroups is the number of
groups in A.
B = grpstats(A,groupvars,whichstats,...,'DataVars',vars)
specifies the variables in A to which the functions in whichstats should
be applied. The output dataset arrays contain one summary statistic
variable for each of the specified variables. vars is a positive integer,
a vector of positive integers, a variable name, a cell array containing
one or more variable names, or a logical vector.
B = grpstats(A,groupvars,whichstats,...,'VarNames',names)
specifies the names of the variables in B. By default, grpstats uses
the names from A for the grouping variables, and constructs names for
the summary statistic variables based on the function name and the
data variable names. The number of variables in B is ngroupvars + 1
+ ndatavars*nfuns, where ngroupvars is the number of variables
specified in groupvars, ndatavars is the number of variables specified
in vars, and nfuns is the number of summary statistics specified in
whichstats.
18-535
dataset.grpstats
Examples Compute blood pressure statistics for the data in hospital.mat, by sex
and smoker status:
load hospital
grpstats(hospital,...
{'Sex','Smoker'},...
{@median,@iqr},...
'DataVars','BloodPressure')
ans =
Sex Smoker GroupCount
Female_0 Female false 40
Female_1 Female true 13
Male_0 Male false 26
Male_1 Male true 21
median_BloodPressure
Female_0 119.5 79
Female_1 129 91
Male_0 119 79
Male_1 129 92
iqr_BloodPressure
Female_0 6.5 5.5
Female_1 8 5.5
Male_0 7 6
Male_1 10.5 4.5
See Also grpstats, summary
18-536
gscatter
Purpose Scatter plot by group
Syntax gscatter(x,y,group)
gscatter(x,y,group,clr,sym,siz)
gscatter(x,y,group,clr,sym,siz,doleg)
gscatter(x,y,group,clr,sym,siz,doleg,xnam,ynam)
h = gscatter(...)
Description gscatter(x,y,group) creates a scatter plot of x and y, grouped by
group. x and y are vectors of the same size. group is a grouping variable
in the form of a categorical variable, vector, string array, or cell array of
strings. (See “Grouped Data” on page 2-34.) Alternatively, group can be
a cell array containing several grouping variables (such as {g1 g2 g3}),
in which case observations are in the same group if they have common
values of all grouping variables. Points in the same group and appear
on the graph with the same marker and color.
gscatter(x,y,group,clr,sym,siz) specifies the color, marker type,
and size for each group. clr is a string array of colors recognized by
the plot function. The default for clr is 'bgrcmyk'. sym is a string
array of symbols recognized by the plot command, with the default
value '.'. siz is a vector of sizes, with the default determined by the
'DefaultLineMarkerSize' property. If you do not specify enough
values for all groups, gscatter cycles through the specified values as
needed.
gscatter(x,y,group,clr,sym,siz,doleg) controls whether a
legend is displayed on the graph (doleg is 'on', the default) or not
(doleg is 'off').
gscatter(x,y,group,clr,sym,siz,doleg,xnam,ynam) specifies the
name to use for the x-axis and y-axis labels. If the x and y inputs are
simple variable names and xnam and ynam are omitted, gscatter labels
the axes with the variable names.
h = gscatter(...) returns an array of handles to the lines on the
graph.
18-537
gscatter
Examples Load the cities data and look at the relationship between the ratings
for climate (first column) and housing (second column) grouped by city
size. We’ll also specify the colors and plotting symbols.
load discrim
gscatter(ratings(:,1),ratings(:,2),group,'br','xo')
See Also “Grouped Data” on page 2-34
gplotmatrix, grpstats, scatter
18-538
qrandstream.gt
Purpose Greater than relation for handles
Syntax h1 > h2
Description h1 > h2 performs element-wise comparisons between handle arrays
h1 and h2. h1 and h2 must be of the same dimensions unless one is a
scalar. The result is a logical array of the same dimensions, where each
element is an element-wise > result.
If one of h1 or h2 is scalar, scalar expansion is performed and the result
will match the dimensions of the array that is not scalar.
tf = gt(h1, h2) stores the result in a logical array of the same
dimensions.
See Also qrandstream, eq, ge, le, lt, ne
18-539
haltonset class
Superclasses qrandset
Purpose Halton quasi-random point sets
Description haltonset is a quasi-random point set class that produces points from
the Halton sequence.
Construction haltonset Construct Halton quasi-random
point set
Methods Inherited Methods
Methods in the following table are inherited from qrandset.
disp Display qrandset object
end Last index in indexing expression
for point set
length Length of point set
ndims Number of dimensions in matrix
net Generate quasi-random point set
scramble Scramble quasi-random point set
size Number of dimensions in matrix
subsref Subscripted reference for
qrandset
Properties Inherited Properties
Properties in the following table are inherited from qrandset.
Dimensions Number of dimensions
Leap Interval between points
18-540
haltonset class
ScrambleMethod Settings that control scrambling
Skip Number of initial points to omit
from sequence
Type Name of sequence on which point
set P is based
Copy Handle. To learn how this affects your use of the class, see
Semantics Comparing Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
References [1] Kocis, L., and W. J. Whiten, "Computational Investigations of
Low-Discrepancy Sequences," ACM Transactions on Mathematical
Software, Vol. 23, No. 2, pp. 266-294, 1997.
See Also sobolset
How To • “Quasi-Random Point Sets” on page 6-18
18-541
haltonset
Purpose Construct Halton quasi-random point set
Syntax p = haltonset(d)
p = haltonset(d,prop1,val1,prop2,val2,...)
Description p = haltonset(d) constructs a d-dimensional point set p of the
haltonset class, with default property settings.
p = haltonset(d,prop1,val1,prop2,val2,...) specifies property
name/value pairs used to construct p.
The object p returned by haltonset encapsulates properties of a
specified quasi-random sequence. The point set is finite, with a length
determined by the Skip and Leap properties and by limits on the size
of point set indices (maximum value of 253). Values of the point set are
not generated and stored in memory until you access p using net or
parenthesis indexing.
Examples Generate a 3-D Halton point set, skip the first 1000 values, and then
retain every 101st point:
p = haltonset(3,'Skip',1e3,'Leap',1e2)
p =
Halton point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : none
Use scramble to apply reverse-radix scrambling:
p = scramble(p,'RR2')
p =
Halton point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : RR2
18-542
haltonset
Use net to generate the first four points:
X0 = net(p,4)
X0 =
0.0928 0.6950 0.0029
0.6958 0.2958 0.8269
0.3013 0.6497 0.4141
0.9087 0.7883 0.2166
Use parenthesis indexing to generate every third point, up to the 11th
point:
X = p(1:3:11,:)
X =
0.0928 0.6950 0.0029
0.9087 0.7883 0.2166
0.3843 0.9840 0.9878
0.6831 0.7357 0.7923
References [1] Kocis, L., and W. J. Whiten. “Computational Investigations of
Low-Discrepancy Sequences.” ACM Transactions on Mathematical
Software. Vol. 23, No. 2, 1997, pp. 266–294.
See Also net, scramble, sobolset
18-543
harmmean
Purpose Harmonic mean
Syntax m = harmmean(X)
harmmean(X,dim)
Description m = harmmean(X) calculates the harmonic mean of a sample. For
vectors, harmmean(x) is the harmonic mean of the elements in x. For
matrices, harmmean(X) is a row vector containing the harmonic means
of each column. For N-dimensional arrays, harmmean operates along the
first nonsingleton dimension of X.
harmmean(X,dim) takes the harmonic mean along dimension dim of X.
The harmonic mean is
n
m=
n
1
∑ xi
i=1
Examples The arithmetic mean is greater than or equal to the harmonic mean.
x = exprnd(1,10,6);
harmonic = harmmean(x)
harmonic =
0.3382 0.3200 0.3710 0.0540 0.4936 0.0907
average = mean(x)
average =
1.3509 1.1583 0.9741 0.5319 1.0088 0.8122
See Also mean, median, geomean, trimmean
18-544
hist3
Purpose Bivariate histogram
Syntax hist3(X)
hist3(X,nbins)
hist3(X,ctrs)
hist3(X,'Edges',edges)
N = hist3(X,...)
[N,C] = hist3(X,...)
hist3(...,param1,val1,param2,val2,...)
Description hist3(X) bins the elements of the m-by-2 matrix X into a 10-by-10 grid
of equally spaced containers, and plots a histogram. Each column of X
corresponds to one dimension in the bin grid.
hist3(X,nbins) plots a histogram using an nbins(1)-by-nbins(2) grid
of bins. hist3(X,'Nbins',nbins) is equivalent to hist3(X,nbins).
hist3(X,ctrs), where ctrs is a two-element cell array of numeric
vectors with monotonically non-decreasing values, uses a 2-D grid
of bins centered on ctrs{1} in the first dimension and on ctrs{2}
in the second. hist3 assigns rows of X falling outside the range of
that grid to the bins along the outer edges of the grid, and ignores
rows of X containing NaNs. hist3(X,'Ctrs',ctrs) is equivalent to
hist3(X,ctrs).
hist3(X,'Edges',edges), where edges is a two-element cell array
of numeric vectors with monotonically non-decreasing values, uses a
2-D grid of bins with edges at edges{1} in the first dimension and at
edges{2} in the second. The (i, j)th bin includes the value X(k,:) if
edges{1}(i) <= X(k,1) < edges{1}(i+1)
edges{2}(j) <= X(k,2) < edges{2}(j+1)
Rows of X that fall on the upper edges of the grid, edges{1}(end) or
edges{2}(end), are counted in the (I,j)th or (i,J)th bins, where I
and J are the lengths of edges{1} and edges{2}. hist3 does not count
rows of X falling outside the range of the grid. Use -Inf and Inf in
edges to include all non-NaN values.
18-545
hist3
N = hist3(X,...) returns a matrix containing the number of elements
of X that fall in each bin of the grid, and does not plot the histogram.
[N,C] = hist3(X,...) returns the positions of the bin centers in a
1-by-2 cell array of numeric vectors, and does not plot the histogram.
hist3(ax,X,...) plots onto an axes with handle ax instead of the
current axes. See the axes reference page for more information about
handles to plots.
hist3(...,param1,val1,param2,val2,...) allows you to specify
graphics parameter name/value pairs to fine-tune the plot.
Examples Example 1
Make a 3-D figure using a histogram with a density plot underneath:
load seamount
dat = [-y,x]; % Grid corrected for negative y-values
hold on
hist3(dat) % Draw histogram in 2D
n = hist3(dat); % Extract histogram data;
% default to 10x10 bins
n1 = n';
n1( size(n,1) + 1 ,size(n,2) + 1 ) = 0;
Generate grid for 2-D projected view of intensities:
xb = linspace(min(dat(:,1)),max(dat(:,1)),size(n,1)+1);
yb = linspace(min(dat(:,2)),max(dat(:,2)),size(n,1)+1);
Make a pseudocolor plot:
h = pcolor(xb,yb,n1);
Set the z-level and colormap of the displayed grid:
set(h, 'zdata', ones(size(n1)) * -max(max(n)))
colormap(hot) % heat map
title('Seamount: ...
18-546
hist3
Data Point Density Histogram and Intensity Map');
grid on
Display the default 3-D perspective view:
view(3);
Example 2
Use the car data to make a histogram on a 7-by-7 grid of bins.
18-547
hist3
load carbig
X = [MPG,Weight];
hist3(X,[7 7]);
xlabel('MPG'); ylabel('Weight');
Make a histogram with semi-transparent bars:
hist3(X,[7 7],'FaceAlpha',.65);
xlabel('MPG'); ylabel('Weight');
set(gcf,'renderer','opengl');
18-548
hist3
Specify bin centers, different in each direction; get back counts, but
don’t make the plot.
cnt = hist3(X, {0:10:50 2000:500:5000});
Example 3
Make a histogram with bars colored according to height.
load carbig
X = [MPG,Weight];
hist3(X,[7 7]);
xlabel('MPG'); ylabel('Weight');
set(gcf,'renderer','opengl');
set(get(gca,'child'),'FaceColor','interp','CDataMode',...
'auto');
18-549
hist3
See Also accumarray, bar, bar3, hist, histc
18-550
categorical.hist
Purpose Plot histogram of categorical data
Syntax hist(Y)
hist(Y,X)
hist(ax,...)
N = hist(...)
[N,X] = hist(...)
Description hist(Y) plots a histogram bar plot of the counts for each level of
the categorical vector Y. If Y is an m-by-n categorical matrix, hist
computes counts for each column of Y, and plots a group of n bars for
each categorical level.
hist(Y,X) plots bars only for the levels specified in X. X is a categorical
vector or a cell array of level names as strings.
hist(ax,...) plots into the axes with handle ax instead of gca.
N = hist(...) returns the counts for each categorical level. If Y is
a matrix, hist works down the columns of Y and returns a matrix of
counts with one column for each coluimn of Y and one row for each
cetegorical level.
[N,X] = hist(...) returns the categorical levels corresponding to
each count in N, or corresponding to each column of N if Y is a matrix.
Examples Create a histogram of age groups from the hospital.mat dataset:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
AgeGroup = ordinal(hospital.Age,labels,[],edges);
AgeGroup = droplevels(AgeGroup);
hist(AgeGroup)
18-551
categorical.hist
See Also hist | categorical.levelcounts | categorical.getlevels
18-552
histfit
Purpose Histogram with normal fit
Syntax histfit(data)
histfit(data,nbins)
histfit(data,nbins,dist)
h = histfit(...)
Description histfit(data) plots a histogram of the values in the vector data using
the number of bins equal to the square root of the number of elements
in data, then superimposes a fitted normal distribution.
histfit(data,nbins) uses nbins bins for the histogram.
histfit(data,nbins,dist) plots a histogram with a density from the
distribution specified by dist, one of the following strings:
• 'beta'
• 'birnbaumsaunders'
• 'exponential'
• 'extreme value' or ev'
• 'gamma'
• 'generalized extreme value' or 'gev'
• 'generalized pareto' or 'gp'
• 'inversegaussian'
• 'logistic'
• 'loglogistic'
• 'lognormal'
• 'nakagami'
• 'negative binomial' or 'nbin'
• 'normal' (default)
• 'poisson'
18-553
histfit
• 'rayleigh'
• 'rician'
• 'tlocationscale'
• 'weibull' or 'wbl'
h = histfit(...) returns a vector of handles h, where h(1) is the
handle to the histogram and h(2) is the handle to the normal curve.
Examples r = normrnd(10,1,100,1);
histfit(r)
h = get(gca,'Children');
set(h(2),'FaceColor',[.8 .8 1])
18-554
histfit
See Also hist, normfit
18-555
hmmdecode
Purpose Hidden Markov model posterior state probabilities
Syntax PSTATES = hmmdecode(seq,TRANS,EMIS)
[PSTATES,logpseq] = hmmdecode(...)
[PSTATES,logpseq,FORWARD,BACKWARD,S] = hmmdecode(...)
hmmdecode(...,'Symbols',SYMBOLS)
Description PSTATES = hmmdecode(seq,TRANS,EMIS) calculates the posterior state
probabilities, PSTATES, of the sequence seq, from a hidden Markov
model. The posterior state probabilities are the conditional probabilities
of being at state k at step i, given the observed sequence of symbols, sym.
You specify the model by a transition probability matrix, TRANS, and
an emissions probability matrix, EMIS. TRANS(i,j) is the probability of
transition from state i to state j. EMIS(k,seq) is the probability that
symbol seq is emitted from state k.
PSTATES is an array with the same length as seq and one row for each
state in the model. The (i, j)th element of PSTATES gives the probability
that the model is in state i at the jth step, given the sequence seq.
Note The function hmmdecode begins with the model in state 1 at step
0, prior to the first emission. hmmdecode computes the probabilities in
PSTATES based on the fact that the model begins in state 1.
[PSTATES,logpseq] = hmmdecode(...) returns logpseq, the
logarithm of the probability of sequence seq, given transition matrix
TRANS and emission matrix EMIS.
[PSTATES,logpseq,FORWARD,BACKWARD,S] = hmmdecode(...) returns
the forward and backward probabilities of the sequence scaled by S.
hmmdecode(...,'Symbols',SYMBOLS) specifies the symbols that are
emitted. SYMBOLS can be a numeric array or a cell array of the names of
the symbols. The default symbols are integers 1 through N, where N is
the number of possible emissions.
18-556
hmmdecode
References [1] Durbin, R., S. Eddy, A. Krogh, and G. Mitchison. Biological
Sequence Analysis. Cambridge, UK: Cambridge University Press, 1998.
Examples trans = [0.95,0.05;
0.10,0.90];
emis = [1/6 1/6 1/6 1/6 1/6 1/6;
1/10 1/10 1/10 1/10 1/10 1/2];
[seq,states] = hmmgenerate(100,trans,emis);
pStates = hmmdecode(seq,trans,emis);
[seq,states] = hmmgenerate(100,trans,emis,...
'Symbols',{'one','two','three','four','five','six'})
pStates = hmmdecode(seq,trans,emis,...
'Symbols',{'one','two','three','four','five','six'});
See Also hmmgenerate, hmmestimate, hmmviterbi, hmmtrain
18-557
hmmestimate
Purpose Hidden Markov model parameter estimates from emissions and states
Syntax [TRANS,EMIS] = hmmestimate(seq,states)
hmmestimate(...,'Symbols',SYMBOLS)
hmmestimate(...,'Statenames',STATENAMES)
hmmestimate(...,'Pseudoemissions',PSEUDOE)
hmmestimate(...,'Pseudotransitions',PSEUDOTR)
Description [TRANS,EMIS] = hmmestimate(seq,states) calculates the maximum
likelihood estimate of the transition, TRANS, and emission, EMIS,
probabilities of a hidden Markov model for sequence, seq, with known
states, states.
hmmestimate(...,'Symbols',SYMBOLS) specifies the symbols that are
emitted. SYMBOLS can be a numeric array or a cell array of the names of
the symbols. The default symbols are integers 1 through N, where N is
the number of possible emissions.
hmmestimate(...,'Statenames',STATENAMES) specifies the names of
the states. STATENAMES can be a numeric array or a cell array of the
names of the states. The default state names are 1 through M, where
M is the number of states.
hmmestimate(...,'Pseudoemissions',PSEUDOE) specifies
pseudocount emission values in the matrix PSEUDO. Use this argument
to avoid zero probability estimates for emissions with very low
probability that might not be represented in the sample sequence.
PSEUDOE should be a matrix of size m-by-n, where m is the number of
states in the hidden Markov model and n is the number of possible
emissions. If the i → k emission does not occur in seq, you can set
PSEUDOE(i,k) to be a positive number representing an estimate of the
expected number of such emissions in the sequence seq.
hmmestimate(...,'Pseudotransitions',PSEUDOTR) specifies
pseudocount transition values. You can use this argument to avoid
zero probability estimates for transitions with very low probability that
might not be represented in the sample sequence. PSEUDOTR should be a
matrix of size m-by-m, where m is the number of states in the hidden
18-558
hmmestimate
Markov model. If the i → j transition does not occur in states, you can
set PSEUDOTR(i,j) to be a positive number representing an estimate of
the expected number of such transitions in the sequence states.
Pseudotransitions and Pseudoemissions
If the probability of a specific transition or emission is very low, the
transition might never occur in the sequence states, or the emission
might never occur in the sequence seq. In either case, the algorithm
returns a probability of 0 for the given transition or emission in TRANS
or EMIS. You can compensate for the absence of transition with the
'Pseudotransitions' and 'Pseudoemissions' arguments. The
simplest way to do this is to set the corresponding entry of PSEUDO or
PSEUDOTR to 1. For example, if the transition i → j does not occur in
states, set PSEUOTR(i,j) = 1. This forces TRANS(i,j) to be positive.
If you have an estimate for the expected number of transitions i → j
in a sequence of the same length as states, and the actual number
of transitions i → j that occur in seq is substantially less than what
you expect, you can set PSEUOTR(i,j) to the expected number. This
increases the value of TRANS(i,j). For transitions that do occur in
states with the frequency you expect, set the corresponding entry of
PSEUDOTR to 0, which does not increase the corresponding entry of TRANS.
If you do not know the sequence of states, use hmmtrain to estimate the
model parameters.
References [1] Durbin, R., S. Eddy, A. Krogh, and G. Mitchison. Biological
Sequence Analysis. Cambridge, UK: Cambridge University Press, 1998.
Examples trans = [0.95,0.05; 0.10,0.90];
emis = [1/6 1/6 1/6 1/6 1/6 1/6;
1/10 1/10 1/10 1/10 1/10 1/2];
[seq,states] = hmmgenerate(1000,trans,emis);
[estimateTR,estimateE] = hmmestimate(seq,states);
18-559
hmmestimate
See Also hmmgenerate, hmmdecode, hmmviterbi, hmmtrain
18-560
hmmgenerate
Purpose Hidden Markov model states and emissions
Syntax [seq,states] = hmmgenerate(len,TRANS,EMIS)
hmmgenerate(...,'Symbols',SYMBOLS)
hmmgenerate(...,'Statenames',STATENAMES)
Description [seq,states] = hmmgenerate(len,TRANS,EMIS) takes a known
Markov model, specified by transition probability matrix TRANS and
emission probability matrix EMIS, and uses it to generate
• A random sequence seq of emission symbols
• A random sequence states of states
The length of both seq and states is len. TRANS(i,j) is the probability
of transition from state i to state j. EMIS(k,l) is the probability that
symbol l is emitted from state k.
Note The function hmmgenerate begins with the model in state 1 at
step 0, prior to the first emission. The model then makes a transition
to state i1, with probability T1i1, and generates an emission ak1 with
probability Ei1k1 . hmmgenerate returns i1 as the first entry of states,
1
and ak1 as the first entry of seq.
hmmgenerate(...,'Symbols',SYMBOLS) specifies the symbols that are
emitted. SYMBOLS can be a numeric array or a cell array of the names of
the symbols. The default symbols are integers 1 through N, where N is
the number of possible emissions.
hmmgenerate(...,'Statenames',STATENAMES) specifies the names of
the states. STATENAMES can be a numeric array or a cell array of the
names of the states. The default state names are 1 through M, where
M is the number of states.
Since the model always begins at state 1, whose transition probabilities
are in the first row of TRANS, in the following example, the first entry of
18-561
hmmgenerate
the output states is be 1 with probability 0.95 and 2 with probability
0.05.
Examples trans = [0.95,0.05;
0.10,0.90];
emis = [1/6 1/6 1/6 1/6 1/6 1/6;
1/10 1/10 1/10 1/10 1/10 1/2];
[seq,states] = hmmgenerate(100,trans,emis)
[seq,states] = hmmgenerate(100,trans,emis,...
'Symbols',{'one','two','three','four','five','six'},...
'Statenames',{'fair';'loaded'})
See Also hmmviterbi, hmmdecode, hmmestimate, hmmtrain
18-562
hmmtrain
Purpose Hidden Markov model parameter estimates from emissions
Syntax [ESTTR,ESTEMIT] = hmmtrain(seq,TRGUESS,EMITGUESS)
hmmtrain(...,'Algorithm',algorithm)
hmmtrain(...,'Symbols',SYMBOLS)
hmmtrain(...,'Tolerance',tol)
hmmtrain(...,'Maxiterations',maxiter)
hmmtrain(...,'Verbose',true)
hmmtrain(...,'Pseudoemissions',PSEUDOE)
hmmtrain(...,'Pseudotransitions',PSEUDOTR)
Description [ESTTR,ESTEMIT] = hmmtrain(seq,TRGUESS,EMITGUESS) estimates
the transition and emission probabilities for a hidden Markov model
using the Baum-Welch algorithm. seq can be a row vector containing
a single sequence, a matrix with one row per sequence, or a cell array
with each cell containing a sequence. TRGUESS and EMITGUESS are
initial estimates of the transition and emission probability matrices.
TRGUESS(i,j) is the estimated probability of transition from state i to
state j. EMITGUESS(i,k) is the estimated probability that symbol k is
emitted from state i.
hmmtrain(...,'Algorithm',algorithm) specifies the training
algorithm. algorithm can be either 'BaumWelch' or 'Viterbi'. The
default algorithm is 'BaumWelch'.
hmmtrain(...,'Symbols',SYMBOLS) specifies the symbols that are
emitted. SYMBOLS can be a numeric array or a cell array of the names of
the symbols. The default symbols are integers 1 through N, where N is
the number of possible emissions.
hmmtrain(...,'Tolerance',tol) specifies the tolerance used for
testing convergence of the iterative estimation process. The default
tolerance is 1e-4.
hmmtrain(...,'Maxiterations',maxiter) specifies the maximum
number of iterations for the estimation process. The default maximum
is 100.
18-563
hmmtrain
hmmtrain(...,'Verbose',true) returns the status of the algorithm at
each iteration.
hmmtrain(...,'Pseudoemissions',PSEUDOE) specifies pseudocount
emission values for the Viterbi training algorithm. Use this argument to
avoid zero probability estimates for emissions with very low probability
that might not be represented in the sample sequence. PSEUDOE should
be a matrix of size m-by-n, where m is the number of states in the
hidden Markov model and n is the number of possible emissions. If the
i→k emission does not occur in seq, you can set PSEUDOE(i,k) to be a
positive number representing an estimate of the expected number of
such emissions in the sequence seq.
hmmtrain(...,'Pseudotransitions',PSEUDOTR) specifies
pseudocount transition values for the Viterbi training algorithm. Use
this argument to avoid zero probability estimates for transitions with
very low probability that might not be represented in the sample
sequence. PSEUDOTR should be a matrix of size m-by-m, where m is the
number of states in the hidden Markov model. If the i→j transition does
not occur in states, you can set PSEUDOTR(i,j) to be a positive number
representing an estimate of the expected number of such transitions in
the sequence states.
If you know the states corresponding to the sequences, use hmmestimate
to estimate the model parameters.
Tolerance
The input argument ’tolerance' controls how many steps the
hmmtrain algorithm executes before the function returns an answer.
The algorithm terminates when all of the following three quantities are
less than the value that you specify for tolerance:
• The log likelihood that the input sequence seq is generated by the
currently estimated values of the transition and emission matrices
• The change in the norm of the transition matrix, normalized by the
size of the matrix
18-564
hmmtrain
• The change in the norm of the emission matrix, normalized by the
size of the matrix
The default value of 'tolerance' is .0001. Increasing the tolerance
decreases the number of steps the hmmtrain algorithm executes before
it terminates.
maxiterations
The maximum number of iterations, 'maxiterations', controls the
maximum number of steps the algorithm executes before it terminates.
If the algorithm executes maxiter iterations before reaching the
specified tolerance, the algorithm terminates and the function returns a
warning. If this occurs, you can increase the value of 'maxiterations'
to make the algorithm reach the desired tolerance before terminating.
References [1] Durbin, R., S. Eddy, A. Krogh, and G. Mitchison. Biological
Sequence Analysis. Cambridge, UK: Cambridge University Press, 1998.
Examples trans = [0.95,0.05;
0.10,0.90];
emis = [1/6, 1/6, 1/6, 1/6, 1/6, 1/6;
1/10, 1/10, 1/10, 1/10, 1/10, 1/2];
seq1 = hmmgenerate(100,trans,emis);
seq2 = hmmgenerate(200,trans,emis);
seqs = {seq1,seq2};
[estTR,estE] = hmmtrain(seqs,trans,emis);
See Also hmmgenerate, hmmdecode, hmmestimate, hmmviterbi
18-565
hmmviterbi
Purpose Hidden Markov model most probable state path
Syntax STATES = hmmviterbi(seq,TRANS,EMIS)
hmmviterbi(...,'Symbols',SYMBOLS)
hmmviterbi(...,'Statenames',STATENAMES)
Description STATES = hmmviterbi(seq,TRANS,EMIS) given a sequence, seq,
calculates the most likely path through the hidden Markov model
specified by transition probability matrix, TRANS, and emission
probability matrix EMIS. TRANS(i,j) is the probability of transition
from state i to state j. EMIS(i,k) is the probability that symbol k is
emitted from state i.
Note The function hmmviterbi begins with the model in state 1 at step
0, prior to the first emission. hmmviterbi computes the most likely path
based on the fact that the model begins in state 1.
hmmviterbi(...,'Symbols',SYMBOLS) specifies the symbols that are
emitted. SYMBOLS can be a numeric array or a cell array of the names of
the symbols. The default symbols are integers 1 through N, where N is
the number of possible emissions.
hmmviterbi(...,'Statenames',STATENAMES) specifies the names of
the states. STATENAMES can be a numeric array or a cell array of the
names of the states. The default state names are 1 through M, where
M is the number of states.
Examples trans = [0.95,0.05;
0.10,0.90];
emis = [1/6 1/6 1/6 1/6 1/6 1/6;
1/10 1/10 1/10 1/10 1/10 1/2];
[seq,states] = hmmgenerate(100,trans,emis);
estimatedStates = hmmviterbi(seq,trans,emis);
18-566
hmmviterbi
[seq,states] = ...
hmmgenerate(100,trans,emis,...
'Statenames',{'fair';'loaded'});
estimatesStates = ...
hmmviterbi(seq,trans,emis,...
'Statenames',{'fair';'loaded'});
References [1] Durbin, R., S. Eddy, A. Krogh, and G. Mitchison. Biological
Sequence Analysis. Cambridge, UK: Cambridge University Press, 1998.
See Also hmmgenerate, hmmdecode, hmmestimate, hmmtrain
18-567
categorical.horzcat
Purpose Horizontal concatenation for categorical arrays
Syntax C = horzcat(dim,A,B,...)
C = horzcat(A,B)
Description C = horzcat(dim,A,B,...) horizontally concatenates the categorical
arrays A,B,... . For matrices, all inputs must have the same number
of rows. For n-D arrays, all inputs must have the same sizes except in
the second dimension. The set of categorical levels for C is the sorted
union of the sets of levels of the inputs, as determined by their labels.
C = horzcat(A,B) is called for the syntax [A B].
See Also cat, vertcat
18-568
dataset.horzcat
Purpose Horizontal concatenation for dataset arrays
Syntax ds = horzcat(ds1, ds2, ...)
Description ds = horzcat(ds1, ds2, ...) horizontally concatenates the dataset
arrays ds1, ds2, ... . You may concatenate dataset arrays that
have duplicate variable names, however, the variables must contain
identical data, and horzcat includes only one copy of the variable in
the output dataset.
Observation names for all dataset arrays that have them must
be identical except for order. horzcat concatenates by matching
observation names when present, or by position for datasets that do not
have observation names.
See Also cat, vertcat
18-569
hougen
Purpose Hougen-Watson model
Syntax yhat = hougen(beta,x)
Description yhat = hougen(beta,x) returns the predicted values of the reaction
rate, yhat, as a function of the vector of parameters, beta, and the
matrix of data, X. beta must have 5 elements and X must have three
columns.
hougen is a utility function for rsmdemo.
The model form is:
1 x2 − x3 / 5
yˆ =
1 + 2 x1 + 3 x2 + 4 x3
References [1] Bates, D. M., and D. G. Watts. Nonlinear Regression Analysis and
Its Applications. Hoboken, NJ: John Wiley & Sons, Inc., 1988.
See Also rsmdemo
18-570
hygecdf
Purpose Hypergeometric cumulative distribution function
Syntax hygecdf(X,M,K,N)
Description hygecdf(X,M,K,N) computes the hypergeometric cdf at each of the
values in X using the corresponding size of the population, M, number of
items with the desired characteristic in the population, K, and number
of samples drawn, N. Vector or matrix inputs for X, M, K, and N must all
have the same size. A scalar input is expanded to a constant matrix
with the same dimensions as the other inputs.
The hypergeometric cdf is
⎛ K ⎞⎛ M − K ⎞
x ⎜ ⎟⎜ ⎟
i N −i ⎠
p = F( x| M, K , N ) = ∑ ⎝ ⎠ ⎝
i =0 ⎛M⎞
⎜ ⎟
⎝N⎠
The result, p, is the probability of drawing up to x of a possible K items
in N drawings without replacement from a group of M objects.
Examples Suppose you have a lot of 100 floppy disks and you know that 20 of them
are defective. What is the probability of drawing zero to two defective
floppies if you select 10 at random?
p = hygecdf(2,100,20,10)
p =
0.6812
See Also cdf, hygepdf, hygeinv, hygestat, hygernd
18-571
hygeinv
Purpose Hypergeometric inverse cumulative distribution function
Syntax hygeinv(P,M,K,N)
Description hygeinv(P,M,K,N) returns the smallest integer X such that the
hypergeometric cdf evaluated at X equals or exceeds P. You can think of
P as the probability of observing X defective items in N drawings without
replacement from a group of M items where K are defective.
Examples Suppose you are the Quality Assurance manager for a floppy disk
manufacturer. The production line turns out floppy disks in batches of
1,000. You want to sample 50 disks from each batch to see if they have
defects. You want to accept 99% of the batches if there are no more
than 10 defective disks in the batch. What is the maximum number of
defective disks should you allow in your sample of 50?
x = hygeinv(0.99,1000,10,50)
x =
3
What is the median number of defective floppy disks in samples of 50
disks from batches with 10 defective disks?
x = hygeinv(0.50,1000,10,50)
x =
0
See Also icdf, hygecdf, hygepdf, hygestat, hygernd
18-572
hygepdf
Purpose Hypergeometric probability density function
Syntax Y = hygepdf(X,M,K,N)
Description Y = hygepdf(X,M,K,N) computes the hypergeometric pdf at each
of the values in X using the corresponding size of the population, M,
number of items with the desired characteristic in the population, K,
and number of samples drawn, N. X, M, K, and N can be vectors, matrices,
or multidimensional arrays that all have the same size. A scalar input
is expanded to a constant array with the same dimensions as the other
inputs.
The parameters in M, K, and N must all be positive integers, with N ≤ M.
The values in X must be less than or equal to all the parameter values.
The hypergeometric pdf is
⎛ K ⎞⎛ M − K ⎞
⎜ ⎟⎜ ⎟
x N−x⎠
y = f ( x| M, K , N ) = ⎝ ⎠ ⎝
⎛M⎞
⎜ ⎟
⎝N⎠
The result, y, is the probability of drawing exactly x of a possible K
items in n drawings without replacement from a group of M objects.
Examples Suppose you have a lot of 100 floppy disks and you know that 20 of them
are defective. What is the probability of drawing 0 through 5 defective
floppy disks if you select 10 at random?
p = hygepdf(0:5,100,20,10)
p =
0.0951 0.2679 0.3182 0.2092 0.0841 0.0215
See Also pdf, hygecdf, hygeinv, hygestat, hygernd
18-573
hygernd
Purpose Hypergeometric random numbers
Syntax R = hygernd(M,K,N)
R = hygernd(M,K,N,v)
R = hygernd(M,K,N,m,n)
Description R = hygernd(M,K,N) generates random numbers from the
hypergeometric distribution with corresponding size of the population,
M, number of items with the desired characteristic in the population, K,
and number of samples drawn, N. M, K, and N can be vectors, matrices, or
multidimensional arrays that all have the same size, which is also the
size of R. A scalar input for M, K, or N is expanded to a constant array
with the same dimensions as the other inputs.
R = hygernd(M,K,N,v) generates random numbers from the
hypergeometric distribution with parameters M, K, and N, where v is a
row vector. If v is a 1-by-2 vector, R is a matrix with v(1) rows and v(2)
columns. If v is 1-by-n, R is an n-dimensional array.
R = hygernd(M,K,N,m,n) generates random numbers from the
hypergeometric distribution with parameters M, K, and N, where scalars
m and n are the row and column dimensions of R.
Examples numbers = hygernd(1000,40,50)
numbers =
1
See Also random, hygepdf, hygecdf, hygeinv, hygestat
18-574
hygestat
Purpose Hypergeometric mean and variance
Syntax [MN,V] = hygestat(M,K,N)
Description [MN,V] = hygestat(M,K,N) returns the mean of and variance for the
hypergeometric distribution with corresponding size of the population,
M, number of items with the desired characteristic in the population, K,
and number of samples drawn, N. Vector or matrix inputs for M, K, and
N must have the same size, which is also the size of MN and V. A scalar
input for M, K, or N is expanded to a constant matrix with the same
dimensions as the other inputs.
The mean of the hypergeometric distribution with parameters M, K, and
N is NK/M, and the variance is NK(M-K)(M-N)/[M^2(M-1)].
Examples The hypergeometric distribution approaches the binomial distribution,
where p = K/M, as M goes to infinity.
[m,v] = hygestat(10.^(1:4),10.^(0:3),9)
m =
0.9000 0.9000 0.9000 0.9000
v =
0.0900 0.7445 0.8035 0.8094
[m,v] = binostat(9,0.1)
m =
0.9000
v =
0.8100
See Also hygepdf, hygecdf, hygeinv, hygernd
18-575
icdf
Purpose Inverse cumulative distribution functions
Syntax Y = icdf(name,X,A)
Y = icdf(name,X,A,B)
Y = icdf(name,X,A,B,C)
Description Y = icdf(name,X,A) computes the inverse cumulative distribution
function for the one-parameter family of distributions specified by
name. Parameter values for the distribution are given in A. The inverse
cumulative distribution function is evaluated at the values in X and its
values are returned in Y.
If X and A are arrays, they must be the same size. If X is a scalar, it is
expanded to a constant matrix the same size as A. If A is a scalar, it is
expanded to a constant matrix the same size as X.
Y is the common size of X and A after any necessary scalar expansion.
Y = icdf(name,X,A,B) computes the inverse cumulative distribution
function for two-parameter families of distributions, where parameter
values are given in A and B.
If X, A, and B are arrays, they must be the same size. If X is a scalar, it is
expanded to a constant matrix the same size as A and B. If either A or B
are scalars, they are expanded to constant matrices the same size as X.
Y is the common size of X, A, and B after any necessary scalar expansion.
Y = icdf(name,X,A,B,C) computes the inverse cumulative distribution
function for three-parameter families of distributions, where parameter
values are given in A, B, and C.
If X, A, B, and C are arrays, they must be the same size. If X is a scalar,
it is expanded to a constant matrix the same size as A, B, and C. If
any of A, B or C are scalars, they are expanded to constant matrices
the same size as X.
Y is the common size of X, A, B and C after any necessary scalar
expansion.
Acceptable strings for name are:
18-576
icdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'beta' or “Beta a b —
'Beta' Distribution”
on page B-4
'bino' or “Binomial n: number p: —
'Binomial' Distribution” of trials probability
on page B-7 of success
for each
trial
'chi2' or “Chi-Square ν: degrees — —
'Chisquare' Distribution” of freedom
on page
B-12
'exp' or “Exponential : mean — —
Distribution”
'Exponential'
on page
B-16
'ev' or “Extreme : location σ: scale —
'Extreme Value parameter parameter
Value' Distribution”
on page
B-19
'f' or 'F' “F ν 1: ν 2: —
Distribution” numerator denominator
'gam' or on page
“Gamma degrees
a : shapeof degrees
b : scale of —
'Gamma' B-25
Distribution” freedom
parameter freedom
parameter
on page
B-27
18-577
icdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'gev' or “Generalized K: shape : location σ: scale
'GeneralizedExtreme parameter parameter parameter
Extreme Value
Value' Distribution”
on page
B-32
'gp' or “Generalized k: tail index σ: scale : threshold
'GeneralizedPareto (shape) parameter (location)
Pareto' Distribution” parameter parameter
on page
B-37
'geo' or “Geometric p: — —
'Geometric' Distribution” probability
'hyge' or on page parameter
“Hypergeometric
M : size of the K: number n: number
B-41
Distribution”
'Hypergeometric' population of items of samples
on page with the drawn
B-43 desired
characteristic
in the
population
'logn' or “Lognormal σ —
'Lognormal' Distribution”
on page
B-51
'nbin' or “Negative r: number p: —
'Negative Binomial of successes probability
Binomial' Distribution” of success
on page in a single
B-72 trial
18-578
icdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'ncf' or “Noncentral ν1: ν 2: δ:
'Noncentral F numerator denominator noncentrality
F' Distribution” degrees of degrees of parameter
on page freedom freedom
B-78
'nct' or “Noncentral ν: degrees δ: —
'Noncentral t of freedom noncentrality
t' Distribution” parameter
'ncx2' or on page
“Noncentral ν: degrees δ: —
'Noncentral B-80
Chi-Square of freedom noncentrality
Chi-square' Distribution” parameter
on page
B-76
'norm' or “Normal : mean σ: standard —
'Normal' Distribution” deviation
on page
B-83
'poiss' or “Poisson λ: mean — —
'Poisson' Distribution”
on page
B-89
'rayl' or “Rayleigh b: scale — —
'Rayleigh' Distribution” parameter
on page
B-91
't' or 'T' “Student’s t ν: degrees — —
Distribution” of freedom
on page
B-95
18-579
icdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'unif' or “Uniform a: lower b: upper —
'Uniform' Distribution endpoint endpoint
(Continuous)” (minimum) (maximum)
on page
B-99
'unid' or “Uniform N: — —
'Discrete Distribution maximum
Uniform' (Discrete)” observable
on page value
B-101
'wbl' or “Weibull a: scale b: shape —
'Weibull' Distribution” parameter parameter
on page
B-103
Examples Compute the icdf of the normal distribution with mean 0 and standard
deviation 1 at inputs 0.1, 0.3, ..., 0.9:
x1 = icdf('Normal',0.1:0.2:0.9,0,1)
x1 =
-1.2816 -0.5244 0 0.5244 1.2816
The order of the parameters is the same as for norminv.
Compute the icdfs of Poisson distributions with rate parameters 0, 1, ...,
4 at inputs 0.1, 0.3, ..., 0.9, respectively:
x2 = icdf('Poisson',0.1:0.2:0.9,0:4)
x2 =
NaN 0 2 4 7
The order of the parameters is the same as for poissinv.
18-580
icdf
See Also cdf, mle, pdf, random
18-581
ProbDistUnivKernel.icdf
Purpose Return inverse cumulative distribution function (ICDF) for
ProbDistUnivKernel object
Syntax Y = icdf(PD, P)
Description Y = icdf(PD, P) returns Y, an array containing the inverse cumulative
distribution function (ICDF) for the ProbDistUnivKernel object PD,
evaluated at values in P.
Input PD An object of the class ProbDistUnivKernel.
Arguments
P A numeric array of values from 0 to 1 where
you want to evaluate the ICDF.
Output Y An array containing the inverse cumulative
Arguments distribution function (ICDF) for the
ProbDistUnivKernel object PD.
See Also icdf
18-582
ProbDistUnivParam.icdf
Purpose Return inverse cumulative distribution function (ICDF) for
ProbDistUnivParam object
Syntax Y = icdf(PD, P)
Description Y = icdf(PD, P) returns Y, an array containing the inverse cumulative
distribution function (ICDF) for the ProbDistUnivParam object PD,
evaluated at values in P.
Input PD An object of the class ProbDistUnivParam.
Arguments
P A numeric array of values from 0 to 1 where
you want to evaluate the ICDF.
Output Y An array containing the inverse cumulative
Arguments distribution function (ICDF) for the
ProbDistUnivParam object PD.
See Also icdf
18-583
piecewisedistribution.icdf
Purpose Inverse cumulative distribution function for piecewise distribution
Syntax X = icdf(obj,P)
Description X = icdf(obj,P) returns an array X of values of the inverse cumulative
distribution function for the piecewise distribution object obj, evaluated
at the values in the array P.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
[p,q] = boundary(obj)
p =
0.1000
0.9000
q =
-1.7766
1.8432
icdf(obj,p)
ans =
-1.7766
1.8432
See Also paretotails, cdf
18-584
inconsistent
Purpose Inconsistency coefficient
Syntax Y = inconsistent(Z)
Y = inconsistent(Z,d)
Description Y = inconsistent(Z) computes the inconsistency coefficient for each
link of the hierarchical cluster tree Z, where Z is an (m-1)-by-3 matrix
generated by the linkage function. The inconsistency coefficient
characterizes each link in a cluster tree by comparing its height with
the average height of other links at the same level of the hierarchy.
The higher the value of this coefficient, the less similar the objects
connected by the link.
Y = inconsistent(Z,d) computes the inconsistency coefficient for
each link in the hierarchical cluster tree Z to depth d, where d is an
integer denoting the number of levels of the cluster tree that are
included in the calculation. By default, d=2.
The output, Y, is an (m-1)-by-4 matrix formatted as follows.
Column Description
1 Mean of the heights of all the links included in the
calculation.
2 Standard deviation of the heights of all the links included
in the calculation.
3 Number of links included in the calculation.
4 Inconsistency coefficient.
For each link, k, the inconsistency coefficient is calculated as:
Y (k, 4) = ( z(k, 3) − Y (k, 1)) / Y (k, 2)
For leaf nodes, nodes that have no further nodes under them, the
inconsistency coefficient is set to 0.
18-585
inconsistent
Examples X = gallery('uniformdata',[10 2],12);
Y = pdist(X);
Z = linkage(Y,'single');
dendrogram(Z)
W = inconsistent(Z,3)
W =
0.1313 0 1.0000 0
0.1386 0 1.0000 0
0.1463 0.0109 2.0000 0.7071
0.2391 0 1.0000 0
0.1951 0.0568 4.0000 0.9425
0.2308 0.0543 4.0000 0.9320
0.2395 0.0748 4.0000 0.7636
0.2654 0.0945 4.0000 0.9203
0.3769 0.0950 3.0000 1.1040
18-586
inconsistent
References [1] Jain, A., and R. Dubes. Algorithms for Clustering Data. Upper
Saddle River, NJ: Prentice-Hall, 1988.
[2] Zahn, C. T. “Graph-theoretical methods for detecting and describing
Gestalt clusters.” IEEE Transactions on Computers. Vol. C-20, Issue
1, 1971, pp. 68–86.
See Also cluster, cophenet, clusterdata, dendrogram, linkage, pdist,
squareform
18-587
ProbDist.InputData property
Purpose Read-only structure containing information about input data to
ProbDist object
Description InputData is a read-only property of the ProbDist class. InputData
is a structure containing information about input data to a ProbDist
object. It includes the following fields:
• data
• cens
• freq
Values Possible values for the three fields in the structure are any data
supplied to the fitdist function:
• data — Data passed to the fitdist function when creating
the ProbDist object. This field is empty if the ProbDist object
was created without fitting to data, that is by using the
ProbDistUnivParam.ProbDistUnivParam constructor.
• cens — The vector supplied with the 'censoring' parameter when
creating the ProbDist object using the fitdist function. This field is
empty if the ProbDist object was created without fitting to data, that
is by using the ProbDistUnivParam.ProbDistUnivParam constructor.
• freq — The vector supplied with the 'frequency' parameter when
creating the ProbDist object using the fitdist function. This field is
empty if the ProbDist object was created without fitting to data, that
is by using the ProbDistUnivParam.ProbDistUnivParam constructor.
Use this information to view and compare the data supplied to create
distributions.
18-588
categorical.int8
Purpose Convert categorical array to signed 8-bit integer array
Syntax B = int8(A)
Description B = int8(A) converts the categorical array A to a signed 8-bit integer
array. Each element of B contains the internal categorical level code for
the corresponding element of A.
Undefined elements of A are assigned the value 0 in B. If A contains
more than intmax('int8') levels, the internal codes will saturate to
intmax('int8') when cast to int8.
See Also For more information on signed integers, see “Integers” in the MATLAB
documentation.
double, uint8
18-589
categorical.int16
Purpose Convert categorical array to signed 16-bit integer array
Syntax B = int16(A)
Description B = int16(A) converts the categorical array A to a signed 16-bit integer
array. Each element of B contains the internal categorical level code for
the corresponding element of A.
Undefined elements of A are assigned the value 0 in B.
See Also For more information on signed integers, see “Integers” in the MATLAB
documentation.
double, uint16
18-590
categorical.int32
Purpose Convert categorical array to signed 32-bit integer array
Syntax B = int32(A)
Description B = int32(A) converts the categorical array A to a signed 32-bit integer
array. Each element of B contains the internal categorical level code for
the corresponding element of A.
Undefined elements of A are assigned the value 0 in B.
See Also For more information on signed integers, see “Integers” in the MATLAB
documentation.
double, uint32
18-591
categorical.int64
Purpose Convert categorical array to signed 64-bit integer array
Syntax B = int64(A)
Description B = int64(A) converts the categorical array A to a signed 64-bit integer
array. Each element of B contains the internal categorical level code for
the corresponding element of A.
Undefined elements of A are assigned the value 0 in B.
See Also For more information on signed integers, see “Integers” in the MATLAB
documentation.
double, uint64
18-592
interactionplot
Purpose Interaction plot for grouped data
Syntax interactionplot(Y,GROUP)
interactionplot(Y,GROUP,'varnames',VARNAMES)
[h,AX,bigax] = interactionplot(...)
Description interactionplot(Y,GROUP) displays the two-factor interaction plot
for the group means of matrix Y with groups defined by entries in the
cell array GROUP. Y is a numeric matrix or vector. If Y is a matrix,
the rows represent different observations and the columns represent
replications of each observation. If Y is a vector, the rows give the
means of each entry in the cell array GROUP. Each cell of GROUP must
contain a grouping variable that can be a categorical variable, numeric
vector, character matrix, or a single-column cell array of strings. (See
“Grouped Data” on page 2-34.) GROUP can also be a matrix whose
columns represent different grouping variables. Each grouping variable
must have the same number of rows as Y. The number of grouping
variables must be greater than 1.
The interaction plot is a matrix plot, with the number of rows and
columns both equal to the number of grouping variables. The grouping
variable names are printed on the diagonal of the plot matrix. The
plot at off-diagonal position (i,j) is the interaction of the two variables
whose names are given at row diagonal (i,i) and column diagonal (j,j),
respectively.
interactionplot(Y,GROUP,'varnames',VARNAMES) displays the
interaction plot with user-specified grouping variable names VARNAMES.
VARNAMES is a character matrix or a cell array of strings, one per
grouping variable. Default names are 'X1', 'X2', ... .
[h,AX,bigax] = interactionplot(...) returns a handle h to the
figure window, a matrix AX of handles to the subplot axes, and a handle
bigax to the big (invisible) axes framing the subplots.
Examples Display interaction plots for data with four 3-level factors named 'A',
'B','C', and 'D':
18-593
interactionplot
y = randn(1000,1); % response
group = ceil(3*rand(1000,4)); % four 3-level factors
interactionplot(y,group,'varnames',{'A','B','C','D'})
See Also “Grouped Data” on page 2-34
maineffectsplot, multivarichart
18-594
categorical.intersect
Purpose Set intersection for categorical arrays
Syntax C = intersect(A,B)
Description C = intersect(A,B) when A and B are categorical arrays returns a
categorical vector C containing the values common to both A and B. The
result C is sorted. The set of categorical levels for C is the sorted union
of the sets of levels of the inputs, as determined by their labels.
[C,IA,IB] = UNION(A,B) also returns index vectors IA and IB such
that C = A(IA) and C = B(IB).
See Also ismember, setdiff, setxor, union, unique
18-595
invpred
Purpose Inverse prediction
Syntax X0 = invpred(X,Y,Y0)
[X0,DXLO,DXUP] = invpred(X,Y,Y0)
[X0,DXLO,DXUP] = invpred(X,Y,Y0,name1,val1,name2,val2,...)
Description X0 = invpred(X,Y,Y0) accepts vectors X and Y of the same length, fits
a simple regression, and returns the estimated value X0 for which the
height of the line is equal to Y0. The output, X0, has the same size as Y0,
and Y0 can be an array of any size.
[X0,DXLO,DXUP] = invpred(X,Y,Y0) also computes 95% inverse
prediction intervals. DXLO and DXUP define intervals with lower bound
X0 DXLO and upper bound X0+DXUP. Both DXLO and DXUP have the same
size as Y0.
The intervals are not simultaneous and are not necessarily finite. Some
intervals may extend from a finite value to -Inf or +Inf, and some may
extend over the entire real line.
[X0,DXLO,DXUP] = invpred(X,Y,Y0,name1,val1,name2,val2,...)
specifies optional argument name/value pairs chosen from the following
list. Argument names are case insensitive and partial matches are
allowed.
Name Value
'alpha' A value between 0 and 1 specifying a
confidence level of 100*(1-alpha)%. Default
is alpha=0.05 for 95% confidence.
'predopt' Either 'observation', the default value to
compute the intervals for X0 at which a new
observation could equal Y0, or 'curve' to
compute intervals for the X0 value at which
the curve is equal to Y0.
18-596
invpred
Examples x = 4*rand(25,1);
y = 10 + 5*x + randn(size(x));
scatter(x,y)
x0 = invpred(x,y,20)
See Also polyfit, polyval, polyconf, polytool
18-597
categorical.ipermute
Purpose Inverse permute dimensions of categorical array
Syntax A = ipermute(B,order)
Description A = ipermute(B,order) is the inverse of permute. ipermute
rearranges the dimensions of the categorical array B so that
permute(A,order) will produce B. The array produced has the same
values of A but the order of the subscripts needed to access any
particular element are rearranged as specified by order. The elements
of order must be a rearrangement of the numbers from 1 to n.
See Also permute
18-598
iqr
Purpose Interquartile range
Syntax y = iqr(X)
iqr(X,dim)
Description y = iqr(X) returns the interquartile range of the values in X. For vector
input, y is the difference between the 75th and the 25th percentiles
of the sample in X. For matrix input, y is a row vector containing the
interquartile range of each column of X. For N-dimensional arrays, iqr
operates along the first nonsingleton dimension of X.
iqr(X,dim) calculates the interquartile range along the dimension
dim of X.
Remarks The IQR is a robust estimate of the spread of the data, since changes in
the upper and lower 25% of the data do not affect it. If there are outliers
in the data, then the IQR is more representative than the standard
deviation as an estimate of the spread of the body of the data. The IQR
is less efficient than the standard deviation as an estimate of the spread
when the data is all from the normal distribution.
Multiply the IQR by 0.7413 to estimate σ (the second parameter of the
normal distribution.)
Examples This Monte Carlo simulation shows the relative efficiency of the IQR to
the sample standard deviation for normal data.
x = normrnd(0,1,100,100);
s = std(x);
s_IQR = 0.7413*iqr(x);
efficiency = (norm(s-1)./norm(s_IQR-1)).^2
efficiency =
0.3297
See Also std, mad, range
18-599
ProbDistUnivKernel.iqr
Purpose Return interquartile range (IQR) for ProbDistUnivKernel object
Syntax Y = iqr(PD)
Description Y = iqr(PD) returns Y, the interquartile range for the
ProbDistUnivKernel object PD. The interquartile range is the distance
between the 75th and 25th percentiles.
Input PD An object of the class ProbDistUnivKernel.
Arguments
Output Y The value of the interquartile range for the
Arguments ProbDistUnivKernel object PD.
See Also iqr
ProbDistUnivKernel.icdf
18-600
ProbDistUnivParam.iqr
Purpose Return interquartile range (IQR) for ProbDistUnivParam object
Syntax Y = iqr(PD)
Description Y = iqr(PD) returns Y, the interquartile range for the
ProbDistUnivParam object PD. The interquartile range is the distance
between the 75th and 25th percentiles.
Input PD An object of the class ProbDistUnivParam.
Arguments
Output Y The value of the interquartile range for the
Arguments ProbDistUnivParam object PD.
See Also iqr
ProbDistUnivParam.icdf
18-601
classregtree.isbranch
Purpose Test node for branch
Syntax ib = isbranch(t)
ib = isbranch(t,nodes)
Description ib = isbranch(t) returns an n-element logical vector ib that is true
for each branch node and false for each leaf node.
ib = isbranch(t,nodes) takes a vector nodes of node numbers and
returns a vector of logical values for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-602
classregtree.isbranch
ib = isbranch(t)
ib =
1
0
1
1
0
1
0
0
18-603
classregtree.isbranch
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, cutvar, numnodes
18-604
categorical.isempty
Purpose True for empty categorical array
Syntax TF = isempty(A)
Description TF = isempty(A) returns true (1) if A is an empty categorical array
and false (0) otherwise. An empty array has no elements, that is
numel(A)==0.
See Also numel, size
18-605
dataset.isempty
Purpose True for empty dataset array
Syntax tf = isempty(A)
Description tf = isempty(A) returns true (1) if A is an empty dataset and false (0)
otherwise. An empty array has no elements, that is prod(size(A))==0.
See Also size
18-606
categorical.isequal
Purpose True if categorical arrays are equal
Syntax TF = isequal(A,B)
TF = isequal(A,B,C,...)
Description TF = isequal(A,B) is true (1) if the categorical arrays A and B are the
same class, have the same size and the same sets of levels, and contain
the same values, and false (0) otherwise.
TF = isequal(A,B,C,...) is true (1) if all the input arguments are
equal.
Elements with undefined levels are not considered equal to each other.
See Also getlabels
18-607
categorical.islevel
Purpose Test for levels
Syntax I = islevel(levels,A)
Description I = islevel(levels,A) returns a logical array I the same size as
the string, cell array of strings, or 2-D character matrix levels. I is
true (1) where the corresponding element of levels is the label of a
level in the categorical array A, even if the level contains no elements.
I is false (0) otherwise.
Examples Display age levels in the data in hospitl.mat, before and after dropping
occupied levels:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
disp(labels')
'0s' '10s' '20s' '30s' '40s' '50s' '60s' '70s' '80s' '90s'
AgeGroup = ordinal(hospital.Age,labels,[],edges);
I = islevel(labels,AgeGroup);
disp(I')
1 1 1 1 1 1 1 1 1 1
AgeGroup = droplevels(AgeGroup);
I = islevel(labels,AgeGroup);
disp(I')
0 0 1 1 1 1 0 0 0 0
See Also ismember, isundefined
18-608
ordinal.ismember
Purpose Test for membership
Syntax I = ismember(A,levels)
[I,IDX] = ismember(A,levels)
Description I = ismember(A,levels) returns a logical array I the same size as the
categorical array A. I is true (1) where the corresponding element of A
is one of the levels specified by the labels in the categorical array, cell
array of strings, or 2-D character array levels. I is false (0) otherwise.
[I,IDX] = ismember(A,levels) also returns an array of indices IDX
containing the highest absolute index in levels for each element in A
whose level is a member of levels, and 0 if there is no such index.
Examples Example 1
For nominal data:
load hospital
sex = hospital.Sex; % Nominal
smokers = hospital.Smoker; % Logical
I = ismember(sex(smokers),'Female');
I(1:5)
ans =
0
1
0
0
0
The use of ismember above is equivalent to:
I = (sex(smokers) == 'Female');
Example 2
For ordinal data:
load hospital
edges = 0:10:100;
18-609
ordinal.ismember
labels = strcat(num2str((0:10:90)','%d'),{'s'});
AgeGroup = ordinal(hospital.Age,labels,[],edges);
I = ismember(AgeGroup(1:5),{'20s','30s'})
I =
1
0
1
0
0
See Also islevel, isundefined
18-610
categorical.ismember
Purpose True for elements of categorical array in set
Syntax TF = ismember(A,levels)
[TF,LOC] = ismember(A,levels)
Description TF = ismember(A,levels) returns a logical array the same size as
the categorical array A, containing true (1) where the level of the
corresponding element of A is equal to one of the levels specified in
levels, and false (0) otherwise. levels is a categorical array, or a cell
array of strings or 2-D character array containing level labels.
[TF,LOC] = ismember(A,levels) also returns an index array LOC
containing the highest absolute index in levels for each element in A
whose level is a member of levels, and 0 if there is no such index.
See Also intersect, islevel, setdiff, setxor, union, unique
18-611
categorical.isscalar
Purpose True if categorical array is scalar
Syntax TF = isscalar(A)
Description TF = isscalar(A) returns true (1) if the categorical array A is a 1-by-1
matrix, and false (0) otherwise.
See Also isempty, isvector, size
18-612
categorical.isundefined
Purpose Test for undefined elements
Syntax I = isundefined(A)
Description I = isundefined(A) returns a logical array I the same size as the
categorical array A. I is true (1) where the corresponding element of
A is not assigned to any level. I is false (0) where the corresponding
element of A is assigned to a level.
Examples Create and display undefined levels in an ordinal array:
A = ordinal([1 2 3 2 1],{'lo','med','hi'})
A =
lo med hi med lo
A = droplevels(A,{'med','hi'})
Warning: OLDLEVELS contains categorical levels that
were present in A, caused some array elements to
have undefined levels.
A =
lo <undefined> <undefined> <undefined> lo
I = isundefined(A)
I =
0 1 1 1 0
See Also islevel, ismember
18-613
qrandstream.isvalid
Purpose Test handle validity
Syntax tf = isvalid(h)
Description tf = isvalid(h) performs an element-wise check for validity on
the handle elements of h. The result is a logical array of the same
dimensions as h, where each element is the element-wise validity result.
A handle is invalid if it has been deleted or if it is an element of a
handle array and has not yet been initialized.
See Also delete, qrandstream
18-614
categorical.isvector
Purpose True if categorical array is vector
Syntax TF = isvector(A)
Description TF = isvector(A) returns true (1) if the categorical array A is a 1-by-n
or n-by-1 vector, where n >= 0, and false (0) otherwise.
See Also isempty, isscalar, size
18-615
gmdistribution.Iters property
Purpose Number of iterations
Description The number of iterations of the algorithm.
Note This property applies only to gmdistribution objects constructed
with fit.
18-616
iwishrnd
Purpose Inverse Wishart random numbers
Syntax W = iwishrnd(Tau,df)
W = iwishrnd(Tau,df,DI)
[W,DI] = iwishrnd(Tau,df)
Description W = iwishrnd(Tau,df) generates a random matrix W from the inverse
Wishart distribution with parameters Tau and df. The inverse of W has
the Wishart distribution with covariance matrix Sigma = inv(Tau)
and with df degrees of freedom. Tau is a symmetric and positive
definite matrix.
W = iwishrnd(Tau,df,DI) expects DI to be the transpose of the inverse
of the Cholesky factor of Tau, so that DI'*DI = inv(Tau), where inv is
the MATLAB inverse function. DI is lower-triangular and the same size
as Tau. If you call iwishrnd multiple times using the same value of Tau,
it is more efficient to supply DI instead of computing it each time.
[W,DI] = iwishrnd(Tau,df) returns DI so you can use it as an input
in future calls to iwishrnd.
Note that different sources use different parametrizations for the
inverse Wishart distribution. This function defines the parameter tau
so that the mean of the output matrix is Tau/(df-d-1) where d is the
dimension of Tau.
See Also wishrnd
“Inverse Wishart Distribution” on page B-46
18-617
jackknife
Purpose Jackknife sampling
Syntax jackstat = jackknife(jackfun,X)
jackstat = jackknife(jackfun,X,Y,...)
jackstat = jackknife(jackfun,...,'Options',option)
Description jackstat = jackknife(jackfun,X) draws jackknife data samples
from the n-by-p data array X, computes statistics on each sample
using the function jackfun, and returns the results in the matrix
jackstat. jackfun is a function handle specified with @. Each of the
n rows of jackstat contains the results of applying jackfun to one
jackknife sample. Row i of jackstat contains the results for the sample
consisting of X with the ith row omitted:
s = x;
s(i,:) = [];
jackstat(i,:) = jackfun(s);
If jackfun returns a matrix or array, then this output is converted to a
row vector for storage in jackstat. If X is a row vector, it is converted
to a column vector.
jackstat = jackknife(jackfun,X,Y,...) accepts additional
arguments to be supplied as inputs to jackfun. They may be scalars,
column vectors, or matrices. jackknife creates each jackknife sample
by sampling with replacement from the rows of the non-scalar data
arguments (these must have the same number of rows). Scalar data
are passed to jackfun unchanged. Non-scalar arguments must have
the same number of rows, and each jackknife sample omits the same
row from these arguments.
jackstat = jackknife(jackfun,...,'Options',option) provides
an option to perform jackknife iterations in parallel, if the Parallel
Computing Toolbox is available. This argument is a struct that you can
create with a call to statset. You can retrieve values of the individual
fields with a call to statget. jackknife uses the following field in
the struct:
18-618
jackknife
'UseParallel' If 'always' and if a matlabpool of the Parallel
Computing Toolbox is open, use multiple processors
to compute jackknife iterations. If the Parallel
Computing Toolbox is not installed, or a matlabpool
is not open, computation occurs in serial mode.
Default is 'never', or serial computation.
Examples Estimate the bias of the MLE variance estimator of random samples
taken from the vector y using jackknife. The bias has a known
formula in this problem, so you can compare the jackknife value to
this formula.
y = exprnd(5,100,1);
m = jackknife(@var,y,1);
n = length(y);
bias = var(y,1)-var(y,0) % Bias formula
bias =
-0.2069
jbias = (n-1)*(mean(m)-var(y,1)) % Jackknife estimate
jbias =
-0.2069
See Also bootstrp, random, randsample, hist, ksdensity
18-619
jbtest
Purpose Jarque-Bera test
Syntax h = jbtest(x)
h = jbtest(x,alpha)
[h,p] = jbtest(...)
[h,p,jbstat] = jbtest(...)
[h,p,jbstat,critval] = jbtest(...)
[h,p,...] = jbtest(x,alpha,mctol)
Description h = jbtest(x) performs a Jarque-Bera test of the null hypothesis
that the sample in vector x comes from a normal distribution with
unknown mean and variance, against the alternative that it does not
come from a normal distribution. The test is specifically designed for
alternatives in “Generating Data Using the Pearson System” on page
6-28 of distributions. The test returns the logical value h = 1 if it rejects
the null hypothesis at the 5% significance level, and h = 0 if it cannot.
The test treats NaN values in x as missing values, and ignores them.
The Jarque-Bera test is a two-sided goodness-of-fit test suitable when a
fully-specified null distribution is unknown and its parameters must
be estimated. The test statistic is
n ⎛⎜ 2 ( k − 3 ) ⎞
2
JB = s + ⎟
6⎜ 4 ⎟
⎝ ⎠
where n is the sample size, s is the sample skewness, and k is the
sample kurtosis. For large sample sizes, the test statistic has a
chi-square distribution with two degrees of freedom.
Jarque-Bera tests often use the chi-square distribution to estimate
critical values for large samples, deferring to the Lilliefors test (see
lillietest) for small samples. jbtest, by contrast, uses a table of
critical values computed using Monte-Carlo simulation for sample sizes
less than 2000 and significance levels between 0.001 and 0.50. Critical
values for a test are computed by interpolating into the table, using the
analytic chi-square approximation only when extrapolating for larger
sample sizes.
18-620
jbtest
h = jbtest(x,alpha) performs the test at significance level alpha.
alpha is a scalar in the range [0.001, 0.50]. To perform the test at a
significance level outside of this range, use the mctol input argument.
[h,p] = jbtest(...) returns the p value p, computed using inverse
interpolation into the table of critical values. Small values of p cast
doubt on the validity of the null hypothesis. jbtest warns when p is
not found within the tabulated range of [0.001, 0.50], and returns either
the smallest or largest tabulated value. In this case, you can use the
mctol input argument to compute a more accurate p value.
[h,p,jbstat] = jbtest(...) returns the test statistic jbstat.
[h,p,jbstat,critval] = jbtest(...) returns the critical value
critval for the test. When jbstat > critval, the null hypothesis is
rejected at significance level alpha.
[h,p,...] = jbtest(x,alpha,mctol) computes a Monte-Carlo
approximation for p directly, rather than interpolating into the table
of pre-computed values. This is useful when alpha or p lie outside
the range of the table. jbtest chooses the number of Monte Carlo
replications, mcreps, large enough to make the Monte Carlo standard
error for p, sqrt(p*(1-p)/mcreps), less than mctol.
Examples Use jbtest to determine if car mileage, in miles per gallon (MPG),
follows a normal distribution across different makes of cars:
load carbig
[h,p] = jbtest(MPG)
h =
1
p =
0.0022
The p value is below the default significance level of 5%, and the test
rejects the null hypothesis that the distribution is normal.
With a log transformation, the distribution becomes closer to normal,
but the p value is still well below 5%:
18-621
jbtest
[h,p] = jbtest(log(MPG))
h =
1
p =
0.0078
Decreasing the significance level makes it harder to reject the null
hypothesis:
[h,p] = jbtest(log(MPG),0.0075)
h =
0
p =
0.0078
References [1] Jarque, C. M., and A. K. Bera. “A test for normality of observations
and regression residuals.” International Statistical Review. Vol. 55,
No. 2, 1987, pp. 163–172.
[2] Deb, P., and M. Sefton. “The Distribution of a Lagrange Multiplier
Test of Normality.” Economics Letters. Vol. 51, 1996, pp. 123–130.
This paper proposed a Monte Carlo simulation for determining the
distribution of the test statistic. The results of this function are based on
an independent Monte Carlo simulation, not the results in this paper.
18-622
johnsrnd
Purpose Johnson system random numbers
Syntax r = johnsrnd(quantiles,m,n)
r = johnsrnd(quantiles)
[r,type] = johnsrnd(...)
[r,type,coefs] = johnsrnd(...)
Description r = johnsrnd(quantiles,m,n) returns an m-by-n matrix of random
numbers drawn from the distribution in the Johnson system that
satisfies the quantile specification given by quantiles. quantiles is
a four-element vector of quantiles for the desired distribution that
correspond to the standard normal quantiles [–1.5 –0.5 0.5 1.5]. In other
words, you specify a distribution from which to draw random values by
designating quantiles that correspond to the cumulative probabilities
[0.067 0.309 0.691 0.933]. quantiles may also be a 2-by-4 matrix whose
first row contains four standard normal quantiles, and whose second
row contains the corresponding quantiles of the desired distribution.
The standard normal quantiles must be spaced evenly.
Note Because r is a random sample, its sample quantiles typically
differ somewhat from the specified distribution quantiles.
r = johnsrnd(quantiles) returns a scalar value.
r = johnsrnd(quantiles,m,n,...) or r =
johnsrnd(quantiles,[m,n,...]) returns an m-by-n-by-... array.
[r,type] = johnsrnd(...) returns the type of the specified
distribution within the Johnson system. type is 'SN', 'SL', 'SB',
or 'SU'. Set m and n to zero to identify the distribution type without
generating any random values.
The four distribution types in the Johnson system correspond to the
following transformations of a normal random variate:
• 'SN' — Identity transformation (normal distribution)
18-623
johnsrnd
• 'SL' — Exponential transformation (lognormal distribution)
• 'SB' — Logistic transformation (bounded)
• 'SU' — Hyperbolic sine transformation (unbounded)
[r,type,coefs] = johnsrnd(...) returns coefficients coefs of
the transformation that defines the distribution. coefs is [gamma,
eta, epsilon, lambda]. If z is a standard normal random
variable and h is one of the transformations defined above, r =
lambda*h((z-gamma)/eta)+epsilon is a random variate from the
distribution type corresponding to h.
Examples Generate random values with longer tails than a standard normal:
r = johnsrnd([-1.7 -.5 .5 1.7],1000,1);
qqplot(r);
18-624
johnsrnd
Generate random values skewed to the right:
r = johnsrnd([-1.3 -.5 .5 1.7],1000,1);
qqplot(r);
18-625
johnsrnd
Generate random values that match some sample data well in the
right-hand tail:
load carbig;
qnorm = [.5 1 1.5 2];
q = quantile(Acceleration, normcdf(qnorm));
r = johnsrnd([qnorm;q],1000,1);
[q;quantile(r,normcdf(qnorm))]
ans =
16.7000 18.2086 19.5376 21.7263
16.8190 18.2474 19.4492 22.4156
18-626
johnsrnd
Determine the distribution type and the coefficients:
[r,type,coefs] = johnsrnd([qnorm;q],0)
r =
[]
type =
SU
coefs =
1.0920 0.5829 18.4382 1.4494
See Also random, pearsrnd
“Johnson System” on page B-48
18-627
dataset.join
Purpose Merge observations
Syntax C = join(A,B)
C = join(A,B,key)
C = join(A,B,param1,val1,param2,val2,...)
[C,IB] = join(...)
C = join(A,B,'Type',TYPE,...)
[C,IA,IB] = join(A,B,'Type',TYPE,...)
Description C = join(A,B) creates a dataset array C by merging observations from
the two dataset arrays A and B. join performs the merge by first finding
key variables, that is, a pair of dataset variables, one in A and one in
B, that share the same name. The key from B must contain unique
values, and must contain all the values that are present in the key
from A. join then uses these key variables to define a many-to-one
correspondence between observations in A and those in B. join uses this
correspondence to replicate the observations in B and combine them
with the observations in A to create C.
C = join(A,B,key) performs the merge using the variable specified
by key as the key variable in both A and B. key is a positive integer, a
variable name, a cell array containing a variable name, or a logical
vector with one true entry.
C contains one observation for each observation in A. Variables in C
include all of the variables from A, as well as one variable corresponding
to each variable in B (except for the key from B). If A and B contain
variables with identical names, join adds the suffix '_left' and
'_right' to the corresponding variables in C.
C = join(A,B,param1,val1,param2,val2,...) specifies optional
parameter name/value pairs to control how the dataset variables in A
and B are used in the merge. Parameters are:
• 'Keys' — Specifies the variable to use as a keys in both A and B.
• 'LeftKeys' — Specifies the variable to use as a keys in A.
• 'RightKeys' — Specifies the variable to use as a keys in B.
18-628
dataset.join
You may provide either the 'Keys' parameter, or both the 'LeftKeys'
and 'RightKeys' parameters. The value for these parameters
is a positive integer, a variable name, a cell array containing a
variable name, or a logical vector with one true entry. 'LeftKeys' or
'RightKeys' must both specify the same number of key variables, and
join pairs the left and right keys are paired in the order specified.
• 'LeftVars' — Specifies the variables from A to include in C. By
default, join includes all variables from A.
• 'RightVars' — Specifies the variables from B to include in C. By
default, join includes all variables from B except the key variable.
You can use 'LeftVars' or 'RightVars' to include or exclude key
variables as well as data variables. The value for these parameters is
a positive integer, a vector of positive integers, a variable name, a cell
array containing one or more variable names, or a logical vector.
[C,IB] = join(...)returns an index vector IB, where join constructs
C by horizontally concatenating A(:,LeftVars) and B(IB,RightVars).
join can also perform more complicated inner and outer join operations
that allow a many-to-many correspondence between A and B, and allow
unmatched observations in either A or B.
C = join(A,B,'Type',TYPE,...)performs the join operation specified
by TYPE. TYPE is one of 'inner', 'leftouter', 'rightouter',
'fullouter', or 'outer' (which is a synonym for 'fullouter'). For an
inner join, C only contains observations corresponding to a combination
of key values that occurred in both A and B. For a left (or right) outer
join, C also contains observations corresponding to keys in A (or B) that
did not match any in B (or A). Variables in C taken from A (or B) contain
null values in those observations. A full outer join is equivalent to a
left and right outer join. C contains variables corresponding to the key
variables from both A and B, and join sorts the observations in C by
the key values.
For inner and outer joins, C contains variables corresponding to the key
variables from both A and B by default, as well as all the remaining
variables. join sorts the observations in the result C by the key values.
18-629
dataset.join
[C,IA,IB] = join(A,B,'Type',TYPE,...) returns index vectors IA
and IB indicating the correspondence between observations in C and
those in A and B. For an inner join, join constructs C by horizontally
concatenating A(IA,LeftVars) and B(IB,RightVars). For an outer
join, IA or IB may also contain zeros, indicating the observations in C
that do not correspond to observations in A or B, respectively.
Examples Create a dataset array from Fisher’s iris data:
load fisheriris
NumObs = size(meas,1);
NameObs = strcat({'Obs'},num2str((1:NumObs)','%-d'));
iris = dataset({nominal(species),'species'},...
{meas,'SL','SW','PL','PW'},...
'ObsNames',NameObs);
Create a separate dataset array with the diploid chromosome counts for
each species of iris:
snames = nominal({'setosa';'versicolor';'virginica'});
CC = dataset({snames,'species'},{[38;108;70],'cc'})
CC =
species cc
setosa 38
versicolor 108
virginica 70
Broadcast the data in CC to the rows of iris using the key variable
species in each dataset:
iris2 = join(iris,CC);
iris2([1 2 51 52 101 102],:)
ans =
species SL SW PL PW cc
Obs1 setosa 5.1 3.5 1.4 0.2 38
18-630
dataset.join
Obs2 setosa 4.9 3 1.4 0.2 38
Obs51 versicolor 7 3.2 4.7 1.4 108
Obs52 versicolor 6.4 3.2 4.5 1.5 108
Obs101 virginica 6.3 3.3 6 2.5 70
Obs102 virginica 5.8 2.7 5.1 1.9 70
Create two datasets and join them using the 'MergeKeys' flag:
% Create two data sets that both contain the key variable
% 'Key1'. The two arrays contain observations with common
% values of Key1, but each array also contains observations
% with values of Key1 not present in the other.
a = dataset({'a' 'b' 'c' 'e' 'h'}',[1 2 3 11 17]',...
'VarNames',{'Key1' 'Var1'})
b = dataset({'a' 'b' 'd' 'e'}',[4 5 6 7]',...
'VarNames',{'Key1' 'Var2'})
% Combine a and b with an outer join, which matches up
% observations with common key values, but also retains
% observations whose key values don't have a match.
% Keep the key values as separate variables in the result.
couter = join(a,b,'key','Key1','Type','outer')
% Join a and b, merging the key values as a single variable
% in the result.
coutermerge = join(a,b,'key','Key1','Type','outer',...
'MergeKeys',true)
% Join a and b, retaining only observations whose key
% values match.
cinner = join(a,b,'key','Key1','Type','inner',...
'MergeKeys',true)
See Also sortrows
18-631
KDTreeSearcher class
Superclasses NeighborSearcher
Purpose Nearest neighbors search using kd-tree
Description A KDTreeSearcher object represents kNN (k-nearest neighbor) search
using a kd-tree. Search objects store information about the data used,
the distance metric and parameters, and the maximal number of data
points in each leaf node. You cannot create this object for sparse
input data. The search performance for this object, compared with the
ExhaustiveSearcher object, tends to be better for smaller dimensions
(10 or fewer) and worse for larger dimensions. For more information on
search objects, see “What Are Search Objects?” on page 12-20.
Construction NS = KDTreeSearcher(X,'Name',Value) constructs a kd-tree based on
X and saves the information in a KDTreeSearcher object where rows of
X correspond to observations and columns correspond to variables. You
can use this tree to find neighbors in X nearest to the query points. See
the following section for optional name/value pairs.
NS = createns(X,'NSMethod','kdtree','Name',Value) creates a
kd-tree based on X using createns and saves the information in a
KDTreeSearcher object where rows of X correspond to observations and
columns correspond to variables. You can use this tree to find neighbors
in X nearest to the query points. See the following section for optional
name/value pairs.
Name/Value Pairs
KDTreeSearcher and createns accept one or more of the following
optional name/value pairs as input:
Distance
A string specifying the default distance metric used when you call
the knnsearch method.
• 'euclidean' — Euclidean distance (default).
• 'cityblock' — City block distance.
18-632
KDTreeSearcher class
• 'chebychev' — Chebychev distance (maximum coordinate
difference).
• 'minkowski' — Minkowski distance.
For more information on these distance metrics, see “Distance
Metrics” on page 12-14.
P
A positive scalar indicating the exponent of the Minkowski
distance. This parameter is only valid when Distance is
'minkowski'. Default is 2.
BucketSize
A positive integer, indicating the maximum number of data points
in each leaf node of the kd-tree. Default is 50.
Properties X
A matrix used to create the object
Distance
A string specifying a built-in distance metric that you provide
when you create the object. This property is the default distance
metric used when you call the knnsearch method to find nearest
neighbors for future query points.
DistParameter
Specifies the additional parameter for the chosen distance metric.
The value is:
• If 'Distance' is 'minkowski': A positive scalar indicating the
exponent of the Minkowski distance.
• Otherwise: Empty.
18-633
KDTreeSearcher class
Methods knnsearch Find k-nearest neighbors using
KDTreeSearcher object
Examples Create a KDTreeSearcher object using the constructor:
load fisheriris
x = meas(:,3:4);
kdtreeobj = kdtreesearcher(x,'distance','minkowski')
kdtreeobj =
KDTreeSearcher
Properties:
BucketSize: 50
X: [150x2 double]
Distance: 'minkowski'
DistParameter: 2
Create a KDTreeSearcher object using createns:
load fisheriris
x = meas(:,3:4);
kdtreeobj = createns(x,'NsMethod','kdtree',...
'distance','minkowski')
kdtreeobj =
KDTreeSearcher
Properties:
BucketSize: 50
X: [150x2 double]
Distance: 'minkowski'
18-634
KDTreeSearcher class
DistParameter: 2
For more in-depth examples using the knnsearch method, see the
method reference page or see “Example: Classifying Query Data Using
knnsearch” on page 12-22.
References [1] Friedman, J. H., Bentely, J., and Finkel, R. A. (1977). An Algorithm
for Finding Best Matches in Logarithmic Expected Time, ACM
Transactions on Mathematical Software 3, 209.
See Also createns | ExhaustiveSearcher | NeighborSearcher
How To • “k-Nearest Neighbor Search” on page 12-17
18-635
ProbDistKernel.Kernel property
Purpose Read-only string specifying name of kernel smoothing function for
ProbDistKernel object
Description Kernel is a read-only property of the ProbDistKernel class. Kernel is
a string specifying the name of the kernel smoothing function used
to create a ProbDistKernel object.
Values 'normal'
'box'
'triangle'
'epanechnikov'
Use this information to view and compare the kernel smoothing
function used to create distributions.
See Also ksdensity
18-636
kmeans
Purpose K-means clustering
Syntax IDX = kmeans(X,k)
[IDX,C] = kmeans(X,k)
[IDX,C,sumd] = kmeans(X,k)
[IDX,C,sumd,D] = kmeans(X,k)
[...] = kmeans(...,param1,val1,param2,val2,...)
Description IDX = kmeans(X,k) partitions the points in the n-by-p data matrix X
into k clusters. This iterative partitioning minimizes the sum, over
all clusters, of the within-cluster sums of point-to-cluster-centroid
distances. Rows of X correspond to points, columns correspond to
variables. kmeans returns an n-by-1 vector IDX containing the cluster
indices of each point. By default, kmeans uses squared Euclidean
distances.
[IDX,C] = kmeans(X,k) returns the k cluster centroid locations in
the k-by-p matrix C.
[IDX,C,sumd] = kmeans(X,k) returns the within-cluster sums of
point-to-centroid distances in the 1-by-k vector sumd.
[IDX,C,sumd,D] = kmeans(X,k) returns distances from each point to
every centroid in the n-by-k matrix D.
[...] = kmeans(...,param1,val1,param2,val2,...) enables
you to specify optional parameter/value pairs to control the iterative
algorithm used by kmeans. Valid parameter strings are listed in the
following table.
18-637
kmeans
Parameter Value
'distance' Distance measure, in p-dimensional space. kmeans
minimizes with respect to this parameter. kmeans
computes centroid clusters differently for the
different supported distance measures.
'sqEuclidean' Squared Euclidean distance
(default). Each centroid is the
mean of the points in that cluster.
'cityblock' Sum of absolute differences, i.e.,
the L1 distance. Each centroid
is the component-wise median of
the points in that cluster.
'cosine' One minus the cosine of the
included angle between points
(treated as vectors). Each
centroid is the mean of the points
in that cluster, after normalizing
those points to unit Euclidean
length.
'correlation' One minus the sample correlation
between points (treated as
sequences of values). Each
centroid is the component-wise
mean of the points in that cluster,
after centering and normalizing
those points to zero mean and
unit standard deviation.
'Hamming' Percentage of bits that differ (only
suitable for binary data). Each
centroid is the component-wise
median of points in that cluster.
18-638
kmeans
Parameter Value
'emptyaction' Action to take if a cluster loses all its member
observations.
'error' Treat an empty cluster as an
error (default).
'drop' Remove any clusters that
become empty. kmeans sets the
corresponding return values in C
and D to NaN.
'singleton' Create a new cluster consisting
of the one point furthest from its
centroid.
'onlinephase' Flag indicating whether kmeans should perform an
online update phase in addition to a batch update
phase. The online phase can be time consuming
for large data sets, but guarantees a solution that
is a local minimum of the distance criterion, that
is, a partition of the data where moving any single
point to a different cluster increases the total sum
of distances.
'on' Perform online update (default).
'off' Do not perform online update.
'options' Options for the iterative algorithm used to minimize
the fitting criterion, as created by statset.
'replicates' Number of times to repeat the clustering, each
with a new set of initial cluster centroid positions.
kmeans returns the solution with the lowest value
for sumd. You can supply 'replicates' implicitly
by supplying a 3D array as the value for the
'start' parameter.
18-639
kmeans
Parameter Value
'start' Method used to choose the initial cluster centroid
positions, sometimes known as seeds.
'sample' Select k observations from X at
random (default).
'uniform' Select k points uniformly at
random from the range of X. Not
valid with Hamming distance.
'cluster' Perform a preliminary clustering
phase on a random 10%
subsample of X. This preliminary
phase is itself initialized using
’sample’.
Matrix k-by-p matrix of centroid starting
locations. In this case, you can
pass in [] for k, and kmeans infers
k from the first dimension of the
matrix. You can also supply a 3-D
array, implying a value for the
'replicates' parameter from
the array’s third dimension.
Algorithm kmeans uses a two-phase iterative algorithm to minimize the sum of
point-to-centroid distances, summed over all k clusters:
1 The first phase uses batch updates, where each iteration consists
of reassigning points to their nearest cluster centroid, all at once,
followed by recalculation of cluster centroids. This phase occasionally
does not converge to solution that is a local minimum, that is, a
partition of the data where moving any single point to a different
cluster increases the total sum of distances. This is more likely
for small data sets. The batch phase is fast, but potentially only
approximates a solution as a starting point for the second phase.
18-640
kmeans
2 The second phase uses online updates, where points are individually
reassigned if doing so will reduce the sum of distances, and cluster
centroids are recomputed after each reassignment. Each iteration
during the second phase consists of one pass though all the points.
The second phase will converge to a local minimum, although there
may be other local minima with lower total sum of distances. The
problem of finding the global minimum can only be solved in general
by an exhaustive (or clever, or lucky) choice of starting points, but
using several replicates with random starting points typically results
in a solution that is a global minimum.
References [1] Seber, G. A. F. Multivariate Observations. Hoboken, NJ: John Wiley
& Sons, Inc., 1984.
[2] Spath, H. Cluster Dissection and Analysis: Theory, FORTRAN
Programs, Examples. Translated by J. Goldschmidt. New York:
Halsted Press, 1985.
Examples The following creates two clusters from separated random data:
X = [randn(100,2)+ones(100,2);...
randn(100,2)-ones(100,2)];
opts = statset('Display','final');
[idx,ctrs] = kmeans(X,2,...
'Distance','city',...
'Replicates',5,...
'Options',opts);
5 iterations, total sum of distances = 284.671
4 iterations, total sum of distances = 284.671
4 iterations, total sum of distances = 284.671
3 iterations, total sum of distances = 284.671
3 iterations, total sum of distances = 284.671
plot(X(idx==1,1),X(idx==1,2),'r.','MarkerSize',12)
hold on
plot(X(idx==2,1),X(idx==2,2),'b.','MarkerSize',12)
18-641
kmeans
plot(ctrs(:,1),ctrs(:,2),'kx',...
'MarkerSize',12,'LineWidth',2)
plot(ctrs(:,1),ctrs(:,2),'ko',...
'MarkerSize',12,'LineWidth',2)
legend('Cluster 1','Cluster 2','Centroids',...
'Location','NW')
See Also linkage, clusterdata, silhouette
18-642
ExhaustiveSearcher.knnsearch
Purpose Find k-nearest neighbors using ExhaustiveSearcher object
Syntax IDX = knnsearch(NS,Y)
[IDX,D] = knnsearch(NS,Y)
[IDX,D] = knnsearch(NS,Y,'Name',Value)
Description IDX = knnsearch(NS,Y) finds the nearest neighbor (closest point) in
NS.X for each point in Y. Rows of Y correspond to observations and
columns correspond to features. Y must have the same number of
columns as NS.X. IDX is a column vector with ny rows, where ny is the
number of rows in Y. Each row in IDX contains the index of observation
in NS.X which has the smallest distance to the corresponding
observation in Y.
[IDX,D] = knnsearch(NS,Y) returns a column vector D containing
the distances between each observation in Y and the corresponding
closest observation in NS.X. That is, D(i) is the distance between
NS.X(IDX(i),:) and Y(i,:).
[IDX,D] = knnsearch(NS,Y,'Name',Value) accepts one or more
comma-separated argument name/value pairs. Specify Name inside
single quotes.
Input Name/Value Pairs
Arguments
K
A positive integer, k, specifying the number of nearest neighbors
in NS.X for each point in Y. Default is 1. IDX and D are ny-by-k
matrices. D sorts the distances in each row in ascending order.
Each row in IDX contains the indices of the k closest neighbors in
NS.X corresponding to the k smallest distances in D.
Distance
• 'euclidean' — Euclidean distance (default).
• 'seuclidean' — Standardized Euclidean distance. Each
coordinate difference between X and each query point is
18-643
ExhaustiveSearcher.knnsearch
scaled by dividing by a scale value S. The default value of
S is NS.DistParameter if NS.Distance is 'seuclidean',
otherwise the default is the standard deviation computed from
X, S=nanstd(X). To specify another value for S, use the 'Scale'
argument.
• 'cityblock' — City block distance.
• 'chebychev' — Chebychev distance (maximum coordinate
difference).
• 'minkowski' — Minkowski distance.
• 'mahalanobis' — Mahalanobis distance, which is computed
using a positive definite covariance matrix C. The default
value of C is nancov(X). To change the value of C, use the Cov
parameter.
• 'cosine' — One minus the cosine of the included angle
between observations (treated as vectors).
• 'correlation' — One minus the sample linear correlation
between observations (treated as sequences of values).
• 'spearman' — One minus the sample Spearman’s rank
correlation between observations (treated as sequences of
values).
• 'hamming' — Hamming distance, which is the percentage of
coordinates that differ.
• 'jaccard' — One minus the Jaccard coefficient, which is the
percentage of nonzero coordinates that differ.
• custom distance function — A distance function specified using
@ (for example, @distfun). A distance function must be of the
form function D2 = distfun(ZI, ZJ), taking as arguments
a 1-by-n vector ZI containing a single row of from X or from
the query points Y, an m2-by-n matrix ZJ containing multiple
rows of X or Y, and returning an m2-by-1 vector of distances D2,
whose jth element is the distance between the observations
ZI and ZJ(j,:).
18-644
ExhaustiveSearcher.knnsearch
Default is NS.Distance. For more information on these distance
metrics, see “Distance Metrics” on page 12-14.
P
A positive scalar, p, indicating the exponent of the Minkowski
distance. This parameter is only valid if knnsearch uses the
'minkowski' distance metric. Default is NS.DistParameter if
NS.Distance is 'minkowski' and 2 otherwise.
Cov
A positive definite matrix indicating the covariance matrix when
computing the Mahalanobis distance. This parameter is only
valid when knnsearch uses the 'mahalanobis' distance metric.
Default is NS.DistParameter if NS.Distance is 'mahalanobis',
or nancov(X) otherwise.
Scale
A vector S with the length equal to the number of columns in
X. Each coordinate of X and each query point is scaled by the
corresponding element of S when computing the standardized
Euclidean distance. This parameter is only valid when Distance
is 'seuclidean'. Default is nanstd(X).
Examples Create an ExhaustiveSearcher object specifying 'cosine' as the
distance metric. Perform a k-nearest neighbors search on the object
using the mahalanobis metric and compare the results:
load fisheriris
x = meas(:,3:4);
exhaustiveobj = ExhaustiveSearcher(x,'Distance','cosine')
exhaustiveobj =
ExhaustiveSearcher
Properties:
X: [150x2 double]
18-645
ExhaustiveSearcher.knnsearch
Distance: 'cosine'
DistParameter: []
% Perform a knnsearch between x and a query point, using
% first cosine then mahalanobis distance metrics:
newpoint = [5 1.45];
[n,d]=knnsearch(exhaustiveobj,newpoint,'k',10);
[nmah,dmah] = knnsearch(exhaustiveobj,newpoint,'k',10,...
'distance','mahalanobis');
% Visualize the results of the two different nearest
% neighbors searches:
% First plot the training data:
gscatter(x(:,1),x(:,2),species)
% Plot an X for the query point:
line(newpoint(1),newpoint(2),'marker','x','color','k',...
'markersize',10,'linewidth',2,'linestyle','none')
% Use circles to denote the cosine nearest neighbors:
line(x(n,1),x(n,2),'color',[.5 .5 .5],'marker','o',...
'linestyle','none','markersize',10)
% Use pentagrams to denote the mahalanobis nearest neighbors:
line(x(nmah,1),x(nmah,2),'color',[.5 .5 .5],'marker','p',...
'linestyle','none','markersize',10)
legend('setosa','versicolor','virginica','query point',...
'cosine','mahalanobis')
set(legend,'location','best')
18-646
ExhaustiveSearcher.knnsearch
Algorithm For information on a specific search algorithm, see “Distance Metrics”
on page 12-14.
See Also createns | ExhaustiveSearcher | KDTreeSearcher.knnsearch |
knnsearch
How To • “k-Nearest Neighbor Search” on page 12-17
• “Distance Metrics” on page 12-14
18-647
KDTreeSearcher.knnsearch
Purpose Find k-nearest neighbors using KDTreeSearcher object
Syntax IDX = knnsearch(NS,Y)
[IDX,D] = knnsearch(NS,Y)
[IDX,D] = knnsearch(NS,Y,'Name',Value)
Description IDX = knnsearch(NS,Y) finds the nearest neighbor (closest point) in
NS.X for each point in Y. Rows of Y correspond to observations and
columns correspond to features. Y must have the same number of
columns as NS.X. IDX is a column vector with ny rows, where ny is the
number of rows in Y. Each row in IDX contains the index of observation
in NS.X which has the smallest distance to the corresponding
observation in Y.
[IDX,D] = knnsearch(NS,Y) returns a column vector D containing
the distances between each observation in Y and the corresponding
closest observation in NS.X. That is, D(i) is the distance between
NS.X(IDX(i),:) and Y(i,:).
[IDX,D] = knnsearch(NS,Y,'Name',Value) accepts one or more
comma-separated name/value pairs. Specify Name inside single quotes.
Input Name/Value Pairs
Arguments
K
A positive integer, k, specifying the number of nearest neighbors
in NS.X for each point in Y. Default is 1. IDX and D are ny-by-k
matrices. D sorts the distances in each row in ascending order.
Each row in IDX contains the indices of the k closest neighbors in
NS.X corresponding to the k smallest distances in D.
Distance
Select one of the following distance algorithms.
• 'euclidean' — Euclidean distance (default).
• 'cityblock' — City block distance.
18-648
KDTreeSearcher.knnsearch
• 'chebychev' — Chebychev distance (maximum coordinate
difference).
• 'minkowski' — Minkowski distance.
Default is NS.Distance. For more information on these distance
metrics, see “Distance Metrics” on page 12-14.
P
A positive scalar, p, indicating the exponent of the Minkowski
distance. This parameter is only valid whenthe Distance is
'minkowski'. Default is NS.DistParameter if NS.Distance is
'minkowski' and 2 otherwise.
Examples Create a KDTreeSearcher object specifying 'minkowski' as the distance
metric with an exponent of 5. Perform a k-nearest neighbors search on
the object using the chebychev metric and compare the results:
load fisheriris
x = meas(:,3:4);
kdtreeNS = KDTreeSearcher(x,'Distance','minkowski','P',5)
kdtreeNS =
KDTreeSearcher
Properties:
BucketSize: 50
X: [150x2 double]
Distance: 'minkowski'
DistParameter: 5
% Perform a knnsearch between X and a query point, using
% first Minkowski then Chebychev distance metrics:
newpoint = [5 1.45];
[n,d]=knnsearch(kdtreeNS,newpoint,'k',10);
[ncb,dcb] = knnsearch(kdtreeNS,newpoint,'k',10,...
18-649
KDTreeSearcher.knnsearch
'distance','chebychev');
% Visualize the results of the two different nearest
% neighbors searches:
% First plot the training data:
gscatter(x(:,1),x(:,2),species)
% Zoom in on the points of interest:
set(gca,'xlim',[4.5 5.5],'ylim',[1 2]); axis square
% Plot an X for the query point:
line(newpoint(1),newpoint(2),'marker','x','color','k',...
'markersize',10,'linewidth',2,'linestyle','none')
% Use circles to denote the Minkowski nearest neighbors:
line(x(n,1),x(n,2),'color',[.5 .5 .5],'marker','o',...
'linestyle','none','markersize',10)
% Use pentagrams to denote the Chebychev nearest neighbors:
line(x(ncb,1),x(ncb,2),'color',[.5 .5 .5],'marker','p',...
'linestyle','none','markersize',10)
legend('setosa','versicolor','virginica','query point',...
'minkowski','chebychev')
set(legend,'location','best')
18-650
KDTreeSearcher.knnsearch
Algorithm For information on a specific search algorithm, see “Distance Metrics”
on page 12-14.
References [1] Friedman, J. H., Bentely, J., and Finkel, R. A. (1977). An Algorithm
for Finding Best Matches in Logarithmic Expected Time, ACM
Transactions on Mathematical Software 3, 209.
See Also createns | ExhaustiveSearcher | ExhaustiveSearcher.knnsearch |
knnsearch
How To • “k-Nearest Neighbor Search” on page 12-17
18-651
KDTreeSearcher.knnsearch
• “Distance Metrics” on page 12-14
18-652
knnsearch
Purpose Find k-nearest neighbors using data
Syntax IDX = knnsearch(X,Y)
[IDX,Dist] = knnsearch(X,Y)
[IDX,Dist] = knnsearch(X,Y,'Name',Value)
Description IDX = knnsearch(X,Y) finds the nearest neighbor in X for each point
in Y. X is an mx-by-n matrix and Y is an my-by-n matrix. Rows of X and
Y correspond to observations and columns correspond to variables. IDX
is a column vector with my rows. Each row in IDX contains the index of
nearest neighbor in X for the corresponding row in Y.
[IDX,Dist] = knnsearch(X,Y) returns an my-by-1 vector Dist
containing the distances between each observation in Y and the
corresponding closest observation in X. That is, Dist(i) is the distance
between X(IDX(i),:) and Y(i,:).
[IDX,Dist] = knnsearch(X,Y,'Name',Value) accepts one or more
optional comma-separated name/value pairs. Specify Name inside single
quotes.
knnsearch does not save a search object. To create a search object,
use createns.
Input Name/Value Pairs
Arguments
K
A positive integer, k, specifying the number of nearest neighbors
in X for each point in Y. Default is 1. IDX and D are ny-by-k
matrices. D sorts the distances in each row in ascending order.
Each row in IDX contains the indices of the k closest neighbors in
X corresponding to the k smallest distances in D.
NSMethod
Nearest neighbors search method. Value is either:
18-653
knnsearch
• 'kdtree' — Creates and uses a kd-tree to find nearest
neighbors. This is the default value when the number of
columns of X is less than 10, X is not sparse, and the distance
measure is one of the following measures. 'kdtree' is only
valid when the distance measure is one of the following:
- 'euclidean'
- 'cityblock'
- 'minkowski'
- 'chebychev'
• 'exhaustive' — Uses the exhaustive search algorithm by
computing the distance values from all the points in X to each
point in Y to find nearest neighbors.
Distance
A string or a function handle specifying the distance metric. The
value can be one of the following:
• 'euclidean' — Euclidean distance (default).
• 'seuclidean' — Standardized Euclidean distance. Each
coordinate difference between rows in X and the query matrix is
scaled by dividing by the corresponding element of the standard
deviation computed from X, S=nanstd(X). To specify another
value for S, use the Scale argument.
• 'cityblock' — City block distance.
• 'chebychev' — Chebychev distance (maximum coordinate
difference).
• 'minkowski' — Minkowski distance. The default exponent is
2. To specify a different exponent, use the 'P' argument.
• 'mahalanobis' — Mahalanobis distance, computed using a
positive definite covariance matrix C. The default value of C is
nancov(X). To change the value of C, use the Cov parameter.
18-654
knnsearch
• 'cosine' — 1 minus the cosine of the included angle between
observations (treated as vectors).
• 'correlation' — One minus the sample linear correlation
between observations (treated as sequences of values).
• 'spearman' — One minus the sample Spearman’s rank
correlation between observations (treated as sequences of
values).
• 'hamming' — Hamming distance, which is the percentage of
coordinates that differ.
• 'jaccard' — One minus the Jaccard coefficient, which is the
percentage of nonzero coordinates that differ.
• custom distance function — A distance function specified using
@ (for example, @distfun). A distance function must be of the
form function D2 = distfun(ZI, ZJ), taking as arguments
a 1-by-n vector ZI containing a single row of X or Y, an m2-by-n
matrix ZJ containing multiple rows of X or Y, and returning
an m2-by-1 vector of distances D2, whose jth element is the
distance between the observations ZI and ZJ(j,:).
For more information on these distance metrics, see “Distance
Metrics” on page 12-14.
P
A positive scalar, p, indicating the exponent of the Minkowski
distance. This parameter is only valid if the Distance is
'minkowski'. Default is 2.
Cov
A positive definite matrix indicating the covariance matrix when
computing the Mahalanobis distance. This parameter is only
valid when Distance is 'mahalanobis'. Default is nancov(X).
Scale
18-655
knnsearch
A vector S containing nonnegative values, with length equal to
the number of columns in X. Each coordinate of X and each query
point is scaled by the corresponding element of S when computing
the standardized Euclidean distance. This argument is only valid
when Distance is 'seuclidean'. Default is nanstd(X).
BucketSize
The maximum number of data points in the leaf node of the
kd-tree. This argument is only meaningful when using the kd-tree
search method. Default is 50.
Examples Find the 10 nearest neighbors in x to each point in y using first the
'minkowski' distance metric with a p value of 5, and then using the
'chebychev' distance metric. Visually compare the results:
load fisheriris
x = meas(:,3:4);
y = [5 1.45;6 2;2.75 .75];
% Perform a knnsearch between x and the query points in y,
% using first Minkowski then Chebychev distance metrics.
[n,d]=knnsearch(x,y,'k',10,'distance','minkowski','p',5);
[ncb,dcb] = knnsearch(x,y,'k',10,...
'distance','chebychev');
% Visualize the results of the two different nearest
% neighbors searches.
% First plot the training data.
gscatter(x(:,1),x(:,2),species)
% Plot an X for the query points.
line(y(:,1),y(:,2),'marker','x','color','k',...
'markersize',10,'linewidth',2,'linestyle','none')
% Use circles to denote the Minkowski nearest neighbors.
line(x(n,1),x(n,2),'color',[.5 .5 .5],'marker','o',...
18-656
knnsearch
'linestyle','none','markersize',10)
% Use pentagrams to denote the Chebychev nearest neighbors.
line(x(ncb,1),x(ncb,2),'color',[.5 .5 .5],'marker','p',...
'linestyle','none','markersize',10)
legend('setosa','versicolor','virginica','query point',...
'minkowski','chebychev')
set(legend,'location','best')
Algorithm For information on a specific search algorithm, see “Distance Metrics”
on page 12-14.
18-657
knnsearch
References [1] Friedman, J. H., Bentely, J., and Finkel, R. A. (1977) An Algorithm
for Finding Best Matches in Logarithmic Expected Time, ACM
Transactions on Mathematical Software 3, 209.
See Also createns | KDTreeSearcher.knnsearch |
ExhaustiveSearcher.knnsearch
How To • “k-Nearest Neighbor Search” on page 12-17
18-658
kruskalwallis
Purpose Kruskal-Wallis test
Syntax p = kruskalwallis(X)
p = kruskalwallis(X,group)
p = kruskalwallis(X,group,displayopt)
[p,table] = kruskalwallis(...)
[p,table,stats] = kruskalwallis(...)
Description p = kruskalwallis(X) performs a Kruskal-Wallis test to compare
samples from two or more groups. Each column of the m-by-n matrix X
represents an independent sample containing m mutually independent
observations. The function compares the medians of the samples in
X, and returns the p value for the null hypothesis that all samples
are drawn from the same population (or equivalently, from different
populations with the same distribution). Note that the Kruskal-Wallis
test is a nonparametric version of the classical one-way ANOVA, and an
extension of the Wilcoxon rank sum test to more than two groups.
If the p value is near zero, this casts doubt on the null hypothesis and
suggests that at least one sample median is significantly different from
the others. The choice of a critical p value to determine whether the
result is judged statistically significant is left to the researcher. It is
common to declare a result significant if the p value is less than 0.05
or 0.01.
The kruskalwallis function displays two figures. The first figure is a
standard ANOVA table, calculated using the ranks of the data rather
than their numeric values. Ranks are found by ordering the data from
smallest to largest across all groups, and taking the numeric index of
this ordering. The rank for a tied observation is equal to the average
rank of all observations tied with it. For example, the following table
shows the ranks for a small sample.
X value 1.4 2.7 1.6 1.6 3.3 0.9 1.1
Rank 3 6 4.5 4.5 7 1 2
18-659
kruskalwallis
The entries in the ANOVA table are the usual sums of squares, degrees
of freedom, and other quantities calculated on the ranks. The usual F
statistic is replaced by a chi-square statistic. The p value measures the
significance of the chi-square statistic.
The second figure displays box plots of each column of X (not the ranks
of X).
p = kruskalwallis(X,group) uses the values in group (a character
array or cell array) as labels for the box plot of the samples in X, when
X is a matrix. Each row of group contains the label for the data in the
corresponding column of X, so group must have length equal to the
number of columns in X. (See “Grouped Data” on page 2-34.)
When X is a vector, kruskalwallis performs a Kruskal-Wallis test on
the samples contained in X, as indexed by input group (a categorical
variable, vector, character array, or cell array). Each element in group
identifies the group (i.e., sample) to which the corresponding element in
vector X belongs, so group must have the same length as X. The labels
contained in group are also used to annotate the box plot.
It is not necessary to label samples sequentially (1, 2, 3, ...). For
example, if X contains measurements taken at three different
temperatures, -27°, 65°, and 110°, you could use these numbers as the
sample labels in group. If a row of group contains an empty cell or
empty string, that row and the corresponding observation in X are
disregarded. NaNs in either input are similarly ignored.
p = kruskalwallis(X,group,displayopt) enables the table and box
plot displays when displayopt is 'on' (default) and suppresses the
displays when displayopt is 'off'.
[p,table] = kruskalwallis(...) returns the ANOVA table
(including column and row labels) in cell array table.
[p,table,stats] = kruskalwallis(...) returns a stats structure
that you can use to perform a follow-up multiple comparison test. The
kruskalwallis test evaluates the hypothesis that all samples come
from populations that have the same median, against the alternative
that the medians are not all the same. Sometimes it is preferable to
18-660
kruskalwallis
perform a test to determine which pairs are significantly different, and
which are not. You can use the multcompare function to perform such
tests by supplying the stats structure as input.
Assumptions
The Kruskal-Wallis test makes the following assumptions about the
data in X:
• All samples come from populations having the same continuous
distribution, apart from possibly different locations due to group
effects.
• All observations are mutually independent.
The classical one-way ANOVA test replaces the first assumption with
the stronger assumption that the populations have normal distributions.
Examples This example compares the material strength study used with the
anova1 function, to see if the nonparametric Kruskal-Wallis procedure
leads to the same conclusion. The example studies the strength of
beams made from three alloys:
strength = [82 86 79 83 84 85 86 87 74 82 ...
78 75 76 77 79 79 77 78 82 79];
alloy = {'st','st','st','st','st','st','st','st',...
'al1','al1','al1','al1','al1','al1',...
'al2','al2','al2','al2','al2','al2'};
This example uses both classical and Kruskal-Wallis ANOVA, omitting
displays:
anova1(strength,alloy,'off')
ans =
1.5264e-004
kruskalwallis(strength,alloy,'off')
ans =
18-661
kruskalwallis
0.0018
Both tests find that the three alloys are significantly different, though
the result is less significant according to the Kruskal-Wallis test. It
is typical that when a data set has a reasonable fit to the normal
distribution, the classical ANOVA test is more sensitive to differences
between groups.
To understand when a nonparametric test may be more appropriate,
let’s see how the tests behave when the distribution is not normal. You
can simulate this by replacing one of the values by an extreme value
(an outlier).
strength(20)=120;
anova1(strength,alloy,'off')
ans =
0.2501
kruskalwallis(strength,alloy,'off')
ans =
0.0060
Now the classical ANOVA test does not find a significant difference, but
the nonparametric procedure does. This illustrates one of the properties
of nonparametric procedures - they are often not severely affected by
changes in a small portion of the data.
References [1] Gibbons, J. D. Nonparametric Statistical Inference. New York:
Marcel Dekker, 1985.
[2] Hollander, M., and D. A. Wolfe. Nonparametric Statistical Methods.
Hoboken, NJ: John Wiley & Sons, Inc., 1999.
See Also “Grouped Data” on page 2-34
anova1, boxplot, friedman, multcompare, ranksum
18-662
ksdensity
Purpose Kernel smoothing density estimate
Syntax [f,xi] = ksdensity(x)
f = ksdensity(x,xi)
ksdensity(...)
ksdensity(ax,...)
[f,xi,u] = ksdensity(...)
[...] = ksdensity(...,param1,val1,param2,val2,...)
Description [f,xi] = ksdensity(x) computes a probability density estimate of
the sample in the vector x. f is the vector of density values evaluated
at the points in xi. The estimate is based on a normal kernel function,
using a window parameter ('width') that is a function of the number of
points in x. The density is evaluated at 100 equally spaced points that
cover the range of the data in x.
f = ksdensity(x,xi) specifies the vector xi of values, where the
density estimate is to be evaluated.
ksdensity(...) without output arguments produces a plot of the
results.
ksdensity(ax,...) plots into axes ax instead of gca.
[f,xi,u] = ksdensity(...) also returns the width of the
kernel-smoothing window.
[...] = ksdensity(...,param1,val1,param2,val2,...) specifies
parameter/value pairs to control the density estimation. Valid
parameter strings and their possible values are as follows:
18-663
ksdensity
Parameter Value
'censoring' A logical vector of the same length as x, indicating
which entries are censoring times. Default is no
censoring.
'kernel' The type of kernel smoother to use. Choose the
value as 'normal' (default), 'box', 'triangle', or
'epanechnikov'.
Alternatively, you can specify some other function,
as a function handle or as a string, e.g., @normpdf
or 'normpdf'. The function must take a single
argument that is an array of distances between data
values and places where the density is evaluated. It
must return an array of the same size containing
corresponding values of the kernel function.
'npoints' The number of equally spaced points in xi. Default
is 100.
'support' • 'unbounded' allows the density to extend over the
whole real line (default).
• 'positive' restricts the density to positive
values.
• A two-element vector gives finite lower and upper
bounds for the support of the density.
'weights' Vector of the same length as x, assigning weight to
each x value.
18-664
ksdensity
Parameter Value
'width' The bandwidth of the kernel-smoothing window. The
default is optimal for estimating normal densities,
but you may want to choose a smaller value to reveal
features such as multiple modes.
'function' The function type to estimate, chosen from among
'pdf', 'cdf', 'icdf', 'survivor', or 'cumhazard'
for the density, cumulative probability, inverse
cumulative probability, survivor, or cumulative
hazard functions, respectively.
For 'icdf',
f=ksdensity(x,yi,...,'function','icdf')
computes the estimated inverse CDF of the values
in x, and evaluates it at the probability values
specified in yi.
In place of the kernel functions listed above, you can specify another
kernel function by using @ (such as @normpdf) or quotes (such as
'normpdf'). ksdensity calls the function with a single argument that
is an array containing distances between data values in x and locations
in xi where the density is evaluated. The function must return an array
of the same size containing corresponding values of the kernel function.
When the 'function' parameter value is 'pdf', this kernel function
returns density values, otherwise it returns cumulative probability
values. Specifying a custom kernel when the 'function' parameter
value is 'icdf' returns an error.
If the 'support' parameter is 'positive', ksdensity transforms x
using a log function, estimates the density of the transformed values,
and transforms back to the original scale. If 'support' is a vector
[L U], ksdensity uses the transformation log((X-L)/(U-X)). The
'width' parameter and u outputs are on the scale of the transformed
values.
Examples Generate a mixture of two normal distributions and plots the estimated
density:
18-665
ksdensity
x = [randn(30,1); 5+randn(30,1)];
[f,xi] = ksdensity(x);
plot(xi,f);
Generate a mixture of two normal distributions, and plot the estimated
cumulative distribution at a specified set of values:
x = [randn(30,1); 5+randn(30,1)];
xi = linspace(-10,15,201);
f = ksdensity(x,xi,'function','cdf');
plot(xi,f);
18-666
ksdensity
Generate a mixture of two normal distributions, and plot the estimated
inverse cumulative distribution function at a specified set of values:
x = [randn(30,1); 5+randn(30,1)];
yi = linspace(.01,.99,99);
g = ksdensity(x,yi,'function','icdf');
plot(yi,g);
18-667
ksdensity
References [1] Bowman, A. W., and A. Azzalini. Applied Smoothing Techniques for
Data Analysis. New York: Oxford University Press, 1997.
See Also hist, @ (function handle)
18-668
kstest
Purpose One-sample Kolmogorov-Smirnov test
Syntax h = kstest(x)
h = kstest(x,CDF)
h = kstest(x,CDF,alpha)
h = kstest(x,CDF,alpha,type)
[h,p,ksstat,cv] = kstest(...)
Description h = kstest(x) performs a Kolmogorov-Smirnov test to compare the
values in the data vector x to a standard normal distribution. The null
hypothesis is that x has a standard normal distribution. The alternative
hypothesis is that x does not have that distribution. The result h is
1 if the test rejects the null hypothesis at the 5% significance level,
0 otherwise.
The test statistic is:
max(| F ( x) − G( x)|)
where F(x) is the empirical cdf and G(x) is the standard normal cdf.
h = kstest(x,CDF) compares the distribution of x to the hypothesized
continuous distribution defined by CDF, which is either a two-column
matrix or a ProbDist object of the ProbDistUnivParam class or
ProbDistUnivKernel class. When CDF is a matrix, column 1 contains
a set of possible x values, and column 2 contains the corresponding
hypothesized cumulative distribution function values G(x). If possible,
define CDF so that column 1 contains the values in x. If there are
values in x not found in column 1 of CDF, kstest approximates G(x)
by interpolation. All values in x must lie in the interval between the
smallest and largest values in the first column of CDF. If the second
argument is empty ([]), kstest uses the standard normal distribution.
The Kolmogorov-Smirnov test requires that CDF be predetermined. It
is not accurate if CDF is estimated from the data. To test x against a
normal distribution without specifying the parameters, use lillietest
instead.
18-669
kstest
h = kstest(x,CDF,alpha) specifies the significance level alpha for the
test. The default is 0.05.
h = kstest(x,CDF,alpha,type) specifies the type of test using one of
the following values for the string type:
• 'unequal' — Tests the alternative hypothesis that the population
cdf is unequal to the specified CDF. This is the default.
• 'larger' — Tests the alternative hypothesis that the population cdf
is larger than the specified CDF. The test statistic does not use the
absolute value.
• 'smaller' — Tests the alternative hypothesis that the population
cdf is smaller than the specified CDF. The test statistic does not use
the absolute value.
[h,p,ksstat,cv] = kstest(...) also returns the p value p, the test
statistic ksstat, and the cutoff value cv for determining if ksstat is
significant.
Examples Generate evenly spaced numbers and perform a Kolmogorov-Smirnov
test to see if they come from a standard normal distribution:
x = -2:1:4
x =
-2 -1 0 1 2 3 4
[h,p,k,c] = kstest(x,[],0.05,0)
h =
0
p =
0.13632
k =
0.41277
c =
0.48342
18-670
kstest
The test fails to reject the null hypothesis that the values come from a
standard normal distribution. This illustrates the difficulty of testing
normality in small samples. (The Lilliefors test, implemented by the
Statistics Toolbox function lillietest, may be more appropriate.)
The following figure illustrates the test statistic:
xx = -3:.1:5;
F = cdfplot(x);
hold on
G = plot(xx,normcdf(xx),'r-');
set(F,'LineWidth',2)
set(G,'LineWidth',2)
legend([F G],...
'Empirical','Standard Normal',...
'Location','NW')
18-671
kstest
The test statistic k is the maximum difference between the curves.
Setting type to 'smaller' tests the alternative that the population
cdf is smaller than the normal cdf:
[h,p,ksstat] = kstest(x,[],0.05,'smaller')
h =
0
p =
0.068181
k =
0.41277
The test statistic is the same as before, but the p value is smaller.
18-672
kstest
Setting type to 'larger' changes the test statistic:
[h,p,k] = kstest(x,[],0.05,'larger')
h =
0
p =
0.77533
k =
0.12706
References [1] Massey, F. J. “The Kolmogorov-Smirnov Test for Goodness of Fit.”
Journal of the American Statistical Association. Vol. 46, No. 253, 1951,
pp. 68–78.
[2] Miller, L. H. “Table of Percentage Points of Kolmogorov Statistics.”
Journal of the American Statistical Association. Vol. 51, No. 273,
1956, pp. 111–121.
[3] Marsaglia, G., W. Tsang, and J. Wang. “Evaluating Kolmogorov’s
Distribution.” Journal of Statistical Software. Vol. 8, Issue 18, 2003.
See Also kstest2, lillietest
18-673
kstest2
Purpose Two-sample Kolmogorov-Smirnov test
Syntax h = kstest2(x1,x2)
h = kstest2(x1,x2,alpha,type)
[h,p] = kstest2(...)
[h,p,ks2stat] = kstest2(...)
Description h = kstest2(x1,x2) performs a two-sample Kolmogorov-Smirnov test
to compare the distributions of the values in the two data vectors x1 and
x2. The null hypothesis is that x1 and x2 are from the same continuous
distribution. The alternative hypothesis is that they are from different
continuous distributions. The result h is 1 if the test rejects the null
hypothesis at the 5% significance level; 0 otherwise.
The test statistic is:
max(| F1( x) − F 2( x)|)
where F1(x) is the proportion of x1 values less than or equal to x and
F1(x) is the proportion of x2 values less than or equal to x.
h = kstest2(x1,x2,alpha) specifies the significance level alpha for
the test. The default is 0.05.
h = kstest2(x1,x2,alpha,type) specifies the type of test using one of
the following values for the string type:
• 'unequal' — Tests the alternative hypothesis that the population
cdfs are unequal. This is the default.
• 'larger' — Tests the alternative hypothesis that the first population
cdf is larger than the second population cdf. The test statistic does
not use the absolute value.
• 'smaller' — Tests the alternative hypothesis that the first
population cdf is smaller than the second population cdf. The test
statistic does not use the absolute value.
18-674
kstest2
[h,p] = kstest2(...) also returns the asymptotic p value p. The
asymptotic p value becomes very accurate for large sample sizes, and
is believed to be reasonably accurate for sample sizes n1 and n2 such
that (n1*n2)/(n1 + n2) >= 4.
[h,p,ks2stat] = kstest2(...) also returns the p value p and the
test statistic ks2stat.
Examples The following test compares the distributions of a small evenly-spaced
sample and a larger normal sample:
x = -1:1:5
y = randn(20,1);
[h,p,k] = kstest2(x,y)
h =
0
p =
0.0774
k =
0.5214
The following figure illustrates the test statistic:
F1 = cdfplot(x);
hold on
F2 = cdfplot(y)
set(F1,'LineWidth',2,'Color','r')
set(F2,'LineWidth',2)
legend([F1 F2],'F1(x)','F2(x)','Location','NW')
18-675
kstest2
The test statistic k is the maximum difference between the curves.
References [1] Massey, F. J. “The Kolmogorov-Smirnov Test for Goodness of Fit.”
Journal of the American Statistical Association. Vol. 46, No. 253, 1951,
pp. 68–78.
[2] Miller, L. H. “Table of Percentage Points of Kolmogorov Statistics.”
Journal of the American Statistical Association. Vol. 51, No. 273,
1956, pp. 111–121.
[3] Marsaglia, G., W. Tsang, and J. Wang. “Evaluating Kolmogorov’s
Distribution.” Journal of Statistical Software. Vol. 8, Issue 18, 2003.
18-676
kstest2
[4] Stephens, M. A. “Use of the Kolmogorov-Smirnov, Cramer-Von
Mises and Related Statistics Without Extensive Tables.” Journal of the
Royal Statistical Society. Series B, Vol. 32, No. 1, 1970, pp. 115–122.
See Also kstest, lillietest
18-677
kurtosis
Purpose Kurtosis
Syntax k = kurtosis(X)
k = kurtosis(X,flag)
k = kurtosis(X,flag,dim)
Description k = kurtosis(X) returns the sample kurtosis of X. For vectors,
kurtosis(x) is the kurtosis of the elements in the vector x. For
matrices kurtosis(X) returns the sample kurtosis for each column
of X. For N-dimensional arrays, kurtosis operates along the first
nonsingleton dimension of X.
k = kurtosis(X,flag) specifies whether to correct for bias (flag is
0) or not (flag is 1, the default). When X represents a sample from a
population, the kurtosis of X is biased, that is, it will tend to differ from
the population kurtosis by a systematic amount that depends on the size
of the sample. You can set flag to 0 to correct for this systematic bias.
k = kurtosis(X,flag,dim) takes the kurtosis along dimension dim
of X.
kurtosis treats NaNs as missing values and removes them.
Algorithm Kurtosis is a measure of how outlier-prone a distribution is. The
kurtosis of the normal distribution is 3. Distributions that are more
outlier-prone than the normal distribution have kurtosis greater than
3; distributions that are less outlier-prone have kurtosis less than 3.
The kurtosis of a distribution is defined as
E( x − )4
k=
4
where μ is the mean of x, σ is the standard deviation of x, and E(t)
represents the expected value of the quantity t. kurtosis computes a
sample version of this population value.
18-678
kurtosis
Note Some definitions of kurtosis subtract 3 from the computed value,
so that the normal distribution has kurtosis of 0. The kurtosis function
does not use this convention.
When you set flag to 1, the following equation applies:
1 n
∑(x − x)
4
n i=1 i
k1 =
2
⎛1 n 2⎞
⎜ ∑ ( xi − x ) ⎟
⎜n ⎟
⎝ i=1 ⎠
When you set flag to 0, the following equation applies:
n −1
k0 =
( n − 2) ( n − 3)
( ( n + 1) k1 − 3 ( n − 1) ) + 3
This bias-corrected formula requires that X contain at least four
elements.
Examples X = randn([5 4])
X =
1.1650 1.6961 -1.4462 -0.3600
0.6268 0.0591 -0.7012 -0.1356
0.0751 1.7971 1.2460 -1.3493
0.3516 0.2641 -0.6390 -1.2704
-0.6965 0.8717 0.5774 0.9846
k = kurtosis(X)
k =
2.1658 1.2967 1.6378 1.9589
See Also mean, moment, skewness, std, var
18-679
categorical.labels property
Purpose Text labels for levels
Description Text labels for levels. Access labels with getlabels.
18-680
qrandstream.le
Purpose Less than or equal relation for handles
Syntax h1 <= h2
Description Handles are equal if they are handles for the same object. All
comparisons use a number associated with each handle object. Nothing
can be assumed about the result of a handle comparison except that the
repeated comparison of two handles in the same MATLAB session will
yield the same result. The order of handle values is purely arbitrary
and has no connection to the state of the handle objects being compared.
h1 <= h2 performs element-wise comparisons between handle arrays
h1 and h2. h1 and h2 must be of the same dimensions unless one is a
scalar. The result is a logical array of the same dimensions, where each
element is an element-wise <= result.
If one of h1 or h2 is scalar, scalar expansion is performed and the result
will match the dimensions of the array that is not scalar.
tf = le(h1, h2) stores the result in a logical array of the same
dimensions.
See Also qrandstream, eq, ge, gt, lt, ne
18-681
qrandset.Leap property
Purpose Interval between points
Description Number of points to leap over and omit for each point taken from the
sequence. The Leap property of a point set contains a positive integer
which specifies the number of points in the sequence to leap over
and omit for every point taken. The default Leap value is 0, which
corresponds to taking every point from the sequence.
Leaping is a technique used to improve the quality of a point set.
However, you must choose the Leap values with care; many Leap values
create sequences that fail to touch on large sub-hyper-rectangles of the
unit hypercube, and so fail to be a uniform quasi-random point set.
Choosing Leap Values for Halton Sets
A known rule for choosing Leap values for Halton sets is to set it to
(P-1) where P is a prime number that has not been used to generate one
of the dimensions, i.e. for a k-dimensional point set P would be the
(k+1)th or greater prime.
Examples Experiment with different leap values:
% No leaping produces the standard Halton sequence.
P = haltonset(5);
P(1:5,:)
% Set a leap of 1. The point set now includes every other
% point from the sequence.
P.Leap = 1;
P(1:5,:)
See Also net | qrandset | Skip | subsref | haltonset
18-682
dataset.length
Purpose Length of dataset array
Syntax n = length(A)
Description n = length(A) returns the number of observations in the dataset A.
length is equivalent to size(A,1).
See Also size
18-683
qrandset.length
Purpose Length of point set
Syntax length(p)
Description length(p) returns the number of points in the point set p. It is
equivalent to size(p, 1).
See Also qrandset, size
18-684
categorical.length
Purpose Length of categorical array
Syntax n = length(A)
Description n = length(A) returns the size of the longest dimension of the
categorical array A when A is not empty. If A is a vector, this is the
same as its length. length is equivalent to max(size(x)) for nonempty
arrays, and 0 for empty arrays.
See Also isempty, isscalar, size
18-685
categorical.levelcounts
Purpose Element counts by level
Syntax C = levelcounts(A)
C = levelcounts(A,dim)
Description C = levelcounts(A) for a categorical vector A counts the number of
elements in A equal to each of the possible levels in A. The output is a
vector C containing those counts, and has as many elements as A has
levels. For matrix A, C is a matrix of column counts. For N-dimensional
arrays, levelcounts operates along the first nonsingleton dimension.
C = levelcounts(A,dim) operates along the dimension dim.
Examples Count the number of patients in each age group in the data in
hospital.mat:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
disp(labels')
'0s' '10s' '20s' '30s' '40s' '50s' '60s' '70s' '80s' '90s'
AgeGroup = ordinal(hospital.Age,labels,[],edges);
I = islevel(labels,AgeGroup);
disp(I')
0 1 1 1 1 1 1 1 1 1
c = levelcounts(AgeGroup);
disp(c')
0 0 15 41 42 2 0 0 0 0
AgeGroup = droplevels(AgeGroup);
I = islevel(labels,AgeGroup);
disp(I')
0 0 1 1 1 1 0 0 0 0
c = levelcounts(AgeGroup);
disp(c')
15 41 42 2
18-686
categorical.levelcounts
See Also islevel, ismember, summary
18-687
leverage
Purpose Leverage
Syntax h = leverage(data)
h = leverage(data,model)
Description h = leverage(data) finds the leverage of each row (point) in the
matrix data for a linear additive regression model.
h = leverage(data,model) finds the leverage on a regression, using a
specified model type, where model can be one of these strings:
• 'linear' - includes constant and linear terms
• 'interaction' - includes constant, linear, and cross product terms
• 'quadratic' - includes interactions and squared terms
• 'purequadratic' - includes constant, linear, and squared terms
Leverage is a measure of the influence of a given observation on a
regression due to its location in the space of the inputs.
Algorithm [Q,R] = qr(x2fx(data,'model'));
leverage = (sum(Q'.*Q'))'
Examples One rule of thumb is to compare the leverage to 2p/n where n is the
number of observations and p is the number of parameters in the model.
For the Hald data set this value is 0.7692.
load hald
h = max(leverage(ingredients,'linear'))
h =
0.7004
Since 0.7004 < 0.7692, there are no high leverage points using this rule.
18-688
leverage
References [1] Goodall, C. R. “Computation Using the QR Decomposition.”
Handbook in Statistics. Vol. 9, Amsterdam: Elsevier/North-Holland,
1993.
See Also regstats
18-689
lhsdesign
Purpose Latin hypercube sample
Syntax X = lhsdesign(n,p)
X = lhsdesign(...,'smooth','off')
X = lhsdesign(...,'criterion',criterion)
X = lhsdesign(...,'iterations',k)
Description X = lhsdesign(n,p) generates a latin hypercube sample X containing
n values on each of p variables. For each column, the n values are
randomly distributed with one from each interval (0,1/n), (1/n,2/n),
..., (1-1/n,1), and they are randomly permuted.
X = lhsdesign(...,'smooth','off') produces points at the
midpoints of the above intervals: 0.5/n, 1.5/n, ..., 1-0.5/n. The
default is 'on'.
X = lhsdesign(...,'criterion',criterion) iteratively generates
latin hypercube samples to find the best one according to the criterion
criterion, which can be one of the following strings.
Criterion Description
'none' No iteration
'maximin' Maximize minimum distance between points
'correlation' Reduce correlation
X = lhsdesign(...,'iterations',k) iterates up to k times in an
attempt to improve the design according to the specified criterion. The
default is k = 5.
See Also haltonset, sobolset, lhsnorm, unifrnd
18-690
lhsnorm
Purpose Latin hypercube sample from normal distribution
Syntax X = lhsnorm(mu,sigma,n)
X = lhsnorm(mu,sigma,n,flag)
[X,Z] = lhsnorm(...)
Description X = lhsnorm(mu,sigma,n) generates a latin hypercube sample X of
size n from the multivariate normal distribution with mean vector mu
and covariance matrix sigma. X is similar to a random sample from the
multivariate normal distribution, but the marginal distribution of each
column is adjusted so that its sample marginal distribution is close to
its theoretical normal distribution.
X = lhsnorm(mu,sigma,n,flag) controls the amount of smoothing in
the sample. If flag is 'off', each column has points equally spaced
on the probability scale. In other words, each column is a permutation
of the values G(0.5/n), G(1.5/n), ..., G(1-0.5/n) where G is
the inverse normal cumulative distribution for that column’s marginal
distribution. If flag is 'on' (the default), each column has points
uniformly distributed on the probability scale. For example, in place
of 0.5/n you use a value having a uniform distribution on the interval
(0/n,1/n).
[X,Z] = lhsnorm(...) also returns Z, the original multivariate
normal sample before the marginals are adjusted to obtain X.
References [1] Stein, M. “Large sample properties of simulations using latin
hypercube sampling.” Technometrics. Vol. 29, No. 2, 1987, pp. 143–151.
Correction, Vol. 32, p. 367.
See Also lhsdesign, mvnrnd
18-691
lillietest
Purpose Lilliefors test
Syntax h = lillietest(x)
h = lillietest(x,alpha)
h = lillietest(x,alpha,distr)
[h,p] = lillietest(...)
[h,p,kstat] = lillietest(...)
[h,p,kstat,critval] = lillietest(...)
[h,p,...] = lillietest(x,alpha,distr,mctol)
Description h = lillietest(x) performs a Lilliefors test of the default null
hypothesis that the sample in vector x comes from a distribution in the
normal family, against the alternative that it does not come from a
normal distribution. The test returns the logical value h = 1 if it rejects
the null hypothesis at the 5% significance level, and h = 0 if it cannot.
The test treats NaN values in x as missing values, and ignores them.
The Lilliefors test is a 2-sided goodness-of-fit test suitable when a
fully-specified null distribution is unknown and its parameters must be
estimated. This is in contrast to the one-sample Kolmogorov-Smirnov
test (see kstest), which requires that the null distribution be
completely specified. The Lilliefors test statistic is the same as for the
Kolmogorov-Smirnov test:
KS = max SCDF ( x) − CDF ( x)
x
where SCDF is the empirical cdf estimated from the sample and CDF is
the normal cdf with mean and standard deviation equal to the mean
and standard deviation of the sample.
lillietest uses a table of critical values computed using Monte Carlo
simulation for sample sizes less than 1000 and significance levels
between 0.001 and 0.50. The table is larger and more accurate than the
table introduced by Lilliefors. Critical values for a test are computed
by interpolating into the table, using an analytic approximation when
extrapolating for larger sample sizes.
18-692
lillietest
h = lillietest(x,alpha) performs the test at significance level
alpha. alpha is a scalar in the range [0.001, 0.50]. To perform the test at
a significance level outside of this range, use the mctol input argument.
h = lillietest(x,alpha,distr) performs the test of the null
hypothesis that x came from the location-scale family of distributions
specified by distr. Acceptable values for distr are 'norm' (normal, the
default), 'exp' (exponential), and 'ev' (extreme value). The Lilliefors
test can not be used when the null hypothesis is not a location-scale
family of distributions.
[h,p] = lillietest(...) returns the p value p, computed using
inverse interpolation into the table of critical values. Small values of
p cast doubt on the validity of the null hypothesis. lillietest warns
when p is not found within the tabulated range of [0.001, 0.50], and
returns either the smallest or largest tabulated value. In this case, you
can use the mctol input argument to compute a more accurate p value.
[h,p,kstat] = lillietest(...) returns the test statistic kstat.
[h,p,kstat,critval] = lillietest(...) returns the critical value
critval for the test. When kstat > critval, the null hypothesis is
rejected at significance level alpha
[h,p,...] = lillietest(x,alpha,distr,mctol) computes a Monte
Carlo approximation for p directly, rather than interpolating into the
table of pre-computed values. This is useful when alpha or p lie outside
the range of the table. lillietest chooses the number of Monte Carlo
replications, mcreps, large enough to make the Monte Carlo standard
error for p, sqrt(p*(1-p)/mcreps), less than mctol.
Examples Use lillietest to determine if car mileage, in miles per gallon (MPG),
follows a normal distribution across different makes of cars:
load carbig.mat
[h,p] = lillietest(MPG)
Warning: P is less than the smallest tabulated value, returning 0.001.
h =
1
18-693
lillietest
p =
1.0000e-003
This is clear evidence for rejecting the null hypothesis of normality,
but the p value returned is just the smallest value in the table of
pre-computed values. To find a more accurate p value for the test, run a
Monte Carlo approximation using the mctol input argument:
[h,p] = lillietest(MPG,0.05,'norm',1e-4)
h =
1
p =
8.3333e-006
References [1] Conover, W. J. Practical Nonparametric Statistics. Hoboken, NJ:
John Wiley & Sons, Inc., 1980.
[2] Lilliefors, H. W. “On the Kolmogorov-Smirnov test for the
exponential distribution with mean unknown.” Journal of the American
Statistical Association. Vol. 64, 1969, pp. 387–389.
[3] Lilliefors, H. W. “On the Kolmogorov-Smirnov test for normality
with mean and variance unknown.” Journal of the American Statistical
Association. Vol. 62, 1967, pp. 399–402.
See Also jbtest, kstest, kstest2, cdfplot
18-694
linhyptest
Purpose Linear hypothesis test
Syntax p = linhyptest(beta,COVB,c,H,dfe)
[p,t,r] = linhyptest(...)
Description p = linhyptest(beta,COVB,c,H,dfe) returns the p value p of
a hypothesis test on a vector of parameters. beta is a vector of k
parameter estimates. COVB is the k-by-k estimated covariance matrix
of the parameter estimates. c and H specify the null hypothesis in the
form H*b = c, where b is the vector of unknown parameters estimated
by beta. dfe is the degrees of freedom for the COVB estimate, or Inf if
COVB is known rather than estimated.
beta is required. The remaining arguments have default values:
• COVB = eye(k)
• c = zeros(k,1)
• H = eye(K)
• dfe = Inf
If H is omitted, c must have k elements and it specifies the null
hypothesis values for the entire parameter vector.
Note The following functions return outputs suitable for use as the
COVB input argument to linhyptest: nlinfit, coxphfit, glmfit,
mnrfit, regstats, robustfit. nlinfit returns COVB directly; the other
functions return COVB in stats.covb.
[p,t,r] = linhyptest(...) also returns the test statistic t and the
rank r of the hypothesis matrix H. If dfe is Inf or is not given, t*r is a
chi-square statistic with r degrees of freedom . If dfe is specified as a
finite value, t is an F statistic with r and dfe degrees of freedom.
18-695
linhyptest
linhyptest performs a test based on an asymptotic normal distribution
for the parameter estimates. It can be used after any estimation
procedure for which the parameter covariances are available, such as
regstats or glmfit. For linear regression, the p-values are exact.
For other procedures, the p-values are approximate, and may be less
accurate than other procedures such as those based on a likelihood ratio.
Examples Fit a multiple linear model to the data in hald.mat:
load hald
stats = regstats(heat,ingredients,'linear');
beta = stats.beta
beta =
62.4054
1.5511
0.5102
0.1019
-0.1441
Perform an F-test that the last two coefficients are both 0:
SIGMA = stats.covb;
dfe = stats.fstat.dfe;
H = [0 0 0 1 0;0 0 0 0 1];
c = [0;0];
[p,F] = linhyptest(beta,SIGMA,c,H,dfe)
p =
0.4668
F =
0.8391
See Also regstats | glmfit | robustfit | mnrfit | nlinfit | coxphfit
18-696
linkage
Purpose Create agglomerative hierarchical cluster tree
Syntax Z = linkage(y)
Z = linkage(y,method)
Z = linkage(X,method,metric)
Z = linkage(X,method,inputs)
Description Z = linkage(y) creates an agglomerative hierarchical cluster tree from
the distances in y. y is a Euclidean distance matrix or a more general
dissimilarity matrix, formatted as a vector, as returned by pdist.
Z is a (m-1)-by-3 matrix, where m is the number of observations in the
original data. Columns 1 and 2 of Z contain cluster indices linked in
pairs to form a binary tree. The leaf nodes are numbered from 1 to m.
Leaf nodes are the singleton clusters from which all higher clusters
are built. Each newly-formed cluster, corresponding to row Z(I,:),
is assigned the index m+I. Z(I,1:2) contains the indices of the two
component clusters that form cluster m+I. There are m-1 higher clusters
which correspond to the interior nodes of the clustering tree. Z(I,3)
contains the linkage distances between the two clusters merged in row
Z(I,:).
For example, suppose there are 30 initial nodes and at step 12 cluster 5
and cluster 7 are combined. Suppose their distance at that time is 1.5.
Then Z(12,:) will be [5, 7, 1.5]. The newly formed cluster will have
index 12 + 30 = 42. If cluster 42 appears in a later row, it means the
cluster created at step 12 is being combined into some larger cluster.
Z = linkage(y,method) creates the tree using the specified method.
Methods differ from one another in how they measure the distance
between clusters. Available methods are listed in the following table.
Method Description
'average' Unweighted average distance (UPGMA).
'centroid' Centroid distance (UPGMC). Y must contain
Euclidean distances.
18-697
linkage
Method Description
'complete' Furthest distance.
'median' Weighted center of mass distance (WPGMC). Y must
contain Euclidean distances.
'single' Shortest distance. This is the default.
'ward' Inner squared distance (minimum variance
algorithm). Y must contain Euclidean distances.
'weighted' Weighted average distance (WPGMA).
Note The 'centroid' and 'median' methods can produce a cluster
tree that is not monotonic. This occurs when the distance from the
union of two clusters, r and s, to a third cluster is less than the distance
from either r or s to that third cluster. In this case, sections of the
dendrogram change direction. This is an indication that you should
use another method.
Z = linkage(X,method,metric) creates a hierarchical cluster tree
from the observations in X. Rows in X correspond to observations and
columns to variables. Pairwise distances are computed internally by
calling pdist. metric is one of the distance metrics accepted by pdist.
Z = linkage(X,method,inputs) allows you to pass extra input
arguments to pdist. inputs is a cell array containing input arguments.
Linkages
The following notation is used to describe the linkages used by the
various methods:
• Cluster r is formed from clusters p and q.
• nr is the number of objects in cluster r.
• xri is the ith object in cluster r.
18-698
linkage
• Single linkage, also called nearest neighbor, uses the smallest
distance between objects in the two clusters:
d(r, s) = min(dist( xri , xsj )), i ∈ (i,..., nr ), j ∈ (1,..., ns )
• Complete linkage, also called furthest neighbor, uses the largest
distance between objects in the two clusters:
d(r, s) = max(dist( xri , xsj )), i ∈ (1,..., nr ), j ∈ (1,..., ns )
• Average linkage uses the average distance between all pairs of objects
in any two clusters:
nr ns
1
d(r, s) =
nr ns
∑ ∑ dist( xri , xsj )
i=1 j =1
• Centroid linkage uses the Euclidean distance between the centroids
of the two clusters:
d(r, s) = xr − xs 2
where
nr
1
xr =
nr
∑ xri
i=1
• Median linkage uses the Euclidean distance between weighted
centroids of the two clusters,
d(r, s) = xr − xs 2
where xr and xs are weighted centroids for the clusters r and s. If
cluster r was created by combining clusters p and q, xr is defined
recursively as
18-699
linkage
1
xr = ( x p + xq )
2
• Ward’s linkage uses the incremental sum of squares; that is, the
increase in the total within-cluster sum of squares as a result of
joining two clusters. The within-cluster sum of squares is defined as
the sum of the squares of the distances between all objects in the
cluster and the centroid of the cluster. The equivalent distance is:
2
2
xr − xs 2
d (r, s) = nr ns
(nr + ns )
where 2
is Euclidean distance, and xr and xs are the centroids of
clusters r and s, as defined in the centroid linkage.
Examples X = [3 1.7; 1 1; 2 3; 2 2.5; 1.2 1; 1.1 1.5; 3 1];
Y = pdist(X);
Z = linkage(Y)
Z =
2.0000 5.0000 0.2000
3.0000 4.0000 0.5000
8.0000 6.0000 0.5099
1.0000 7.0000 0.7000
11.0000 9.0000 1.2806
12.0000 10.0000 1.3454
See Also cluster, clusterdata, cophenet, dendrogram, inconsistent, kmeans,
pdist, silhouette, squareform
18-700
logncdf
Purpose Lognormal cumulative distribution function
Syntax P = logncdf(X,mu,sigma)
[P,PLO,PUP] = logncdf(X,mu,sigma,pcov,alpha)
Description P = logncdf(X,mu,sigma) returns values at X of the lognormal
cdf with distribution parameters mu and sigma. mu and sigma are
the mean and standard deviation, respectively, of the associated
normal distribution. X, mu, and sigma can be vectors, matrices, or
multidimensional arrays that all have the same size. A scalar input
for X, mu, or sigma is expanded to a constant array with the same
dimensions as the other inputs.
[P,PLO,PUP] = logncdf(X,mu,sigma,pcov,alpha) returns confidence
bounds for P when the input parameters mu and sigma are estimates.
pcov is the covariance matrix of the estimated parameters. alpha
specifies 100(1 - alpha)% confidence bounds. The default value of alpha
is 0.05. PLO and PUP are arrays of the same size as P containing the
lower and upper confidence bounds.
logncdf computes confidence bounds for P using a normal
approximation to the distribution of the estimate
X − ˆ
ˆ
and then transforming those bounds to the scale of the output P. The
computed bounds give approximately the desired confidence level when
you estimate mu, sigma, and pcov from large samples, but in smaller
samples other methods of computing the confidence bounds might be
more accurate.
The lognormal cdf is
−(ln(t) − )2
1 xe 2 2
p = F ( x| , ) =
2 ∫0 t
dt
18-701
logncdf
Examples x = (0:0.2:10);
y = logncdf(x,0,1);
plot(x,y); grid;
xlabel('x'); ylabel('p');
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 102–105.
See Also cdf, lognpdf, logninv, lognstat, lognfit, lognlike, lognrnd
“Lognormal Distribution” on page B-51
18-702
lognfit
Purpose Lognormal parameter estimates
Syntax parmhat = lognfit(data)
[parmhat,parmci] = lognfit(data)
[parmhat,parmci] = lognfit(data,alpha)
[...] = lognfit(data,alpha,censoring)
[...] = lognfit(data,alpha,censoring,freq)
[...] = lognfit(data,alpha,censoring,freq,options)
Description parmhat = lognfit(data) returns a vector of maximum likelihood
estimates parmhat(1) = mu and parmhat(2) = sigma of parameters
for a lognormal distribution fitting data. mu and sigma are the mean and
standard deviation, respectively, of the associated normal distribution.
[parmhat,parmci] = lognfit(data) returns 95% confidence intervals
for the parameter estimates mu and sigma in the 2-by-2 matrix parmci.
The first column of the matrix contains the lower and upper confidence
bounds for parameter mu, and the second column contains the confidence
bounds for parameter sigma.
[parmhat,parmci] = lognfit(data,alpha) returns 100(1 - alpha) %
confidence intervals for the parameter estimates, where alpha is a value
in the range (0 1) specifying the width of the confidence intervals. By
default, alpha is 0.05, which corresponds to 95% confidence intervals.
[...] = lognfit(data,alpha,censoring) accepts a Boolean vector
censoring, of the same size as data, which is 1 for observations that
are right-censored and 0 for observations that are observed exactly.
[...] = lognfit(data,alpha,censoring,freq) accepts a frequency
vector, freq, of the same size as data. Typically, freq contains integer
frequencies for the corresponding elements in data, but can contain
any nonnegative values. Pass in [] for alpha, censoring, or freq to
use their default values.
[...] = lognfit(data,alpha,censoring,freq,options) accepts a
structure, options, that specifies control parameters for the iterative
algorithm the function uses to compute maximum likelihood estimates
when there is censoring. The lognormal fit function accepts an options
18-703
lognfit
structure which can be created using the function statset. Enter
statset('lognfit') to see the names and default values of the
parameters that lognfit accepts in the options structure. See the
reference page for statset for more information about these options.
Examples This example generates 100 independent samples of lognormally
distributed data with µ = 0 and σ = 3. parmhat estimates µ and σ and
parmci gives 99% confidence intervals around parmhat. Notice that
parmci contains the true values of µ and σ.
data = lognrnd(0,3,100,1);
[parmhat,parmci] = lognfit(data,0.01)
parmhat =
-0.2480 2.8902
parmci =
-1.0071 2.4393
0.5111 3.5262
See Also mle, lognlike, lognpdf, logncdf, logninv, lognstat, lognrnd
“Lognormal Distribution” on page B-51
18-704
logninv
Purpose Lognormal inverse cumulative distribution function
Syntax X = logninv(P,mu,sigma)
[X,XLO,XUP] = logninv(P,mu,sigma,pcov,alpha)
Description X = logninv(P,mu,sigma) returns values at P of the inverse lognormal
cdf with distribution parameters mu and sigma. mu and sigma are
the mean and standard deviation, respectively, of the associated
normal distribution. mu and sigma can be vectors, matrices, or
multidimensional arrays that all have the same size, which is also the
size of X. A scalar input for P, mu, or sigma is expanded to a constant
array with the same dimensions as the other inputs.
[X,XLO,XUP] = logninv(P,mu,sigma,pcov,alpha) returns confidence
bounds for X when the input parameters mu and sigma are estimates.
pcov is the covariance matrix of the estimated parameters. alpha
specifies 100(1 - alpha)% confidence bounds. The default value of alpha
is 0.05. XLO and XUP are arrays of the same size as X containing the
lower and upper confidence bounds.
logninv computes confidence bounds for P using a normal
approximation to the distribution of the estimate
ˆ + ˆ q
where q is the Pth quantile from a normal distribution with mean 0 and
standard deviation 1. The computed bounds give approximately the
desired confidence level when you estimate mu, sigma, and pcov from
large samples, but in smaller samples other methods of computing the
confidence bounds might be more accurate.
The lognormal inverse function is defined in terms of the lognormal
cdf as
x = F −1 ( p| , ) = { x : F ( x | , ) = p}
where
18-705
logninv
−(ln(t) − )2
1 xe 2 2
p = F ( x| , ) =
2 ∫0 t
dt
Examples p = (0.005:0.01:0.995);
crit = logninv(p,1,0.5);
plot(p,crit)
xlabel('Probability'); ylabel('Critical Value'); grid
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
Hoboken, NJ: Wiley-Interscience, 2000. pp. 102–105.
See Also icdf, logncdf, lognpdf, lognstat, lognfit, lognlike, lognrnd
“Lognormal Distribution” on page B-51
18-706
lognlike
Purpose Lognormal negative log-likelihood
Syntax nlogL = lognlike(params,data)
[nlogL,avar] = lognlike(params,data)
[...] = lognlike(params,data,censoring)
[...] = lognlike(params,data,censoring,freq)
Description nlogL = lognlike(params,data) returns the negative log-likelihood of
data for the lognormal distribution with parameters params. params(1)
is the mean of the associated normal distribution, mu, and params(2)
is the standard deviation of the associated normal distribution, sigma.
The values of mu and sigma are scalars, and the output nlogL is a scalar.
[nlogL,avar] = lognlike(params,data) returns the inverse of
Fisher’s information matrix. If the input parameter value in params
is the maximum likelihood estimate, avar is its asymptotic variance.
avar is based on the observed Fisher’s information, not the expected
information.
[...] = lognlike(params,data,censoring) accepts a Boolean
vector, censoring, of the same size as data, which is 1 for observations
that are right-censored and 0 for observations that are observed exactly.
[...] = lognlike(params,data,censoring,freq) accepts a
frequency vector, freq, of the same size as data. The vector freq
typically contains integer frequencies for the corresponding elements in
data, but can contain any nonnegative values. Pass in [] for censoring
to use its default value.
See Also lognfit, lognpdf, logncdf, logninv, lognstat, lognrnd
“Lognormal Distribution” on page B-51
18-707
lognpdf
Purpose Lognormal probability density function
Syntax Y = lognpdf(X,mu,sigma)
Description Y = lognpdf(X,mu,sigma) returns values at X of the lognormal
pdf with distribution parameters mu and sigma. mu and sigma are
the mean and standard deviation, respectively, of the associated
normal distribution. X, mu, and sigma can be vectors, matrices, or
multidimensional arrays that all have the same size, which is also the
size of Y. A scalar input for X, mu, or sigma is expanded to a constant
array with the same dimensions as the other inputs.
The lognormal pdf is
−( ln x − )
2
1 2
y = f ( x| , ) =
2
e
x 2
The normal and lognormal distributions are closely related. If X
is distributed lognormally with parameters µ and σ, then log(X) is
distributed normally with mean µ and standard deviation σ.
The mean m and variance v of a lognormal random variable are
functions of µ and σ that can be calculated with the lognstat function.
They are:
(
m = exp + 2 / 2 )
( )( ( ) )
v = exp 2 + 2 exp 2 − 1
A lognormal distribution with mean m and variance v has parameters
= log ⎛⎜ m2 / v + m2 ⎞⎟
⎝ ⎠
(
= log v / m2 + 1 )
18-708
lognpdf
The lognormal distribution is applicable when the quantity of interest
must be positive, since log(X) exists only when X is positive.
Examples x = (0:0.02:10);
y = lognpdf(x,0,1);
plot(x,y); grid;
xlabel('x'); ylabel('p')
References [1] Mood, A. M., F. A. Graybill, and D. C. Boes. Introduction to the
Theory of Statistics. 3rd ed., New York: McGraw-Hill, 1974. pp.
540–541.
See Also pdf, logncdf, logninv, lognstat, lognfit, lognlike, lognrnd
“Lognormal Distribution” on page B-51
18-709
lognrnd
Purpose Lognormal random numbers
Syntax R = lognrnd(mu,sigma)
R = lognrnd(mu,sigma,v)
R = lognrnd(mu,sigma,m,n)
Description R = lognrnd(mu,sigma) returns an array of random numbers
generated from the lognormal distribution with parameters mu and
sigma. mu and sigma are the mean and standard deviation, respectively,
of the associated normal distribution. mu and sigma can be vectors,
matrices, or multidimensional arrays that have the same size, which
is also the size of R. A scalar input for mu or sigma is expanded to a
constant array with the same dimensions as the other input.
R = lognrnd(mu,sigma,v) returns an array of random numbers
generated from the lognormal distribution with parameters mu and
sigma, where v is a row vector. If v is a 1-by-2 vector, R is a matrix with
v(1) rows and v(2) columns. If v is 1-by-n, R is an n-dimensional array.
R = lognrnd(mu,sigma,m,n) returns an array of random numbers
generated from the lognormal distribution with parameters mu and
sigma, where scalars m and n are the row and column dimensions of R.
The normal and lognormal distributions are closely related. If X
is distributed lognormally with parameters µ and σ, then log(X) is
distributed normally with mean µ and standard deviation σ.
The mean m and variance v of a lognormal random variable are
functions of µ and σ that can be calculated with the lognstat function.
They are:
(
m = exp + 2 / 2 )
( )( ( ) )
v = exp 2 + 2 exp 2 − 1
A lognormal distribution with mean m and variance v has parameters
18-710
lognrnd
= log ⎛⎜ m2 / v + m2 ⎞⎟
⎝ ⎠
(
= log v / m2 + 1 )
Examples Generate one million lognormally distributed random numbers with
mean 1 and variance 2:
m = 1;
v = 2;
mu = log((m^2)/sqrt(v+m^2));
sigma = sqrt(log(v/(m^2)+1));
[M,V]= lognstat(mu,sigma)
M =
1
V =
2.0000
X = lognrnd(mu,sigma,1,1e6);
MX = mean(X)
MX =
0.9974
VX = var(X)
VX =
1.9776
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
Hoboken, NJ: Wiley-Interscience, 2000. pp. 102–105.
See Also random, lognpdf, logncdf, logninv, lognstat, lognfit, lognlike
“Lognormal Distribution” on page B-51
18-711
lognstat
Purpose Lognormal mean and variance
Syntax [M,V] = lognstat(mu,sigma)
Description [M,V] = lognstat(mu,sigma) returns the mean of and variance of the
lognormal distribution with parameters mu and sigma. mu and sigma
are the mean and standard deviation, respectively, of the associated
normal distribution. mu and sigma can be vectors, matrices, or
multidimensional arrays that all have the same size, which is also the
size of M and V. A scalar input for mu or sigma is expanded to a constant
array with the same dimensions as the other input.
The normal and lognormal distributions are closely related. If X
is distributed lognormally with parameters µ and σ, then log(X) is
distributed normally with mean µ and standard deviation σ.
The mean m and variance v of a lognormal random variable are
functions of µ and σ that can be calculated with the lognstat function.
They are:
(
m = exp + 2 / 2 )
( )( ( ) )
v = exp 2 + 2 exp 2 − 1
A lognormal distribution with mean m and variance v has parameters
= log ⎛⎜ m2 / v + m2 ⎞⎟
⎝ ⎠
(
= log v / m2 + 1 )
Examples Generate one million lognormally distributed random numbers with
mean 1 and variance 2:
m = 1;
v = 2;
18-712
lognstat
mu = log((m^2)/sqrt(v+m^2));
sigma = sqrt(log(v/(m^2)+1));
[M,V]= lognstat(mu,sigma)
M =
1
V =
2.0000
X = lognrnd(mu,sigma,1,1e6);
MX = mean(X)
MX =
0.9974
VX = var(X)
VX =
1.9776
References [1] Mood, A. M., F. A. Graybill, and D. C. Boes. Introduction to the
Theory of Statistics. 3rd ed., New York: McGraw-Hill, 1974. pp.
540–541.
See Also lognpdf, logncdf, logninv, lognfit, lognlike, lognrnd
“Lognormal Distribution” on page B-51
18-713
paretotails.lowerparams
Purpose Lower Pareto tails parameters
Syntax params = lowerparams(obj)
Description params = lowerparams(obj) returns the 2-element vector params of
shape and scale parameters, respectively, of the lower tail of the Pareto
tails object obj. lowerparams does not return a location parameter.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
lowerparams(obj)
ans =
-0.1901 1.1898
upperparams(obj)
ans =
0.3646 0.5103
See Also paretotails, upperparams
18-714
qrandstream.lt
Purpose Less than relation for handles
Syntax h1 < h2
Description h1 < h2 performs element-wise comparisons between handle arrays
h1 and h2. h1 and h2 must be of the same dimensions unless one is a
scalar. The result is a logical array of the same dimensions, where each
element is an element-wise < result.
If one of h1 or h2 is scalar, scalar expansion is performed and the result
will match the dimensions of the array that is not scalar.
tf = lt(h1, h2) stores the result in a logical array of the same
dimensions.
See Also qrandstream, eq, ge, gt, le, ne
18-715
lsline
Purpose Add least-squares line to scatter plot
Syntax lsline
h = lsline
Description lsline superimposes a least-squares line on each scatter plot in the
current axes. Scatter plots are produced by the MATLAB scatter and
plot functions. Data points connected with solid, dashed, or dash-dot
lines (LineStyle '-', '--', or '.-') are not considered to be scatter
plots by lsline, and are ignored.
h = lsline returns a column vector of handles h to the least-squares
lines.
Examples Use lsline together with scatter plots produced by scatter and
various line styles of plot:
x = 1:10;
y1 = x + randn(1,10);
scatter(x,y1,25,'b','*')
hold on
y2 = 2*x + randn(1,10);
plot(x,y2,'mo')
y3 = 3*x + randn(1,10);
plot(x,y3,'rx:')
y4 = 4*x + randn(1,10);
plot(x,y4,'g+--')
lsline
18-716
lsline
See Also scatter, plot, refline, refcurve, gline
18-717
mad
Purpose Mean or median absolute deviation
Syntax y = mad(X)
Y = mad(X,1)
Y = mad(X,0)
Description y = mad(X) returns the mean absolute deviation of the values in X. For
vector input, y is mean(abs(X-mean(X)). For a matrix input, y is a
row vector containing the mean absolute deviation of each column of
X. For N-dimensional arrays, mad operates along the first nonsingleton
dimension of X.
Y = mad(X,1) returns the median absolute deviation of the values in X.
For vector input, y is median(abs(X-median(X)). For a matrix input, y is
a row vector containing the median absolute deviation of each column of
X. For N-dimensional arrays, mad operates along the first nonsingleton
dimension of X.
Y = mad(X,0) is the same as mad(X), and returns the mean absolute
deviation of the values in X.
mad(X,flag,dim) computes absolute deviations along the dimension
dim of X. flag is 0 or 1 to indicate mean or median absolute deviation,
respectively.
mad treats NaNs as missing values and removes them.
For normally distributed data, multiply mad by one of the following
factors to obtain an estimate of the normal scale parameter σ:
• sigma = 1.253*mad(X,0) — For mean absolute deviation
• sigma = 1.4826*mad(X,1) — For median absolute deviation
Examples The following compares the robustness of different scale estimates for
normally distributed data in the presence of outliers:
x = normrnd(0,1,1,50);
xo = [x 10]; % Add outlier
18-718
mad
r1 = std(xo)/std(x)
r1 =
1.7385
r2 = mad(xo,0)/mad(x,0)
r2 =
1.2306
r3 = mad(xo,1)/mad(x,1)
r3 =
1.0602
References [1] Mosteller, F., and J. Tukey. Data Analysis and Regression. Upper
Saddle River, NJ: Addison-Wesley, 1977.
[2] Sachs, L. Applied Statistics: A Handbook of Techniques. New York:
Springer-Verlag, 1984, p. 253.
See Also std | range | iqr
18-719
mahal
Purpose Mahalanobis distance
Syntax d = mahal(Y,X)
Description d = mahal(Y,X) computes the Mahalanobis distance (in squared units)
of each observation in Y from the reference sample in matrix X. If Y is
n-by-m, where n is the number of observations and m is the dimension
of the data, d is n-by-1. X and Y must have the same number of columns,
but can have different numbers of rows. X must have more rows than
columns.
For observation I, the Mahalanobis distance is defined by d(I) =
(Y(I,:)-mu)*inv(SIGMA)*(Y(I,:)-mu)', where mu and SIGMA are
the sample mean and covariance of the data in X. mahal performs an
equivalent, but more efficient, computation.
Examples Generate some correlated bivariate data in X and compare the
Mahalanobis and squared Euclidean distances of observations in Y:
X = mvnrnd([0;0],[1 .9;.9 1],100);
Y = [1 1;1 -1;-1 1;-1 -1];
d1 = mahal(Y,X) % Mahalanobis
d1 =
1.3592
21.1013
23.8086
1.4727
d2 = sum((Y-repmat(mean(X),4,1)).^2, 2) % Squared Euclidean
d2 =
1.9310
1.8821
2.1228
2.0739
scatter(X(:,1),X(:,2))
18-720
mahal
hold on
scatter(Y(:,1),Y(:,2),100,d1,'*','LineWidth',2)
hb = colorbar;
ylabel(hb,'Mahalanobis Distance')
legend('X','Y','Location','NW')
The observations in Y with equal coordinate values are much closer to
X in Mahalanobis distance than observations with opposite coordinate
values, even though all observations are approximately equidistant
from the mean of X in Euclidean distance. The Mahalanobis distance,
by considering the covariance of the data and the scales of the different
variables, is useful for detecting outliers in such cases.
See Also pdist, mahal
18-721
gmdistribution.mahal
Purpose Mahalanobis distance to component means
Syntax D = mahal(obj,X)
Description D = mahal(obj,X) computes the Mahalanobis distance (in squared
units) of each observation in X to the mean of each of the k components
of the Gaussian mixture distribution defined by obj. obj is an object
created by gmdistribution or fit. X is an n-by-d matrix, where n is
the number of observations and d is the dimension of the data. D is
n-by-k, with D(I,J) the distance of observation I from the mean of
component J.
Examples Generate data from a mixture of two bivariate Gaussian distributions
using the mvnrnd function:
MU1 = [1 2];
SIGMA1 = [2 0; 0 .5];
MU2 = [-3 -5];
SIGMA2 = [1 0; 0 1];
X = [mvnrnd(MU1,SIGMA1,1000);mvnrnd(MU2,SIGMA2,1000)];
scatter(X(:,1),X(:,2),10,'.')
hold on
18-722
gmdistribution.mahal
Fit a two-component Gaussian mixture model:
obj = gmdistribution.fit(X,2);
h = ezcontour(@(x,y)pdf(obj,[x y]),[-8 6],[-8 6]);
18-723
gmdistribution.mahal
Compute the Mahalanobis distance of each point in X to the mean
of each component of obj:
D = mahal(obj,X);
delete(h)
scatter(X(:,1),X(:,2),10,D(:,1),'.')
hb = colorbar;
ylabel(hb,'Mahalanobis Distance to Component 1')
18-724
gmdistribution.mahal
See Also gmdistribution, cluster, posterior, mahal
18-725
maineffectsplot
Purpose Main effects plot for grouped data
Syntax maineffectsplot(Y,GROUP)
maineffectsplot(Y,GROUP,param1,val1,param2,val2,...)
[figh,AXESH] = maineffectsplot(...)
Description maineffectsplot(Y,GROUP) displays main effects plots for the group
means of matrix Y with groups defined by entries in the cell array
GROUP. Y is a numeric matrix or vector. If Y is a matrix, the rows
represent different observations and the columns represent replications
of each observation. Each cell of GROUP must contain a grouping variable
that can be a categorical variable, numeric vector, character matrix, or
single-column cell array of strings. (See “Grouped Data” on page 2-34.)
GROUP can also be a matrix whose columns represent different grouping
variables. Each grouping variable must have the same number of rows
as Y. The number of grouping variables must be greater than 1.
The display has one subplot per grouping variable, with each subplot
showing the group means of Y as a function of one grouping variable.
maineffectsplot(Y,GROUP,param1,val1,param2,val2,...) specifies
one or more of the following name/value pairs:
• 'varnames' — Grouping variable names in a character matrix or a
cell array of strings, one per grouping variable. Default names are
'X1', 'X2', ... .
• 'statistic' — String values that indicate whether the group mean
or the group standard deviation should be plotted. Use 'mean' or
'std'. The default is 'mean'. If the value is 'std', Y is required
to have multiple columns.
• 'parent' — A handle to the figure window for the plots. The default
is the current figure window.
[figh,AXESH] = maineffectsplot(...) returns the handle figh to
the figure window and an array of handles AXESH to the subplot axes.
18-726
maineffectsplot
Examples Display main effects plots for car weight with two grouping variables,
model year and number of cylinders:
load carsmall;
maineffectsplot(Weight,{Model_Year,Cylinders}, ...
'varnames',{'Model Year','# of Cylinders'})
See Also “Grouped Data” on page 2-34
interactionplot, multivarichart
18-727
manova1
Purpose One-way multivariate analysis of variance
Syntax d = manova1(X,group)
d = manova1(X,group,alpha)
[d,p] = manova1(...)
[d,p,stats] = manova1(...)
Description d = manova1(X,group) performs a one-way Multivariate Analysis of
Variance (MANOVA) for comparing the multivariate means of the
columns of X, grouped by group. X is an m-by-n matrix of data values,
and each row is a vector of measurements on n variables for a single
observation. group is a grouping variable defined as a categorical
variable, vector, string array, or cell array of strings. Two observations
are in the same group if they have the same value in the group array.
(See “Grouped Data” on page 2-34.) The observations in each group
represent a sample from a population.
The function returns d, an estimate of the dimension of the space
containing the group means. manova1 tests the null hypothesis that the
means of each group are the same n-dimensional multivariate vector,
and that any difference observed in the sample X is due to random
chance. If d = 0, there is no evidence to reject that hypothesis. If d = 1,
then you can reject the null hypothesis at the 5% level, but you cannot
reject the hypothesis that the multivariate means lie on the same line.
Similarly, if d = 2 the multivariate means may lie on the same plane in
n-dimensional space, but not on the same line.
d = manova1(X,group,alpha) gives control of the significance level,
alpha. The return value d will be the smallest dimension having
p > alpha, where p is a p value for testing whether the means lie in a
space of that dimension.
[d,p] = manova1(...) also returns a p, a vector of p-values for testing
whether the means lie in a space of dimension 0, 1, and so on. The
largest possible dimension is either the dimension of the space, or one
less than the number of groups. There is one element of p for each
dimension up to, but not including, the largest.
18-728
manova1
If the ith p value is near zero, this casts doubt on the hypothesis that
the group means lie on a space of i-1 dimensions. The choice of a
critical p value to determine whether the result is judged statistically
significant is left to the researcher and is specified by the value of the
input argument alpha. It is common to declare a result significant if
the p value is less than 0.05 or 0.01.
[d,p,stats] = manova1(...) also returns stats, a structure
containing additional MANOVA results. The structure contains the
following fields.
Field Contents
W Within-groups sum of squares and cross-products
matrix
B Between-groups sum of squares and cross-products
matrix
T Total sum of squares and cross-products matrix
dfW Degrees of freedom for W
dfB Degrees of freedom for B
dfT Degrees of freedom for T
lambda Vector of values of Wilk’s lambda test statistic for
testing whether the means have dimension 0, 1, etc.
chisq Transformation of lambda to an approximate
chi-square distribution
chisqdf Degrees of freedom for chisq
eigenval Eigenvalues of W-1B
eigenvec Eigenvectors of W-1B; these are the coefficients for
the canonical variables C, and they are scaled so the
within-group variance of the canonical variables is 1
18-729
manova1
Field Contents
canon Canonical variables C, equal to XC*eigenvec, where XC
is X with columns centered by subtracting their means
mdist A vector of Mahalanobis distances from each point
to the mean of its group
gmdist A matrix of Mahalanobis distances between each pair
of group means
The canonical variables C are linear combinations of the original
variables, chosen to maximize the separation between groups.
Specifically, C(:,1) is the linear combination of the X columns that has
the maximum separation between groups. This means that among all
possible linear combinations, it is the one with the most significant F
statistic in a one-way analysis of variance. C(:,2) has the maximum
separation subject to it being orthogonal to C(:,1), and so on.
You may find it useful to use the outputs from manova1 along with other
functions to supplement your analysis. For example, you may want to
start with a grouped scatter plot matrix of the original variables using
gplotmatrix. You can use gscatter to visualize the group separation
using the first two canonical variables. You can use manovacluster to
graph a dendrogram showing the clusters among the group means.
Assumptions
The MANOVA test makes the following assumptions about the data
in X:
• The populations for each group are normally distributed.
• The variance-covariance matrix is the same for each population.
• All observations are mutually independent.
Examples you can use manova1 to determine whether there are differences in
the averages of four car characteristics, among groups defined by the
country where the cars were made.
18-730
manova1
load carbig
[d,p] = manova1([MPG Acceleration Weight Displacement],...
Origin)
d =
3
p =
0
0.0000
0.0075
0.1934
There are four dimensions in the input matrix, so the group means
must lie in a four-dimensional space. manova1 shows that you cannot
reject the hypothesis that the means lie in a 3-D subspace.
References [1] Krzanowski, W. J. Principles of Multivariate Analysis: A User’s
Perspective. New York: Oxford University Press, 1988.
See Also “Grouped Data” on page 2-34
anova1, canoncorr, gscatter, gplotmatrix, manovacluster
18-731
manovacluster
Purpose Dendrogram of group mean clusters following MANOVA
Syntax manovacluster(stats)
manovacluster(stats,method)
H = manovacluster(stats,method)
Description manovacluster(stats) generates a dendrogram plot of the group
means after a multivariate analysis of variance (MANOVA). stats is
the output stats structure from manova1. The clusters are computed
by applying the single linkage method to the matrix of Mahalanobis
distances between group means.
See dendrogram for more information on the graphical output from this
function. The dendrogram is most useful when the number of groups
is large.
manovacluster(stats,method) uses the specified method in place of
single linkage. method can be any of the following character strings
that identify ways to create the cluster hierarchy. (See linkage for
additional information.)
Method Description
'single' Shortest distance (default)
'complete' Largest distance
'average' Average distance
'centroid' Centroid distance
'ward' Incremental sum of squares
H = manovacluster(stats,method) returns a vector of handles to the
lines in the figure.
Examples Let’s analyze the larger car data set to determine which countries
produce cars with the most similar characteristics.
load carbig
18-732
manovacluster
X = [MPG Acceleration Weight Displacement];
[d,p,stats] = manova1(X,Origin);
manovacluster(stats)
See Also cluster, dendrogram, linkage, manova1
18-733
CompactTreeBagger.margin
Purpose Classification margin
Syntax mar = margin(B,X,Y)
mar = margin(B,X,Y,'param1',val1,'param2',val2,...)
Description mar = margin(B,X,Y) computes the classification margins for
predictors X given true response Y. The Y can be either a numeric vector,
character matrix, cell array of strings, categorical vector or logical
vector. mar is a numeric array of size Nobs-by-NTrees, where Nobs is
the number of rows of X and Y, and NTrees is the number of trees in the
ensemble B. For observation I and tree J, mar(I,J) is the difference
between the score for the true class and the largest score for other
classes. This method is available for classification ensembles only.
mar = margin(B,X,Y,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'mode' String indicating how the method computes errors.
If set to 'cumulative' (default), margin computes
cumulative errors and mar is an Nobs-by-NTrees
matrix, where the first column gives error from
trees(1), second column gives error fromtrees(1:2)
etc, up to trees(1:NTrees). If set to 'individual',
mar is a Nobs-by-NTrees matrix, where each element
is an error from each tree in the ensemble. If set
to 'ensemble', mar a single column of length Nobs
showing the cumulative margins for the entire
ensemble.
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements in the
input vector, and a scalar for 'ensemble' mode. For
example, in the 'cumulative' mode, the first element
18-734
CompactTreeBagger.margin
gives error from trees(1), the second element gives
error from trees(1:2) etc.
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. The method uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple non-weighted majority vote. You cannot use
this argument in the 'individual' mode.
'useifort' Logical matrix of size Nobs-by-NTrees indicating
which trees should be used to make predictions for
each observation. By default the method uses all trees
for all observations.
See Also TreeBagger.margin
18-735
TreeBagger.margin
Purpose Classification margin
Syntax mar = margin(B,X,Y)
mar = margin(B,X,Y,'param1',val1,'param2',val2,...)
Description mar = margin(B,X,Y) computes the classification margins for
predictors X given true response Y. The Y can be either a numeric vector,
character matrix, cell array of strings, categorical vector or logical
vector. mar is a numeric array of size Nobs-by-NTrees, where Nobs is
the number of rows of X and Y, and NTrees is the number of trees in the
ensemble B. For observation I and tree J, mar(I,J) is the difference
between the score for the true class and the largest score for other
classes. This method is available for classification ensembles only.
mar = margin(B,X,Y,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'mode' String indicating how the method computes errors.
If set to 'cumulative' (default), margin computes
cumulative errors and mar is an Nobs-by-NTrees
matrix, where the first column gives error from
trees(1), second column gives error fromtrees(1:2)
etc, up to trees(1:NTrees). If set to 'individual',
mar is a Nobs-by-NTrees matrix, where each element
is an error from each tree in the ensemble. If set
to 'ensemble', mar a single column of length Nobs
showing the cumulative margins for the entire
ensemble.
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements in the
input vector, and a scalar for 'ensemble' mode. For
example, in the 'cumulative' mode, the first element
18-736
TreeBagger.margin
gives error from trees(1), the second element gives
error from trees(1:2) etc.
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. The method uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple non-weighted majority vote. You cannot use
this argument in the 'individual' mode.
'useifort' Logical matrix of size Nobs-by-NTrees indicating
which trees should be used to make predictions for
each observation. By default the method uses all trees
for all observations.
See Also CompactTreeBagger.margin
18-737
mdscale
Purpose Nonclassical multidimensional scaling
Syntax Y = mdscale(D,p)
[Y,stress] = mdscale(D,p)
[Y,stress,disparities] = mdscale(D,p)
[...] = mdscale(...,param1,val1,param2,val2,...)
Description Y = mdscale(D,p) performs nonmetric multidimensional scaling on
the n-by-n dissimilarity matrix D, and returns Y, a configuration of
n points (rows) in p dimensions (columns). The Euclidean distances
between points in Y approximate a monotonic transformation of the
corresponding dissimilarities in D. By default, mdscale uses Kruskal’s
normalized stress1 criterion.
You can specify D as either a full n-by-n matrix, or in upper triangle
form such as is output by pdist. A full dissimilarity matrix must be real
and symmetric, and have zeros along the diagonal and non-negative
elements everywhere else. A dissimilarity matrix in upper triangle
form must have real, non-negative entries. mdscale treats NaNs in D as
missing values, and ignores those elements. Inf is not accepted.
You can also specify D as a full similarity matrix, with ones along the
diagonal and all other elements less than one. mdscale transforms a
similarity matrix to a dissimilarity matrix in such a way that distances
between the points returned in Y approximate sqrt(1-D). To use a
different transformation, transform the similarities prior to calling
mdscale.
[Y,stress] = mdscale(D,p) returns the minimized stress, i.e., the
stress evaluated at Y.
[Y,stress,disparities] = mdscale(D,p) returns the disparities,
that is, the monotonic transformation of the dissimilarities D.
[...] = mdscale(...,param1,val1,param2,val2,...) enables you
to specify optional parameter name/value pairs that control further
details of mdscale. The parameters are
18-738
mdscale
• 'Criterion'— The goodness-of-fit criterion to minimize. This also
determines the type of scaling, either non-metric or metric, that
mdscale performs. Choices for non-metric scaling are:
- 'stress' — Stress normalized by the sum of squares of the
inter-point distances, also known as stress1. This is the default.
- 'sstress' — Squared stress, normalized with the sum of 4th
powers of the inter-point distances.
Choices for metric scaling are:
- 'metricstress' — Stress, normalized with the sum of squares of
the dissimilarities.
- 'metricsstress' — Squared stress, normalized with the sum of
4th powers of the dissimilarities.
- 'sammon' — Sammon’s nonlinear mapping criterion. Off-diagonal
dissimilarities must be strictly positive with this criterion.
- 'strain' — A criterion equivalent to that used in classical
multidimensional scaling.
• 'Weights' — A matrix or vector the same size as D, containing
nonnegative dissimilarity weights. You can use these to weight the
contribution of the corresponding elements of D in computing and
minimizing stress. Elements of D corresponding to zero weights are
effectively ignored.
• 'Start' — Method used to choose the initial configuration of points
for Y. The choices are
- 'cmdscale' — Use the classical multidimensional scaling solution.
This is the default. 'cmdscale' is not valid when there are zero
weights.
- 'random' — Choose locations randomly from an appropriately
scaled p-dimensional normal distribution with uncorrelated
coordinates.
- An n-by-p matrix of initial locations, where n is the size of the
matrix D and p is the number of columns of the output matrix Y.
18-739
mdscale
In this case, you can pass in [] for p and mdscale infers p from
the second dimension of the matrix. You can also supply a 3-D
array, implying a value for 'Replicates' from the array’s third
dimension.
• 'Replicates' — Number of times to repeat the scaling, each with a
new initial configuration. The default is 1.
• 'Options' — Options for the iterative algorithm used to minimize
the fitting criterion. Pass in an options structure created by statset.
For example,
opts = statset(param1,val1,param2,val2, ...);
[...] = mdscale(...,'Options',opts)
The choices of statset parameters are
- 'Display' — Level of display output. The choices are 'off' (the
default), 'iter', and 'final'.
- 'MaxIter' — Maximum number of iterations allowed. The default
is 200.
- 'TolFun' — Termination tolerance for the stress criterion and its
gradient. The default is 1e-4.
- 'TolX'— Termination tolerance for the configuration location step
size. The default is 1e-4.
Examples load cereal.mat
X = [Calories Protein Fat Sodium Fiber ...
Carbo Sugars Shelf Potass Vitamins];
% Take a subset from a single manufacturer.
X = X(strmatch('K',Mfg),:);
% Create a dissimilarity matrix.
dissimilarities = pdist(X);
% Use non-metric scaling to recreate the data in 2D,
18-740
mdscale
% and make a Shepard plot of the results.
[Y,stress,disparities] = mdscale(dissimilarities,2);
distances = pdist(Y);
[dum,ord] = sortrows([disparities(:) dissimilarities(:)]);
plot(dissimilarities,distances,'bo', ...
dissimilarities(ord),disparities(ord),'r.-');
xlabel('Dissimilarities'); ylabel('Distances/Disparities')
legend({'Distances' 'Disparities'},'Location','NW');
% Do metric scaling on the same dissimilarities.
[Y,stress] = ...
mdscale(dissimilarities,2,'criterion','metricsstress');
distances = pdist(Y);
plot(dissimilarities,distances,'bo', ...
[0 max(dissimilarities)],[0 max(dissimilarities)],'k:');
xlabel('Dissimilarities'); ylabel('Distances')
See Also cmdscale, pdist, statset
18-741
CompactTreeBagger.mdsProx
Purpose Multidimensional scaling of proximity matrix
Syntax [SC,EIGEN] = mdsProx(B,X)
[SC,EIGEN] = mdsProx(B,X,'param1',val1,'param2',val2,...)
Description [SC,EIGEN] = mdsProx(B,X) applies classical multidimensional
scaling to the proximity matrix computed for the data in the matrix X,
and returns scaled coordinates SC and eigenvalues EIGEN of the scaling
transformation. The method applies multidimensional scaling to the
matrix of distances defined as 1-prox, where prox is the proximity
matrix returned by the proximity method.
You can supply the proximity matrix directly by using the 'data'
parameter.
[SC,EIGEN] = mdsProx(B,X,'param1',val1,'param2',val2,...)
specifies optional parameter name/value pairs:
'data' Flag indicating how the method treats the X input
argument. If set to 'predictors' (default), mdsProx
assumes X to be a matrix of predictors and used
for computation of the proximity matrix. If set to
'proximity', the method treats X as a proximity matrix
returned by the proximity method.
'colors' If you supply this argument, mdsProx makes overlaid
scatter plots of two scaled coordinates using specified
colors for different classes. You must supply the colors
as a string with one character for each color. If there are
more classes in the data than characters in the supplied
string, mdsProx only plots the first C classes, where C is
the length of the string. For regression or if you do not
provide the vector of true class labels, the method uses
the first color for all observations in X.
18-742
CompactTreeBagger.mdsProx
'labels' Vector of true class labels for a classification ensemble.
True class labels can be either a numeric vector,
character matrix, or cell array of strings. If supplied,
this vector must have as many elements as there are
observations (rows) in X. This argument has no effect
unless you also supply the 'colors' argument.
'mdscoords' Indices of the two scaled coordinates to plot. By default,
mdsProx makes a scatter plot of the first and second
scaled coordinates which correspond to the two largest
eigenvalues. You can specify any other two or three
indices not exceeding the dimensionality of the scaled
data. This argument has no effect unless you also supply
the 'colors' argument.
See Also cmdscale, TreeBagger.mdsProx, proximity
18-743
TreeBagger.mdsProx
Purpose Multidimensional scaling of proximity matrix
Syntax [S,E] = mdsProx(B)
[S,E] = mdsProx(B,'param1',val1,'param2',val2,...)
Description [S,E] = mdsProx(B) returns scaled coordinates, S, and eigenvalues,
E, for the proximity matrix in the ensemble B. An earlier call to
fillProximities(B) must create the proximity matrix.
[S,E] = mdsProx(B,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'keep' Array of indices of observations in the training data
to use for multidimensional scaling. By default, this
argument is set to 'all'. If you provide numeric or
logical indices, the method uses only the subset of the
training data specified by these indices to compute
the scaled coordinates and eigenvalues.
'colors' If you supply this argument, mdsProx makes overlaid
scatter plots of two scaled coordinates using specified
colors for different classes. You must supply the colors
as a string with one character for each color. If there
are more classes in the data than characters in the
supplied string, mdsProx only plots the first C classes,
where C is the length of the string. For regression or if
you do not provide the vector of true class labels, the
method uses the first color for all observations in X.
'mdscoords' Indices of the two scaled coordinates to plot. By
default, mdsProx makes a scatter plot of the first and
second scaled coordinates which correspond to the two
largest eigenvalues. You can specify any other two or
three indices not exceeding the dimensionality of the
scaled data. This argument has no effect unless you
also supply the 'colors' argument.
See Also cmdscale, CompactTreeBagger.mdsProx, fillProximities
18-744
ProbDistUnivKernel.median
Purpose Return median of ProbDistUnivKernel object
Syntax M = median(PD)
Description M = median(PD) returns M, the median of the ProbDistUnivKernel
object PD.
Input PD An object of the class ProbDistUnivKernel.
Arguments
Output M The median of the ProbDistUnivKernel object
Arguments PD.
See Also median
18-745
ProbDistUnivParam.median
Purpose Return median of ProbDistUnivParam object
Syntax M = median(PD)
Description M = median(PD) returns M, the median of the ProbDistUnivParam
object PD.
Input PD An object of the class ProbDistUnivParam.
Arguments
Output M The median of the ProbDistUnivParam object
Arguments PD.
See Also median
18-746
ProbDistUnivParam.mean
Purpose Return mean of ProbDistUnivParam object
Syntax M = mean(PD)
Description M = mean(PD) returns M, the mean of the ProbDistUnivParam object PD.
Input PD An object of the class ProbDistUnivParam.
Arguments
Output M The mean of the ProbDistUnivParam object
Arguments PD.
See Also mean
18-747
CompactTreeBagger.meanMargin
Purpose Mean classification margin
Syntax mar = meanMargin(B,X,Y)
mar = meanMargin(B,X,Y,'param1',val1,'param2',val2,...)
Description mar = meanMargin(B,X,Y) computes average classification margins for
predictors X given true response Y. The Y can be either a numeric vector,
character matrix, cell array of strings, categorical vector or logical
vector. meanMargin averages the margins over all observations (rows)
in X for each tree. mar is a matrix of size 1-by-NTrees, where NTrees
is the number of trees in the ensemble B. This method is available for
classification ensembles only.
mar = meanMargin(B,X,Y,'param1',val1,'param2',val2,...)
specifies optional parameter name/value pairs:
'mode' String indicating how meanMargin computes errors.
If set to 'cumulative' (default), is a vector of length
NTrees where the first element gives mean margin
from trees(1), second column gives mean margins
from trees(1:2) etc, up to trees(1:NTrees). If set
to 'individual', mar is a vector of length NTrees,
where each element is a mean margin from each tree
in the ensemble . If set to 'ensemble', mar is a scalar
showing the cumulative mean margin for the entire
ensemble.
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements in the
input vector, and a scalar for 'ensemble' mode. For
example, in the 'cumulative' mode, the first element
gives mean margin from trees(1), the second
element gives mean margin from trees(1:2) etc.
18-748
CompactTreeBagger.meanMargin
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. meanMargin uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple nonweighted majority vote. You cannot use
this argument in the 'individual' mode.
See Also TreeBagger.meanMargin
18-749
TreeBagger.meanMargin
Purpose Mean classification margin
Syntax mar = meanMargin(B,X,Y)
mar = meanMargin(B,X,Y,'param1',val1,'param2',val2,...)
Description mar = meanMargin(B,X,Y) computes average classification margins for
predictors X given true response Y. The Y can be either a numeric vector,
character matrix, cell array of strings, categorical vector or logical
vector. meanMargin averages the margins over all observations (rows)
in X for each tree. mar is a matrix of size 1-by-NTrees, where NTrees
is the number of trees in the ensemble B. This method is available for
classification ensembles only.
mar = meanMargin(B,X,Y,'param1',val1,'param2',val2,...)
specifies optional parameter name/value pairs:
'mode' String indicating how meanMargin computes errors.
If set to 'cumulative' (default), is a vector of length
NTrees where the first element gives mean margin
from trees(1), second column gives mean margins
from trees(1:2) etc, up to trees(1:NTrees). If set
to 'individual', mar is a vector of length NTrees,
where each element is a mean margin from each tree
in the ensemble . If set to 'ensemble', mar is a scalar
showing the cumulative mean margin for the entire
ensemble .
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements in the
input vector, and a scalar for 'ensemble' mode. For
example, in the 'cumulative' mode, the first element
gives mean margin from trees(1), the second
element gives mean margin from trees(1:2) etc.
18-750
TreeBagger.meanMargin
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. meanMargin uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple nonweighted majority vote. You cannot use
this argument in the 'individual' mode.
See Also CompactTreeBagger.meanMargin
18-751
TreeBagger.MergeLeaves property
Purpose Flag to merge leaves that do not improve risk
Description The MergeLeaves property is true if decision trees have their leaves
with the same parent merged for splits that do not decrease the total
risk, and false otherwise. The default value is false.
See Also classregtree
18-752
ordinal.mergelevels
Purpose Merge levels
Syntax B = mergelevels(A,oldlevels,newlevel)
B = mergelevels(A,oldlevels)
Description B = mergelevels(A,oldlevels,newlevel) merges two or more levels
of the categorical array A into a single new level. oldlevels is a cell
array of strings or a 2-D character matrix that specifies the levels
to be merged. Any elements of A that have levels in oldlevels are
assigned the new level in the corresponding elements of B. newlevel is
a character string that specifies the label for the new level. For ordinal
arrays, the levels of A specified by oldlevels must be consecutive, and
mergelevels inserts the new level to preserve the order of the levels.
B = mergelevels(A,oldlevels) merges two or more levels of A. For
nominal arrays, mergelevels uses the first label in oldlevels as the
label for the new level. For ordinal arrays, mergelevels uses the label
corresponding to the lowest level in oldlevels as the label for the new
level.
Examples Example 1
For nominal data:
load fisheriris
species = nominal(species);
species = mergelevels(species,...
{'setosa','virginica'},'parent');
species = setlabels(species,'hybrid','versicolor');
getlabels(species)
ans =
'hybrid' 'parent'
18-753
ordinal.mergelevels
Example 2
For ordinal data:
A = ordinal([1 2 3 2 1],{'lo','med','hi'})
A =
lo med hi med lo
A = mergelevels(A,{'lo','med'},'bad')
A =
bad bad hi bad bad
See Also addlevels, droplevels, islevel, reorderlevels, getlabels
18-754
CompactTreeBagger.Method property
Purpose Method used by trees (classification or regression)
Description The Method property is 'classification' for classification ensembles
and 'regression' for regression ensembles.
18-755
TreeBagger.Method property
Purpose Method used by trees (classification or regression)
Description The Method property is 'classification' for classification ensembles
and 'regression' for regression ensembles.
18-756
mhsample
Purpose Metropolis-Hastings sample
Syntax smpl = mhsample(start,nsamples,'pdf',pdf,'proppdf',proppdf,
'proprnd',proprnd)
smpl = mhsample(...,'symmetric',sym)
smpl = mhsample(...,'burnin',K)
smpl = mhsample(...,'thin',m)
smpl = mhsample(...,'nchain',n)
[smpl,accept] = mhsample(...)
Description smpl =
mhsample(start,nsamples,'pdf',pdf,'proppdf',proppdf,'proprnd',proprnd)
draws nsamples random samples from a target stationary distribution
pdf using the Metropolis-Hastings algorithm.
start is a row vector containing the start value of the Markov
Chain, nsamples is an integer specifying the number of samples to be
generated, and pdf, proppdf, and proprnd are function handles created
using @. proppdf defines the proposal distribution density, and proprnd
defines the random number generator for the proposal distribution. pdf
and proprnd take one argument as an input with the same type and
size as start. proppdf takes two arguments as inputs with the same
type and size as start.
smpl is a column vector or matrix containing the samples. If the log
density function is preferred, 'pdf' and 'proppdf' can be replaced
with 'logpdf' and 'logproppdf'. The density functions used in
Metropolis-Hastings algorithm are not necessarily normalized.
The proposal distribution q(x,y) gives the probability density for
choosing x as the next point when y is the current point. It is sometimes
written as q(x|y).
If the proppdf or logproppdf satisfies q(x,y) = q(y,x), that is, the
proposal distribution is symmetric, mhsample implements Random
Walk Metropolis-Hastings sampling. If the proppdf or logproppdf
satisfies q(x,y) = q(x), that is, the proposal distribution is independent of
current values, mhsample implements Independent Metropolis-Hastings
sampling.
18-757
mhsample
smpl = mhsample(...,'symmetric',sym) draws nsamples random
samples from a target stationary distribution pdf using the
Metropolis-Hastings algorithm. sym is a logical value that indicates
whether the proposal distribution is symmetric. The default value is
false, which corresponds to the asymmetric proposal distribution. If sym
is true, for example, the proposal distribution is symmetric, proppdf
and logproppdf are optional.
smpl = mhsample(...,'burnin',K) generates a Markov chain with
values between the starting point and the kth point omitted in the
generated sequence. Values beyond the kth point are kept. k is a
nonnegative integer with default value of 0.
smpl = mhsample(...,'thin',m) generates a Markov chain with
m-1 out of m values omitted in the generated sequence. m is a positive
integer with default value of 1.
smpl = mhsample(...,'nchain',n) generates n Markov chains using
the Metropolis-Hastings algorithm. n is a positive integer with a default
value of 1. smpl is a matrix containing the samples. The last dimension
contains the indices for individual chains.
[smpl,accept] = mhsample(...) also returns accept, the acceptance
rate of the proposed distribution. accept is a scalar if a single chain is
generated and is a vector if multiple chains are generated.
Examples Estimate the second order moment of a Gamma distribution using the
Independent Metropolis-Hastings sampling.
alpha = 2.43;
beta = 1;
pdf = @(x)gampdf(x,alpha,beta); %target distribution
proppdf = @(x,y)gampdf(x,floor(alpha),floor(alpha)/alpha);
proprnd = @(x)sum(...
exprnd(floor(alpha)/alpha,floor(alpha),1));
nsamples = 5000;
smpl = mhsample(1,nsamples,'pdf',pdf,'proprnd',proprnd,...
'proppdf',proppdf);
xxhat = cumsum(smpl.^2)./(1:nsamples)';
18-758
mhsample
plot(1:nsamples,xxhat)
Generate random samples from N(0,1) using the Random Walk
Metropolis-Hastings sampling.
delta = .5;
pdf = @(x) normpdf(x);
proppdf = @(x,y) unifpdf(y-x,-delta,delta);
proprnd = @(x) x + rand*2*delta - delta;
nsamples = 15000;
x = mhsample(1,nsamples,'pdf',pdf,'proprnd',proprnd,'symmetric',1);
histfit(x,50)
h = get(gca,'Children');
set(h(2),'FaceColor',[.8 .8 1])
18-759
mhsample
See Also slicesample, rand
18-760
TreeBagger.MinLeaf property
Purpose Minimum number of observations per tree leaf
Description The MinLeaf property specifies the minimum number of observations
per tree leaf. The default values are 1 for classification and 5 for
regression. For classregtree training, the 'minparent' value is set to
2*MinLeaf.
See Also classregtree
18-761
mle
Purpose Maximum likelihood estimates
Syntax phat = mle(data)
[phat,pci] = mle(data)
[...] = mle(data,'distribution',dist)
[...] = mle(data,...,name1,val1,name2,val2,...)
[...] = mle(data,'pdf',pdf,'cdf',cdf,'start',start,...)
[...] = mle(data,'logpdf',logpdf,'logsf',logsf,'start',start,
...)
[...] = mle(data,'nloglf',nloglf,'start',start,...)
Description phat = mle(data) returns maximum likelihood estimates (MLEs) for
the parameters of a normal distribution, computed using the sample
data in the vector data.
[phat,pci] = mle(data) returns MLEs and 95% confidence intervals
for the parameters.
[...] = mle(data,'distribution',dist) computes parameter
estimates for the distribution specified by dist. Acceptable strings
for dist are:
• 'beta'
• 'bernoulli'
• 'binomial'
• 'birnbaumsaunders'
• 'discrete uniform' or 'unid'
• 'exponential'
• 'extreme value' or 'ev'
• 'gamma'
• 'generalized extreme value' or 'gev'
• 'generalized pareto' or 'gp'
• 'geometric'
18-762
mle
• 'inversegaussian'
• 'logistic'
• 'loglogistic'
• 'lognormal'
• 'nakagami'
• 'negative binomial' or 'nbin'
• 'normal'
• 'poisson'
• 'rayleigh'
• 'rician'
• 'tlocationscale'
• 'uniform'
• 'weibull' or 'wbl'
[...] = mle(data,...,name1,val1,name2,val2,...) specifies
optional argument name/value pairs chosen from the following list.
Name Value
'censoring' A Boolean vector of the same size as data,
containing ones when the corresponding
elements of data are right-censored
observations and zeros when the corresponding
elements are exact observations. The default
is that all observations are observed exactly.
Censoring is not supported for all distributions.
'frequency' A vector of the same size as data, containing
nonnegative integer frequencies for the
corresponding elements in data. The default is
one observation per element of data.
18-763
mle
Name Value
'alpha' A value between 0 and 1 specifying a confidence
level of 100(1-alpha)% for pci. The default is
0.05 for 95% confidence.
'ntrials' A scalar, or a vector of the same size as data,
containing the total number of trials for the
corresponding element of data. Applies only to
the binomial distribution.
'options' A structure created by a call to statset,
containing numerical options for the fitting
algorithm. Not applicable to all distributions.
mle can also fit custom distributions that you define using distribution
functions, in one of three ways.
[...] = mle(data,'pdf',pdf,'cdf',cdf,'start',start,...)
returns MLEs for the parameters of the distribution defined by the
probability density and cumulative distribution functions pdf and
cdf. pdf and cdf are function handles created using the @ sign. They
accept as inputs a vector data and one or more individual distribution
parameters, and return vectors of probability density values and
cumulative probability values, respectively. If the 'censoring'
name/value pair is not present, you can omit the 'cdf' name/value
pair. mle computes the estimates by numerically maximizing the
distribution’s log-likelihood, and start is a vector containing initial
values for the parameters.
[...] =
mle(data,'logpdf',logpdf,'logsf',logsf,'start',start,...)
returns MLEs for the parameters of the distribution defined by the log
probability density and log survival functions logpdf and logsf.
logpdf and logsf are function handles created using the @ sign. They
accept as inputs a vector data and one or more individual distribution
parameters, and return vectors of logged probability density values and
logged survival function values, respectively. This form is sometimes
more robust to the choice of starting point than using pdf and cdf
18-764
mle
functions. If the 'censoring' name/value pair is not present, you can
omit the 'logsf' name/value pair. start is a vector containing initial
values for the distribution’s parameters.
[...] = mle(data,'nloglf',nloglf,'start',start,...)
returns MLEs for the parameters of the distribution whose negative
log-likelihood is given by nloglf. nloglf is a function handle, specified
using the @ sign, that accepts the four input arguments:
• params - a vector of distribution parameter values
• data - a vector of data
• cens - a Boolean vector of censoring values
• freq - a vector of integer data frequencies
nloglf must accept all four arguments even if you do not supply the
'censoring' or 'frequency' name/value pairs (see above). However,
nloglf can safely ignore its cens and freq arguments in that case.
nloglf returns a scalar negative log-likelihood value and, optionally,
a negative log-likelihood gradient vector (see the 'GradObj' statset
parameter below). start is a vector containing initial values for the
distribution’s parameters.
pdf, cdf, logpdf, logsf, or nloglf can also be cell arrays whose first
element is a function handle as defined above, and whose remaining
elements are additional arguments to the function. mle places these
arguments at the end of the argument list in the function call.
The following optional argument name/value pairs are valid only when
'pdf' and 'cdf', 'logpdf' and 'logcdf', or 'nloglf' are given:
• 'lowerbound' — A vector the same size as start containing lower
bounds for the distribution parameters. The default is -Inf.
• 'upperbound' — A vector the same size as start containing upper
bounds for the distribution parameters. The default is Inf.
• 'optimfun' — A string, either 'fminsearch' or 'fmincon', naming
the optimization function to be used in maximizing the likelihood.
18-765
mle
The default is 'fminsearch'. You can only specify 'fmincon' if
Optimization Toolbox software is available.
When fitting a custom distribution, use the 'options' parameter
to control details of the maximum likelihood optimization. See
statset('mlecustom') for parameter names and default values. mle
interprets the following statset parameters for custom distribution
fitting as follows:
Parameter Value
'GradObj' 'on' or 'off', indicating whether or not fmincon
can expect the function provided with the 'nloglf'
name/value pair to return the gradient vector of
the negative log-likelihood as a second output. The
default is 'off'. Ignored when using fminsearch.
'DerivStep' The relative difference used in finite difference
derivative approximations when using fmincon, and
'GradObj' is 'off'. 'DerivStep' can be a scalar, or
the same size as 'start'. The default is eps^(1/3).
Ignored when using fminsearch.
'FunValCheck' 'on' or 'off', indicating whether or not mle should
check the values returned by the custom distribution
functions for validity. The default is 'on'. A poor
choice of starting point can sometimes cause these
functions to return NaNs, infinite values, or out of
range values if they are written without suitable
error-checking.
'TolBnd' An offset for upper and lower bounds when using
fmincon. mle treats upper and lower bounds as
strict inequalities (i.e., open bounds). With fmincon,
this is approximated by creating closed bounds
inset from the specified upper and lower bounds by
TolBnd. The default is 1e-6.
18-766
mle
Examples The following returns an MLE and a 95% confidence interval for the
success probability of a binomial distribution with 20 trials:
data = binornd(20,0.75,100,1); % Simulated data, p = 0.75
[phat,pci] = mle(data,'distribution','binomial',...
'alpha',.05,'ntrials',20)
phat =
0.7370
pci =
0.7171
0.7562
See Also betafit, binofit, evfit, expfit, gamfit, gevfit, gpfit, lognfit,
nbinfit, normfit, mlecov, poissfit, raylfit, statset, unifit,
wblfit
18-767
mlecov
Purpose Asymptotic covariance of maximum likelihood estimators
Syntax ACOV = mlecov(params,data,...)
ACOV = mlecov(params,data,'pdf',pdf,'cdf',cdf)
ACOV = mlecov(params,data,'logpdf',logpdf,'logsf',logsf)
ACOV = mlecov(params,data,'nloglf',nloglf)
[...] = mlecov(params,data,...,param1,val1,param2,val2,...)
Description ACOV = mlecov(params,data,...) returns an approximation to the
asymptotic covariance matrix of the maximum likelihood estimators of
the parameters for a specified distribution. The following paragraphs
describe how to specify the distribution. mlecov computes a finite
difference approximation to the Hessian of the log-likelihood at the
maximum likelihood estimates params, given the observed data, and
returns the negative inverse of that Hessian. ACOV is a p-by-p matrix,
where p is the number of elements in params.
You must specify a distribution after the input argument data, as
follows.
ACOV = mlecov(params,data,'pdf',pdf,'cdf',cdf) enables you
to define a distribution by its probability density and cumulative
distribution functions, pdf and cdf, respectively. pdf and cdf are
function handles that you create using the @ sign. They accept a vector
of data and one or more individual distribution parameters as inputs
and return vectors of probability density function values and cumulative
distribution values, respectively. If the 'censoring' name/value pair
(see below) is not present, you can omit the 'cdf' name/value pair.
ACOV = mlecov(params,data,'logpdf',logpdf,'logsf',logsf)
enables you to define a distribution by its log probability density and
log survival functions, logpdf and logsf, respectively. logpdf and
logsf are function handles that you create using the @ sign. They
accept as inputs a vector of data and one or more individual distribution
parameters, and return vectors of logged probability density values
and logged survival function values, respectively. If the 'censoring'
name/value pair (see below) is not present, you can omit the 'logsf'
name/value pair.
18-768
mlecov
ACOV = mlecov(params,data,'nloglf',nloglf) enables you to define
a distribution by its log-likelihood function. nloglf is a function
handle, specified using the @ sign, that accepts the following four input
arguments:
• params — Vector of distribution parameter values
• data — Vector of data
• cens — Boolean vector of censoring values
• freq — Vector of integer data frequencies
nloglf must accept all four arguments even if you do not supply the
'censoring' or 'frequency' name/value pairs (see below). However,
nloglf can safely ignore its cens and freq arguments in that case.
nloglf returns a scalar negative log-likelihood value and, optionally,
the negative log-likelihood gradient vector (see the 'gradient'
name/value pair below).
pdf, cdf, logpdf, logsf, and nloglf can also be cell arrays whose first
element is a function handle, as defined above, and whose remaining
elements are additional arguments to the function. The mle function
places these arguments at the end of the argument list in the function
call.
[...] =
mlecov(params,data,...,param1,val1,param2,val2,...) specifies
optional parameter name/value pairs chosen from the following table.
18-769
mlecov
Parameter Value
'censoring' Boolean vector of the same size as data, containing
1’s when the corresponding elements of data are
right-censored observations and 0’s when the
corresponding elements are exact observations. The
default is that all observations are observed exactly.
Censoring is not supported for all distributions.
'frequency' A vector of the same size as data containing
nonnegative frequencies for the corresponding
elements in data. The default is one observation per
element of data.
'options' A structure opts containing numerical options for
the finite difference Hessian calculation. You create
opts by calling statset. The applicable statset
parameters are:
• 'GradObj' — 'on' or 'off', indicating whether
or not the function provided with the 'nloglf'
name/value pair can return the gradient vector of
the negative log-likelihood as its second output. The
default is 'off'.
• 'DerivStep' — Relative step size used in finite
difference for Hessian calculations. Can be a
scalar, or the same size as params. The default is
eps^(1/4). A smaller value might be appropriate if
'GradObj' is 'on'.
Examples Create the following function:
function logpdf = betalogpdf(x,a,b)
logpdf = (a-1)*log(x)+(b-1)*log(1-x)-betaln(a,b);
Fit a beta distribution to some simulated data, and compute the
approximate covariance matrix of the parameter estimates:
18-770
mlecov
x = betarnd(1.23,3.45,25,1);
phat = mle(x,'dist','beta')
acov = mlecov(phat,x,'logpdf',@betalogpdf)
See Also mle
18-771
mnpdf
Purpose Multinomial probability density function
Syntax Y = mnpdf(X,PROB)
Description Y = mnpdf(X,PROB) returns the pdf for the multinomial distribution
with probabilities PROB, evaluated at each row of X. X and PROB are
m-by-k matrices or 1-by-k vectors, where k is the number of multinomial
bins or categories. Each row of PROB must sum to one, and the sample
sizes for each observation (rows of X) are given by the row sums
sum(X,2). Y is an m-by-k matrix, and mnpdf computes each row of Y
using the corresponding rows of the inputs, or replicates them if needed.
Examples % Compute the distribution
p = [1/2 1/3 1/6]; % Outcome probabilities
n = 10; % Sample size
x1 = 0:n;
x2 = 0:n;
[X1,X2] = meshgrid(x1,x2);
X3 = n-(X1+X2);
Y = mnpdf([X1(:),X2(:),X3(:)],repmat(p,(n+1)^2,1));
% Plot the distribution
Y = reshape(Y,n+1,n+1);
bar3(Y)
set(gca,'XTickLabel',0:n)
set(gca,'YTickLabel',0:n)
xlabel('x_1')
ylabel('x_2')
zlabel('Probability Mass')
title('Trinomial Distribution')
18-772
mnpdf
Note that the visualization does not show x3, which is determined by
the constraint x1 + x2 + x3 = n.
See Also mnrnd
“Multinomial Distribution” on page B-54
18-773
mnrfit
Purpose Multinomial logistic regression
Syntax B = mnrfit(X,Y)
B = mnrfit(X,Y,param1,val1,param2,val2,...)
[B,dev] = mnrfit(...)
[B,dev,stats] = mnrfit(...)
Description B = mnrfit(X,Y) returns a matrix B of coefficient estimates for a
multinomial logistic regression of the responses in Y on the predictors
in X. X is an n-by-p matrix of p predictors at each of n observations.
Y is an n-by-k matrix, where Y(i,j) is the number of outcomes of
the multinomial category j for the predictor combinations given by
X(i,:). The sample sizes for each observation are given by the row
sums sum(Y,2).
Alternatively, Y can be an n-by-1 column vector of scalar integers from
1 to k indicating the value of the response for each observation, and
all sample sizes are taken to be 1.
The result B is a (p+1)-by-(k–1) matrix of estimates, where each column
corresponds to the estimated intercept term and predictor coefficients,
one for each of the first k–1 multinomial categories. The estimates for
the kth category are taken to be zero.
Note mnrfit automatically includes a constant term in all models. Do
not enter a column of 1s directly into X.
mnrfit treats NaNs in either X or Y as missing values, and ignores them.
B = mnrfit(X,Y,param1,val1,param2,val2,...) allows you to
specify optional parameter name/value pairs to control the model fit.
Parameters are:
• 'model' — The type of model to fit; one of the text strings 'nominal'
(the default), 'ordinal', or 'hierarchical'
18-774
mnrfit
• 'interactions' — Determines whether the model includes an
interaction between the multinomial categories and the coefficients.
Specify as 'off' to fit a model with a common set of coefficients
for the predictor variables, across all multinomial categories. This
is often described as parallel regression. Specify as 'on' to fit a
model with different coefficients across categories. In all cases,
the model has different intercepts across categories. Thus, B is a
vector containing k–1+p coefficient estimates when 'interaction'
is 'off', and a (p+1)-by-(k–1) matrix when it is 'on'. The default
is 'off' for ordinal models, and 'on' for nominal and hierarchical
models.
• 'link' — The link function to use for ordinal and hierarchical
models. The link function defines the relationship g(μij) = xibj between
the mean response for the ith observation in the jth category, μij , and
the linear combination of predictors xibj. Specify the link parameter
value as one of the text strings 'logit'(the default), 'probit',
'comploglog', or 'loglog'. You may not specify the 'link'
parameter for nominal models; these always use a multivariate
logistic link.
• 'estdisp' — Specify as 'on' to estimate a dispersion parameter for
the multinomial distribution in computing standard errors, or 'off'
(the default) to use the theoretical dispersion value of 1.
[B,dev] = mnrfit(...) returns the deviance of the fit dev.
[B,dev,stats] = mnrfit(...) returns a structure stats that
contains the following fields:
• dfe — Degrees of freedom for error
• s — Theoretical or estimated dispersion parameter
• sfit — Estimated dispersion parameter
• se — Standard errors of coefficient estimates B
• coeffcorr — Estimated correlation matrix for B
• covb — Estimated covariance matrix for B
18-775
mnrfit
• t — t statistics for B
• p — p-values for B
• resid — Residuals
• residp — Pearson residuals
• residd — Deviance residuals
Examples Fit multinomial logistic regression models to data with one predictor
variable and three categories in the response variable:
x = [-3 -2 -1 0 1 2 3]';
Y = [1 11 13; 2 9 14; 6 14 5; 5 10 10; 5 14 6; 7 13 5;...
8 11 6];
bar(x,Y,'stacked'); ylim([0 25]);
18-776
mnrfit
% Now fit a nominal model for the individual response
% category probabilities, with separate slopes on the
% single predictor variable, x, for each
% category:
% The first row of betaHatNom contains the intercept terms
% for the first two response categories. The second row
% contains the slopes.
betaHatNom = mnrfit(x,Y,'model','nominal',...
'interactions','on')
% Compute the predicted probabilities for the three
18-777
mnrfit
% response categories:
xx = linspace(-4,4)';
pHatNom = mnrval(betaHatNom,xx,'model','nominal',...
'interactions','on');
line(xx,cumsum(25*pHatNom,2),'LineWidth',2);
18-778
mnrfit
Fit a "parallel" ordinal model for the cumulative response category
probabilities, with a common slope on the single predictor variable, x,
across all categories:
% The first two elements of betaHatOrd are the
% intercept terms for the first two response categories.
% The last element of betaHatOrd is the common slope.
betaHatOrd = mnrfit(x,Y,'model','ordinal',...
'interactions','off')
% Compute the predicted cumulative probabilities for the
% first two response categories. The cumulative
% probability for the third category is always 1.
pHatOrd = mnrval(betaHatOrd,xx,'type','cumulative',...
'model','ordinal','interactions','off');
bar(x,cumsum(Y,2),'grouped'); ylim([0 25]);
line(xx,25*pHatOrd,'LineWidth',2);
18-779
mnrfit
References [1] McCullagh, P., and J. A. Nelder. Generalized Linear Models. New
York: Chapman & Hall, 1990.
See Also mnrval, glmfit, glmval, regress, regstats
“Multinomial Distribution” on page B-54
18-780
mnrnd
Purpose Multinomial random numbers
Syntax r = mnrnd(n,p)
R = mnrnd(n,p,m)
R = mnrnd(N,P)
Description r = mnrnd(n,p) returns random values r from the multinomial
distribution with parameters n and p. n is a positive integer specifying
the number of trials (sample size) for each multinomial outcome. p is
a 1-by-k vector of multinomial probabilities, where k is the number of
multinomial bins or categories. p must sum to one. (If p does not sum to
one, r consists entirely of NaN values.) r is a 1-by-k vector, containing
counts for each of the k multinomial bins.
R = mnrnd(n,p,m) returns m random vectors from the multinomial
distribution with parameters n and p. R is a m-by-k matrix, where
k is the number of multinomial bins or categories. Each row of R
corresponds to one multinomial outcome.
R = mnrnd(N,P) generates outcomes from different multinomial
distributions. P is a m-by-k matrix, where k is the number of
multinomial bins or categories and each of the m rows contains a
different set of multinomial probabilities. Each row of P must sum to
one. (If any row of P does not sum to one, the corresponding row of R
consists entirely of NaN values.) N is a m-by-1 vector of positive integers
or a single positive integer (replicated by mnrnd to a m-by-1 vector). R
is a m-by-k matrix. Each row of R is generated using the corresponding
rows of N and P.
Examples Generate 2 random vectors with the same probabilities:
n = 1e3;
p = [0.2,0.3,0.5];
R = mnrnd(n,p,2)
R =
215 282 503
194 303 503
18-781
mnrnd
Generate 2 random vectors with different probabilities:
n = 1e3;
P = [0.2, 0.3, 0.5; ...
0.3, 0.4, 0.3;];
R = mnrnd(n,P)
R =
186 290 524
290 389 321
See Also mnpdf
“Multinomial Distribution” on page B-54
18-782
mnrval
Purpose Multinomial logistic regression values
Syntax PHAT = mnrval(B,X)
YHAT = mnrval(B,X,ssize)
[...,DLO,DHI] = mnrval(B,X,...,stats)
[...] = mnrval(...,param1,val1,param2,val2,...)
Description PHAT = mnrval(B,X) computes predicted probabilities for the
multinomial logistic regression model with predictors X. B contains
intercept and coefficient estimates as returned by the mnrfit function.
X is an n-by-p matrix of p predictors at each of n observations. PHAT is an
n-by-k matrix of predicted probabilities for each multinomial category.
Note mnrval automatically includes a constant term in all models. Do
not enter a column of 1s directly into X.
YHAT = mnrval(B,X,ssize) computes predicted category counts for
sample sizes ssize. ssize is an n-by-1 column vector of positive
integers.
[...,DLO,DHI] = mnrval(B,X,...,stats) also computes 95%
confidence bounds on the predicted probabilities PHAT or counts YHAT.
stats is the structure returned by the mnrfit function. DLO and DHI
define a lower confidence bound of PHAT or YHAT minus DLO and an
upper confidence bound of PHAT or YHAT plus DHI. Confidence bounds
are nonsimultaneous and they apply to the fitted curve, not to new
observations.
[...] = mnrval(...,param1,val1,param2,val2,...) allows you to
specify optional parameter name/value pairs to control the predicted
values. These parameters must be set to the corresponding values used
with the mnrfit function to compute B. Parameters are:
• 'model' — The type of model that was fit by mnrfit; one of the text
strings 'nominal' (the default), 'ordinal', or 'hierarchical'.
18-783
mnrval
• 'interactions' — Determines whether the model fit by mnrfit
included an interaction between the multinomial categories and the
coefficients. The default is 'off' for ordinal models, and 'on' for
nominal and hierarchical models.
• 'link' — The link function that was used by mnrfit for ordinal
and hierarchical models. Specify the link parameter value as one of
the text strings 'logit'(the default), 'probit', 'comploglog', or
'loglog'. You may not specify the 'link' parameter for nominal
models; these always use a multivariate logistic link.
• 'type' — Set to 'category' (the default) to return predictions
and confidence bounds for the probabilities (or counts) of the k
multinomial categories. Set to 'cumulative' to return predictions
and confidence bounds for the cumulative probabilities (or counts) of
the first k–1 multinomial categories, as an n-by-(k–1) matrix. The
predicted cumulative probability for the kth category is 1. Set to
'conditional' to return predictions and confidence bounds in terms
of the first k–1 conditional category probabilities, i.e., the probability
for category j, given an outcome in category j or higher. When
'type' is 'conditional', and you supply the sample size argument
ssize, the predicted counts at each row of X are conditioned on the
corresponding element of ssize, across all categories.
• 'confidence' — The confidence level for the confidence bounds; a
value between 0 and 1. The default is 0.95.
Examples Fit multinomial logistic regression models to data with one predictor
variable and three categories in the response variable:
x = [-3 -2 -1 0 1 2 3]';
Y = [1 11 13; 2 9 14; 6 14 5; 5 10 10; 5 14 6; 7 13 5;...
8 11 6];
bar(x,Y,'stacked');
ylim([0 25]);
18-784
mnrval
% Now fit a nominal model for the individual response
% category probabilities, with separate slopes on the
% single predictor variable, x, for each
% category:
% The first row of betaHatNom contains the intercept terms
% for the first two response categories. The second row
% contains the slopes.
betaHatNom = mnrfit(x,Y,'model','nominal',...
'interactions','on')
% Compute the predicted probabilities for the three
% response categories:
18-785
mnrval
xx = linspace(-4,4)';
pHatNom = mnrval(betaHatNom,xx,'model','nominal',...
'interactions','on');
line(xx,cumsum(25*pHatNom,2),'LineWidth',2);
18-786
mnrval
Fit a "parallel" ordinal model for the cumulative response category
probabilities, with a common slope on the single predictor variable, x,
across all categories:
% The first two elements of betaHatOrd are the
% intercept terms for the first two response categories.
% The last element of betaHatOrd is the common slope.
betaHatOrd = mnrfit(x,Y,'model','ordinal',...
'interactions','off')
% Compute the predicted cumulative probabilities for the
% first two response categories. The cumulative
% probability for the third category is always 1.
pHatOrd = mnrval(betaHatOrd,xx,'type','cumulative',...
'model','ordinal','interactions','off');
bar(x,cumsum(Y,2),'grouped');
ylim([0 25]);
line(xx,25*pHatOrd,'LineWidth',2);
18-787
mnrval
References [1] McCullagh, P., and J. A. Nelder. Generalized Linear Models. New
York: Chapman & Hall, 1990.
See Also mnrfit, glmfit, glmval
“Multinomial Distribution” on page B-54
18-788
moment
Purpose Central moments
Syntax m = moment(X,order)
moment(X,order,dim)
Description m = moment(X,order) returns the central sample moment of X specified
by the positive integer order. For vectors, moment(x,order) returns
the central moment of the specified order for the elements of x. For
matrices, moment(X,order) returns central moment of the specified
order for each column. For N-dimensional arrays, moment operates
along the first nonsingleton dimension of X.
moment(X,order,dim) takes the moment along dimension dim of X.
Remarks Note that the central first moment is zero, and the second central
moment is the variance computed using a divisor of n rather than n –
1, where n is the length of the vector x or the number of rows in the
matrix X.
The central moment of order k of a distribution is defined as
mk = E( x − ) k
where E(x) is the expected value of x.
Examples X = randn([6 5])
X =
1.1650 0.0591 1.2460 -1.2704 -0.0562
0.6268 1.7971 -0.6390 0.9846 0.5135
0.0751 0.2641 0.5774 -0.0449 0.3967
0.3516 0.8717 -0.3600 -0.7989 0.7562
-0.6965 -1.4462 -0.1356 -0.7652 0.4005
1.6961 -0.7012 -1.3493 0.8617 -1.3414
m = moment(X,3)
m =
-0.0282 0.0571 0.1253 0.1460 -0.4486
18-789
moment
See Also kurtosis, mean, skewness, std, var
18-790
gmdistribution.Mu property
Purpose Input matrix of means MU
Description Input matrix of means mu.
18-791
multcompare
Purpose Multiple comparison test
Syntax c = multcompare(stats)
c = multcompare(stats,param1,val1,param2,val2,...)
[c,m] = multcompare(...)
[c,m,h] = multcompare(...)
[c,m,h,gnames] = multcompare(...)
Description c = multcompare(stats) performs a multiple comparison test using
the information in the stats structure, and returns a matrix c of
pairwise comparison results. It also displays an interactive graph of the
estimates with comparison intervals around them. See “Examples”
on page 18-797.
In a one-way analysis of variance, you compare the means of several
groups to test the hypothesis that they are all the same, against the
general alternative that they are not all the same. Sometimes this
alternative may be too general. You may need information about which
pairs of means are significantly different, and which are not. A test that
can provide such information is called a multiple comparison procedure.
When you perform a simple t-test of one group mean against another,
you specify a significance level that determines the cutoff value of the
t statistic. For example, you can specify the value alpha = 0.05 to
insure that when there is no real difference, you will incorrectly find
a significant difference no more than 5% of the time. When there are
many group means, there are also many pairs to compare. If you
applied an ordinary t-test in this situation, the alpha value would apply
to each comparison, so the chance of incorrectly finding a significant
difference would increase with the number of comparisons. Multiple
comparison procedures are designed to provide an upper bound on the
probability that any comparison will be incorrectly found significant.
The output c contains the results of the test in the form of a five-column
matrix. Each row of the matrix represents one test, and there is one
row for each pair of groups. The entries in the row indicate the means
being compared, the estimated difference in means, and a confidence
interval for the difference.
18-792
multcompare
For example, suppose one row contains the following entries.
2.0000 5.0000 1.9442 8.2206 14.4971
These numbers indicate that the mean of group 2 minus the mean of
group 5 is estimated to be 8.2206, and a 95% confidence interval for
the true mean is [1.9442, 14.4971].
In this example the confidence interval does not contain 0.0, so the
difference is significant at the 0.05 level. If the confidence interval did
contain 0.0, the difference would not be significant at the 0.05 level.
The multcompare function also displays a graph with each group mean
represented by a symbol and an interval around the symbol. Two means
are significantly different if their intervals are disjoint, and are not
significantly different if their intervals overlap. You can use the mouse
to select any group, and the graph will highlight any other groups that
are significantly different from it.
c = multcompare(stats,param1,val1,param2,val2,...) specifies
one or more of the parameter name/value pairs described in the
following table.
Parameter Values
'alpha' Scalar between 0 and 1 that determines the
confidence levels of the intervals in the matrix
c and in the figure (default is 0.05). The
confidence level is 100(1-alpha)%.
'display' Either 'on' (the default) to display a graph
of the estimates with comparison intervals
around them, or 'off' to omit the graph. See
“Examples” on page 18-797.
'ctype' Specifies the type of critical value to use for the
multiple comparison. “Values of ctype” on page
18-795 describes the allowed values for ctype.
18-793
multcompare
Parameter Values
'dimension' A vector specifying the dimension or dimensions
over which the population marginal means
are to be calculated. Use only if you create
stats with the function anovan. The default
is 1 to compute over the first dimension. See
“Dimension Parameter” on page 18-797 for more
information.
'estimate' Specifies the estimate to be compared. The
allowable values of estimate depend on the
function that was the source of the stats
structure, as described in “Values of estimate”
on page 18-796
[c,m] = multcompare(...) returns an additional matrix m. The first
column of m contains the estimated values of the means (or whatever
statistics are being compared) for each group, and the second column
contains their standard errors.
[c,m,h] = multcompare(...) returns a handle h to the comparison
graph. Note that the title of this graph contains instructions for
interacting with the graph, and the x-axis label contains information
about which means are significantly different from the selected mean.
If you plan to use this graph for presentation, you may want to omit
the title and the x-axis label. You can remove them using interactive
features of the graph window, or you can use the following commands.
title('')
xlabel('')
[c,m,h,gnames] = multcompare(...) returns gnames, a cell array
with one row for each group, containing the names of the groups.
The intervals shown in the graph are computed so that to a very close
approximation, two estimates being compared are significantly different
if their intervals are disjoint, and are not significantly different if their
intervals overlap. (This is exact for multiple comparison of means from
18-794
multcompare
anova1, if all means are based on the same sample size.) You can click
on any estimate to see which means are significantly different from it.
Values of ctype
The following table describes the allowed values for the parameter
ctype.
Value Description
'hsd' or Use Tukey’s honestly significant difference
'tukey-kramer' criterion. This is the default, and it is based on the
Studentized range distribution. It is optimal for
balanced one-way ANOVA and similar procedures
with equal sample sizes. It has been proven
to be conservative for one-way ANOVA with
different sample sizes. According to the unproven
Tukey-Kramer conjecture, it is also accurate for
problems where the quantities being compared
are correlated, as in analysis of covariance with
unbalanced covariate values.
'lsd' Use Tukey’s least significant difference procedure.
This procedure is a simple t-test. It is reasonable
if the preliminary test (say, the one-way ANOVA
F statistic) shows a significant difference. If it is
used unconditionally, it provides no protection
against multiple comparisons.
'bonferroni' Use critical values from the t distribution, after a
Bonferroni adjustment to compensate for multiple
comparisons. This procedure is conservative, but
usually less so than the Scheffé procedure.
18-795
multcompare
Value Description
'dunn-sidak' Use critical values from the t distribution, after
an adjustment for multiple comparisons that was
proposed by Dunn and proved accurate by Sidák.
This procedure is similar to, but less conservative
than, the Bonferroni procedure.
'scheffe' Use critical values from Scheffé’s S procedure,
derived from the F distribution. This procedure
provides a simultaneous confidence level for
comparisons of all linear combinations of the
means, and it is conservative for comparisons of
simple differences of pairs.
Values of estimate
The allowable values of the parameter 'estimate' depend on the
function that was the source of the stats structure, according to the
following table.
Source Values
'anova1' Ignored. Always compare the group means.
'anova2' Either 'column' (the default) or 'row' to
compare column or row means.
'anovan' Ignored. Always compare the population
marginal means as specified by the dim
argument.
'aoctool' Either 'slope', 'intercept', or 'pmm' to
compare slopes, intercepts, or population
marginal means. If the analysis of covariance
model did not include separate slopes, then
'slope' is not allowed. If it did not include
separate intercepts, then no comparisons are
possible.
18-796
multcompare
Source Values
'friedman' Ignored. Always compare average column ranks.
'kruskalwallis' Ignored. Always compare average group ranks.
Dimension Parameter
The dimension parameter is a vector specifying the dimension or
dimensions over which the population marginal means are to be
calculated. For example, if dim = 1, the estimates that are compared
are the means for each value of the first grouping variable, adjusted by
removing effects of the other grouping variables as if the design were
balanced. If dim = [1 3], population marginal means are computed for
each combination of the first and third grouping variables, removing
effects of the second grouping variable. If you fit a singular model, some
cell means may not be estimable and any population marginal means
that depend on those cell means will have the value NaN.
Population marginal means are described by Milliken and Johnson
(1992) and by Searle, Speed, and Milliken (1980). The idea behind
population marginal means is to remove any effect of an unbalanced
design by fixing the values of the factors specified by dim, and averaging
out the effects of other factors as if each factor combination occurred
the same number of times. The definition of population marginal
means does not depend on the number of observations at each
factor combination. For designed experiments where the number of
observations at each factor combination has no meaning, population
marginal means can be easier to interpret than simple means ignoring
other factors. For surveys and other studies where the number of
observations at each combination does have meaning, population
marginal means may be harder to interpret.
Examples Example 1
The following example performs a 1-way analysis of variance (ANOVA)
and displays group means with their names.
load carsmall
18-797
multcompare
[p,t,st] = anova1(MPG,Origin,'off');
[c,m,h,nms] = multcompare(st,'display','off');
[nms num2cell(m)]
ans =
'USA' [21.1328] [0.8814]
'Japan' [31.8000] [1.8206]
'Germany' [28.4444] [2.3504]
'France' [23.6667] [4.0711]
'Sweden' [22.5000] [4.9860]
'Italy' [28.0000] [7.0513]
multcompare also displays the following graph of the estimates with
comparison intervals around them.
You can click the graphs of each country to compare its mean to those of
other countries.
18-798
multcompare
Example 2
The following continues the example described in the anova1 reference
page, which is related to testing the material strength in structural
beams. From the anova1 output you found significant evidence that
the three types of beams are not equivalent in strength. Now you can
determine where those differences lie. First you create the data arrays
and you perform one-way ANOVA.
strength = [82 86 79 83 84 85 86 87 74 82 ...
78 75 76 77 79 79 77 78 82 79];
alloy = {'st','st','st','st','st','st','st','st',...
'al1','al1','al1','al1','al1','al1',...
'al2','al2','al2','al2','al2','al2'};
[p,a,s] = anova1(strength,alloy);
Among the outputs is a structure that you can use as input to
multcompare.
[c,m,h,nms] = multcompare(s);
[nms num2cell(c)]
ans =
'st' [1] [2] [ 3.6064] [ 7] [10.3936]
'al1' [1] [3] [ 1.6064] [ 5] [ 8.3936]
'al2' [2] [3] [-5.6280] [-2] [ 1.6280]
18-799
multcompare
The third row of the output matrix shows that the differences in
strength between the two alloys is not significant. A 95% confidence
interval for the difference is [-5.6, 1.6], so you cannot reject the
hypothesis that the true difference is zero.
The first two rows show that both comparisons involving the first group
(steel) have confidence intervals that do not include zero. In other
words, those differences are significant. The graph shows the same
information.
See Also anova1, anova2, anovan, aoctool, friedman, kruskalwallis
References [1] Hochberg, Y., and A. C. Tamhane. Multiple Comparison Procedures.
Hoboken, NJ: John Wiley & Sons, 1987.
[2] Milliken, G. A., and D. E. Johnson. Analysis of Messy Data, Volume
1: Designed Experiments. Boca Raton, FL: Chapman & Hall/CRC Press,
1992.
18-800
multcompare
[3] Searle, S. R., F. M. Speed, and G. A. Milliken. “Population marginal
means in the linear model: an alternative to least-squares means.”
American Statistician. 1980, pp. 216–221.
18-801
multivarichart
Purpose Multivari chart for grouped data
Syntax multivarichart(y,GROUP)
multivarichart(Y)
multivarichart(...,param1,val1,param2,val2,...)
[charthandle,AXESH] = multivarichart(...)
Description multivarichart(y,GROUP) displays the multivari chart for the vector
y grouped by entries in the cell array GROUP. Each cell of GROUP must
contain a grouping variable that can be a categorical variable, numeric
vector, character matrix, or single-column cell array of strings. (See
“Grouped Data” on page 2-34.) GROUP can also be a matrix whose
columns represent different grouping variables. Each grouping variable
must have the same number of elements as y. The number of grouping
variables must be 2, 3, or 4.
Each subplot of the plot matrix contains a multivari chart for the first
and second grouping variables. The x-axis in each subplot indicates
values of the first grouping variable. The legend at the bottom of the
figure window indicates values of the second grouping variable. The
subplot at position (i,j) is the multivari chart for the subset of y at the
ith level of the third grouping variable and the jth level of the fourth
grouping variable. If the third or fourth grouping variable is absent, it
is considered to have only one level.
multivarichart(Y) displays the multivari chart for a matrix Y. The
data in different columns represent changes in one factor. The data in
different rows represent changes in another factor.
multivarichart(...,param1,val1,param2,val2,...) specifies one
or more of the following name/value pairs:
• 'varnames' — Grouping variable names in a character matrix or a
cell array of strings, one per grouping variable. Default names are
'X1', 'X2', ... .
• 'plotorder' — A string with the value 'sorted' or a vector
containing a permutation of the integers from 1 to the number of
grouping variables.
18-802
multivarichart
If 'plotorder' is a string with value 'sorted', the grouping
variables are rearranged in descending order according to the
number of levels in each variable.
If 'plotorder' is a vector, it indicates the order in which each
grouping variable should be plotted. For example, [2,3,1,4]
indicates that the second grouping variable should be used as the
x-axis of each subplot, the third grouping variable should be used as
the legend, the first grouping variable should be used as the columns
of the plot, and the fourth grouping variable should be used as the
rows of the plot.
[charthandle,AXESH] = multivarichart(...) returns a handle
charthandle to the figure window and a matrix AXESH of handles to
the subplot axes.
Examples Display a multivari chart for data with two grouping variables:
y = randn(100,1); % response
group = [ceil(3*rand(100,1)) ceil(2*rand(100,1))];
multivarichart(y,group)
18-803
multivarichart
Display a multivari chart for data with four grouping variables:
y = randn(1000,1); % response
group = {ceil(2*rand(1000,1)),ceil(3*rand(1000,1)), ...
ceil(2*rand(1000,1)),ceil(3*rand(1000,1))};
multivarichart(y,group)
18-804
multivarichart
See Also “Grouped Data” on page 2-34
maineffectsplot, interactionplot
18-805
mvncdf
Purpose Multivariate normal cumulative distribution function
Syntax y = mvncdf(X)
y = mvncdf(X,mu,SIGMA)
y = mvncdf(xl,xu,mu,SIGMA)
[y,err] = mvncdf(...)
[...] = mvncdf(...,options)
Description y = mvncdf(X) returns the cumulative probability of the multivariate
normal distribution with zero mean and identity covariance matrix,
evaluated at each row of X. Rows of the n-by-d matrix X correspond
to observations or points, and columns correspond to variables or
coordinates. y is an n-by-1 vector.
y = mvncdf(X,mu,SIGMA) returns the cumulative probability of the
multivariate normal distribution with mean mu and covariance SIGMA,
evaluated at each row of X. mu is a 1-by-d vector, and SIGMA is a d-by-d
symmetric, positive definite matrix. mu can also be a scalar value,
which mvncdf replicates to match the size of X. If the covariance
matrix is diagonal, containing variances along the diagonal and zero
covariances off the diagonal, SIGMA may also be specified as a 1-by-d
vector containing just the diagonal. Pass in the empty matrix [] for mu
to use as its default value when you want to only specify SIGMA.
The multivariate normal cumulative probability at X is defined as the
probability that a random vector V, distributed as multivariate normal,
will fall within the semi-infinite rectangle with upper limits defined by
X, for example, Pr{V(1) ≤ X(1),V(2) ≤ X(2),...,V(d) ≤ X(d)}.
y = mvncdf(xl,xu,mu,SIGMA) returns the multivariate normal
cumulative probability evaluated over the rectangle with lower and
upper limits defined by xl and xu, respectively.
[y,err] = mvncdf(...) returns an estimate of the error in y. For
bivariate and trivariate distributions, mvncdf uses adaptive quadrature
on a transformation of the t density, based on methods developed by
Drezner and Wesolowsky and by Genz, as described in the references.
The default absolute error tolerance for these cases is 1e-8. For four
or more dimensions, mvncdf uses a quasi-Monte Carlo integration
18-806
mvncdf
algorithm based on methods developed by Genz and Bretz, as described
in the references. The default absolute error tolerance for these cases
is 1e-4.
[...] = mvncdf(...,options) specifies control parameters for the
numerical integration used to compute y. This argument can be created
by a call to statset. Choices of statset parameters:
• 'TolFun' — Maximum absolute error tolerance. Default is 1e-8
when d < 4, or 1e-4 when d ≥ 4.
• 'MaxFunEvals' — Maximum number of integrand evaluations
allowed when d ≥ 4. Default is 1e7. 'MaxFunEvals' is ignored when
d < 4.
• 'Display' — Level of display output. Choices are 'off' (the
default), 'iter', and 'final'. 'Display' is ignored when d < 4.
Examples mu = [1 -1]; SIGMA = [.9 .4; .4 .3];
[X1,X2] = meshgrid(linspace(-1,3,25)',linspace(-3,1,25)');
X = [X1(:) X2(:)];
p = mvncdf(X,mu,SIGMA);
surf(X1,X2,reshape(p,25,25));
18-807
mvncdf
References [1] Drezner, Z. “Computation of the Trivariate Normal Integral.”
Mathematics of Computation. Vol. 63, 1994, pp. 289–294.
[2] Drezner, Z., and G. O. Wesolowsky. “On the Computation of the
Bivariate Normal Integral.” Journal of Statistical Computation and
Simulation. Vol. 35, 1989, pp. 101–107.
[3] Genz, A. “Numerical Computation of Rectangular Bivariate and
Trivariate Normal and t Probabilities.” Statistics and Computing. Vol.
14, No. 3, 2004, pp. 251–260.
[4] Genz, A., and F. Bretz. “Numerical Computation of Multivariate
t Probabilities with Application to Power Calculation of Multiple
18-808
mvncdf
Contrasts.” Journal of Statistical Computation and Simulation. Vol.
63, 1999, pp. 361–378.
[5] Genz, A., and F. Bretz. “Comparison of Methods for the Computation
of Multivariate t Probabilities.” Journal of Computational and
Graphical Statistics. Vol. 11, No. 4, 2002, pp. 950–971.
See Also mvnpdf, mvnrnd
“Multivariate Normal Distribution” on page B-58
18-809
mvnpdf
Purpose Multivariate normal probability density function
Syntax y = mvnpdf(X)
y = mvnpdf(X,MU)
y = mvnpdf(X,MU,SIGMA)
Description y = mvnpdf(X) returns the n-by-1 vector y, containing the probability
density of the multivariate normal distribution with zero mean and
identity covariance matrix, evaluated at each row of the n-by-d matrix
X. Rows of X correspond to observations and columns correspond to
variables or coordinates.
y = mvnpdf(X,MU) returns the density of the multivariate normal
distribution with mean mu and identity covariance matrix, evaluated
at each row of X. MU is a 1-by-d vector, or an n-by-d matrix. If MU is a
matrix, the density is evaluated for each row of X with the corresponding
row of MU. MU can also be a scalar value, which mvnpdf replicates to
match the size of X.
y = mvnpdf(X,MU,SIGMA) returns the density of the multivariate
normal distribution with mean MU and covariance SIGMA, evaluated
at each row of X. SIGMA is a d-by-d matrix, or a d-by-d-by-n array,
in which case the density is evaluated for each row of X with the
corresponding page of SIGMA, i.e., mvnpdf computes y(i) using X(i,:)
and SIGMA(:,:,i). If the covariance matrix is diagonal, containing
variances along the diagonal and zero covariances off the diagonal,
SIGMA may also be specified as a 1-by-d vector or a 1-by-d-by-n array,
containing just the diagonal. Specify [] for MU to use its default value
when you want to specify only SIGMA.
If X is a 1-by-d vector, mvnpdf replicates it to match the leading
dimension of mu or the trailing dimension of SIGMA.
Examples mu = [1 -1];
SIGMA = [.9 .4; .4 .3];
X = mvnrnd(mu,SIGMA,10);
p = mvnpdf(X,mu,SIGMA);
18-810
mvnpdf
See Also mvncdf, mvnrnd, normpdf
“Multivariate Normal Distribution” on page B-58
18-811
mvregress
Purpose Multivariate linear regression
Syntax beta = mvregress(X,Y)
[beta,SIGMA] = mvregress(X,Y)
[beta,SIGMA,RESID] = mvregress(X,Y)
[beta,SIGMA,RESID,COVB] = mvregress(...)
[beta,SIGMA,RESID,objective] = mvregress(...)
[...] = mvregress(X,Y,param1,val1,param2,val2,...)
Description beta = mvregress(X,Y) returns a vector beta of coefficient estimates
for a multivariate regression of the d-dimensional responses in Y on the
predictors in X. If d = 1, X can be an n-by-p matrix of p predictors at each
of n observations. If d ≥ 1, X can be a cell array of length n, with each
cell containing a d-by-p design matrix for one multivariate observation.
If all observations have the same d-by-p design matrix, X can be a single
cell. Y is n-by-d. beta is p-by-1.
Note To include a constant term in a model, a matrix X should contain
a column of 1s. If X is a cell array X, each cell should contain a matrix
with a column of 1s.
mvregress treats NaNs in X or Y as missing values. Missing values in X
are ignored. Missing values in Y are handled according to the value of
the 'algorithm' parameter described below.
[beta,SIGMA] = mvregress(X,Y) also returns a d-by-d matrix SIGMA
for the estimated covariance of Y.
[beta,SIGMA,RESID] = mvregress(X,Y) also returns an n-by-d matrix
RESID of residuals.
The RESID values corresponding to missing values in Y are the
differences between the conditionally imputed values for Y and the
fitted values. The SIGMA estimate is not the sample covariance matrix
of the RESID matrix.
18-812
mvregress
[beta,SIGMA,RESID,COVB] = mvregress(...) also returns a matrix
COVB for the estimated covariance of the coefficients. By default, or if
the 'varformat' parameter is 'beta' (see below), COVB is the estimated
covariance matrix of beta. If the 'varformat' parameter is 'full',
COVB is the combined estimated covariance matrix for beta and SIGMA.
[beta,SIGMA,RESID,objective] = mvregress(...) also returns the
value of the objective function, or log likelihood, objective, after the
last iteration.
[...] = mvregress(X,Y,param1,val1,param2,val2,...) specifies
additional parameter name/value pairs chosen from the following:
• 'algorithm' — Either 'ecm' to compute the maximum likelihood
estimates via the ECM algorithm, 'cwls' to perform least squares
(optionally conditionally weighted by an input covariance matrix),
or 'mvn' to omit observations with missing data and compute the
ordinary multivariate normal estimates. The default is 'mvn' for
complete data, 'ecm' for missing data when the sample size is
sufficient to estimate all parameters, and 'cwls' otherwise.
• 'covar0' — A d-by-d matrix to be used as the initial estimate for
SIGMA. The default is the identity matrix. For the 'cwls' algorithm,
this matrix is usually a diagonal matrix used as a weighting at each
iteration. The 'cwls' algorithm uses the initial value of SIGMA at
each iteration, without changing it.
• 'covtype' — Either 'full', to allow a full covariance matrix, or
'diagonal', to restrict the covariance matrix to be diagonal. The
default is 'full'.
• 'maxiter' — Maximum number of iterations. The default is 100.
• 'outputfcn' — An output function called with three arguments:
1. A vector of current parameter estimates.
2. A structure with fields 'Covar' for the current value of the
covariance matrix, 'iteration' for the current iteration number,
and 'fval' for the current value of the objective function.
18-813
mvregress
3. A text string that is 'init' when called during initialization,
'iter' when called after an iteration, and 'done' when called after
completion.
The function should return logical true if the iterations should stop,
or logical false if they should continue.
• 'param0' — A vector of p elements to be used as the initial estimate
for beta. Default is a zero vector. Not used for the 'mvn' algorithm.
• 'tolbeta' — Convergence tolerance for beta. The default is
sqrt(eps). Iterations continue until the tolbeta and tolobj
conditions are met. The test for convergence at iteration k is
norm(beta(k)-beta(k-1))<sqrt(p)*tolbeta*(1+norm(beta(k)))
where p = length(beta).
• 'tolobj' — Convergence tolerance for changes in the objective
function. The default is eps^(3/4). The test is
abs(obj(k)-obj(k-1)) < tolobj*(1+abs(obj(k)))
where obj is the objective function. If both tolobj and tolbeta are
0, the function performs maxiter iterations with no convergence test.
• 'beta0' — A vector of p elements to be used as the initial estimate
for beta. Default is a zero vector. Not used for the 'mvn' algorithm.
• 'covar0' — A d-by-d matrix to be used as the initial estimate for
SIGMA. Default is the identity matrix. For the 'cwls' algorithm, this
matrix is usually a diagonal matrix and it is not changed during
the iterations, so the input value is used as the weighting matrix at
each iteration.
• 'outputfcn' — An output function.
• 'varformat' — Either 'beta' to compute COVB for beta only
(default), or 'full' to compute COVB for both beta and SIGMA.
• 'vartype' — Either 'hessian' to compute COVB using the Hessian
or observed information (default), or 'fisher' to compute COVB using
18-814
mvregress
the complete-data Fisher or expected information. The 'hessian'
method takes into account the increased uncertainties due to missing
data, while the 'fisher' method does not.
Examples Predict regional flu estimates based on Google™ queries using the
national CDC estimates as a predictor:
load flu
% response: regional queries
y = double(flu(:,2:end-1));
% predictor: national CDC estimates
x = flu.WtdILI;
[nobs,nregions] = size(y);
% Create and fit model with separate intercepts but
% common slope
X = cell(nobs,1);
for j=1:nobs
X{j} = [eye(nregions), repmat(x(j),nregions,1)];
end
[beta,sig,resid,vars,loglik] = mvregress(X,y);
% Plot raw data with fitted lines
B = [beta(1:nregions)';repmat(beta(end),1,nregions)]
axes1 = axes('Position',[0.13 0.5838 0.6191 0.3412]);
xx = linspace(.5,3.5)';
h = plot(x,y,'x', xx, [ones(size(xx)),xx]*B,'-');
for j=1:nregions;
set(h(nregions+j),'color',get(h(j),'color'));
end
regions = flu.Properties.VarNames;
legend1 = legend(regions{2:end-1});
set(legend1,'Position', [0.7733 0.1967 0.2161 0.6667]);
% Create and fit model with separate intercepts and slopes
18-815
mvregress
for j=1:nobs
X{j} = [eye(nregions), x(j)*eye(nregions)];
end
[beta,sig,resid,vars,loglik2] = mvregress(X,y);
% Plot raw data with fitted lines
B = [beta(1:nregions)';beta(nregions+1:end)']
axes2 = axes('Parent',gcf,'Position',...
[0.13 0.11 0.6191 0.3412]);
h = plot(x,y,'x', xx, [ones(size(xx)),xx]*B,'-');
for j=1:nregions;
set(h(nregions+j),'color',get(h(j),'color'));
end
% Likelihood ratio test for significant difference
chisq = 2*(loglik2-loglik)
p = 1-chi2cdf(chisq, nregions-1)
chisq =
96.4556
p =
18-816
mvregress
References [1] Little, Roderick J. A., and Donald B. Rubin. Statistical Analysis with
Missing Data. 2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 2002.
[2] Meng, Xiao-Li, and Donald B. Rubin. “Maximum Likelihood
Estimation via the ECM Algorithm.” Biometrika. Vol. 80, No. 2, 1993,
pp. 267–278.
[3] Sexton, Joe, and A. R. Swensen. “ECM Algorithms that Converge at
the Rate of EM.” Biometrika. Vol. 87, No. 3, 2000, pp. 651–662.
18-817
mvregress
[4] Dempster, A. P., N. M. Laird, and D. B. Rubin. “Maximum
Likelihood from Incomplete Data via the EM Algorithm.” Journal of the
Royal Statistical Society. Series B, Vol. 39, No. 1, 1977, pp. 1–37.
See Also mvregresslike, manova1
“Multivariate Normal Distribution” on page B-58
18-818
mvregresslike
Purpose Negative log-likelihood for multivariate regression
Syntax nlogL = mvregresslike(X,Y,b,SIGMA,alg)
[nlogL,COVB] = mvregresslike(...)
[nlogL,COVB] = mvregresslike(...,type,format)
Description nlogL = mvregresslike(X,Y,b,SIGMA,alg) computes the negative
log-likelihood nlogL for a multivariate regression of the d-dimensional
multivariate observations in the n-by-d matrix Y on the predictor
variables in the matrix or cell array X, evaluated for the p-by-1 column
vector b of coefficient estimates and the d-by-d matrix SIGMA specifying
the covariance of a row of Y. If d = 1, X can be an n-by-p design matrix
of predictor variables. For any value of d, X can also be a cell array
of length n, with each cell containing a d-by-p design matrix for one
multivariate observation. If all observations have the same d-by-p
design matrix, X can be a single cell.
NaN values in X or Y are taken as missing. Observations with missing
values in X are ignored. Treatment of missing values in Y depends on
the algorithm specified by alg.
alg should match the algorithm used by mvregress to obtain the
coefficient estimates b, and must be one of the following:
• 'ecm' — ECM algorithm
• 'cwls' — Least squares conditionally weighted by SIGMA
• 'mvn' — Multivariate normal estimates computed after omitting
rows with any missing values in Y
[nlogL,COVB] = mvregresslike(...) also returns an estimated
covariance matrix COVB of the parameter estimates b.
[nlogL,COVB] = mvregresslike(...,type,format) specifies the type
and format of COVB.
type is either:
18-819
mvregresslike
• 'hessian' — To use the Hessian or observed information. This
method takes into account the increased uncertainties due to missing
data. This is the default.
• 'fisher' — To use the Fisher or expected information. This method
uses the complete data expected information, and does not include
uncertainty due to missing data.
format is either:
• 'beta' — To compute COVB for b only. This is the default.
• 'full' — To compute COVB for both b and SIGMA.
See Also mvregress, manova1
“Multivariate Normal Distribution” on page B-58
18-820
mvnrnd
Purpose Multivariate normal random numbers
Syntax R = mvnrnd(MU,SIGMA)
r = mvnrnd(MU,SIGMA,cases)
Description R = mvnrnd(MU,SIGMA) returns an n-by-d matrix R of random vectors
chosen from the multivariate normal distribution with mean MU, and
covariance SIGMA. MU is an n-by-d matrix, and mvnrnd generates each
row of R using the corresponding row of mu. SIGMA is a d-by-d symmetric
positive semi-definite matrix, or a d-by-d-by-n array. If SIGMA is an
array, mvnrnd generates each row of R using the corresponding page of
SIGMA, i.e., mvnrnd computes R(i,:) using MU(i,:) and SIGMA(:,:,i).
If the covariance matrix is diagonal, containing variances along the
diagonal and zero covariances off the diagonal, SIGMA may also be
specified as a 1-by-d vector or a 1-by-d-by-n array, containing just the
diagonal. If MU is a 1-by-d vector, mvnrnd replicates it to match the
trailing dimension of SIGMA.
r = mvnrnd(MU,SIGMA,cases) returns a cases-by-d matrix R of
random vectors chosen from the multivariate normal distribution with
a common 1-by-d mean vector MU, and a common d-by-d covariance
matrix SIGMA.
Examples mu = [2 3];
SIGMA = [1 1.5; 1.5 3];
r = mvnrnd(mu,SIGMA,100);
plot(r(:,1),r(:,2),'+')
18-821
mvnrnd
See Also mvnpdf, mvncdf, normrnd
“Multivariate Normal Distribution” on page B-58
18-822
mvtcdf
Purpose Multivariate t cumulative distribution function
Syntax y = mvtcdf(X,C,DF)
y = mvtcdf(xl,xu,C,DF)
[y,err] = mvtcdf(...)
[...] = mvntdf(...,options)
Description y = mvtcdf(X,C,DF) returns the cumulative probability of the
multivariate t distribution with correlation parameters C and degrees
of freedom DF, evaluated at each row of X. Rows of the n-by-d matrix
X correspond to observations or points, and columns correspond to
variables or coordinates. y is an n-by-1 vector.
C is a symmetric, positive definite, d-by-d matrix, typically a correlation
matrix. If its diagonal elements are not 1, mvtcdf scales C to correlation
form. mvtcdf does not rescale X. DF is a scalar, or a vector with n
elements.
The multivariate t cumulative probability at X is defined as the
probability that a random vector T, distributed as multivariate t, will
fall within the semi-infinite rectangle with upper limits defined by X,
i.e., Pr{T(1)≤X(1),T(2)≤X(2),...T(d)≤X(d)}.
y = mvtcdf(xl,xu,C,DF) returns the multivariate t cumulative
probability evaluated over the rectangle with lower and upper limits
defined by xl and xu, respectively.
[y,err] = mvtcdf(...) returns an estimate of the error in y. For
bivariate and trivariate distributions, mvtcdf uses adaptive quadrature
on a transformation of the t density, based on methods developed
by Genz, as described in the references. The default absolute error
tolerance for these cases is 1e-8. For four or more dimensions, mvtcdf
uses a quasi-Monte Carlo integration algorithm based on methods
developed by Genz and Bretz, as described in the references. The
default absolute error tolerance for these cases is 1e-4.
[...] = mvntdf(...,options) specifies control parameters for the
numerical integration used to compute y. This argument can be created
by a call to statset. Choices of statset parameters are:
18-823
mvtcdf
• 'TolFun' — Maximum absolute error tolerance. Default is 1e-8
when d < 4, or 1e-4 when d ≥ 4.
• 'MaxFunEvals' — Maximum number of integrand evaluations
allowed when d ≥ 4. Default is 1e7. 'MaxFunEvals' is ignored when
d < 4.
• 'Display' — Level of display output. Choices are 'off' (the
default), 'iter', and 'final'. 'Display' is ignored when d < 4.
Examples C = [1 .4; .4 1]; df = 2;
[X1,X2] = meshgrid(linspace(-2,2,25)',linspace(-2,2,25)');
X = [X1(:) X2(:)];
p = mvtcdf(X,C,df);
surf(X1,X2,reshape(p,25,25));
18-824
mvtcdf
References [1] Genz, A. “Numerical Computation of Rectangular Bivariate and
Trivariate Normal and t Probabilities.” Statistics and Computing. Vol.
14, No. 3, 2004, pp. 251–260.
[2] Genz, A., and F. Bretz. “Numerical Computation of Multivariate
t Probabilities with Application to Power Calculation of Multiple
Contrasts.” Journal of Statistical Computation and Simulation. Vol.
63, 1999, pp. 361–378.
[3] Genz, A., and F. Bretz. “Comparison of Methods for the Computation
of Multivariate t Probabilities.” Journal of Computational and
Graphical Statistics. Vol. 11, No. 4, 2002, pp. 950–971.
18-825
mvtcdf
See Also mvtpdf, mvtrnd
“Multivariate t Distribution” on page B-64
18-826
mvtpdf
Purpose Multivariate t probability density function
Syntax y = mvtpdf(X,C,df)
Description y = mvtpdf(X,C,df) returns the probability density of the multivariate
t distribution with correlation parameters C and degrees of freedom df,
evaluated at each row of X. Rows of the n-by-d matrix X correspond
to observations or points, and columns correspond to variables or
coordinates. C is a symmetric, positive definite, d-by-d matrix, typically
a correlation matrix. If its diagonal elements are not 1, mvtpdf scales C
to correlation form. mvtcdf does not rescale X. df is a scalar, or a vector
with n elements. y is an n-by-1 vector.
Examples Visualize a multivariate t distribution:
[X1,X2] = meshgrid(linspace(-2,2,25)',linspace(-2,2,25)');
X = [X1(:) X2(:)];
C = [1 .4; .4 1];
df = 2;
p = mvtpdf(X,C,df);
surf(X1,X2,reshape(p,25,25))
18-827
mvtpdf
See Also mvtcdf, mvtrnd
“Multivariate t Distribution” on page B-64
18-828
mvtrnd
Purpose Multivariate t random numbers
Syntax R = mvtrnd(C,df,cases)
R = mvtrnd(C,df)
Description R = mvtrnd(C,df,cases) returns a matrix of random numbers chosen
from the multivariate t distribution, where C is a correlation matrix.
df is the degrees of freedom and is either a scalar or is a vector with
cases elements. If p is the number of columns in C, then the output
R has cases rows and p columns.
Let t represent a row of R. Then the distribution of t is that of a vector
having a multivariate normal distribution with mean 0, variance 1, and
covariance matrix C, divided by an independent chi-square random
value having df degrees of freedom. The rows of R are independent.
C must be a square, symmetric and positive definite matrix. If its
diagonal elements are not all 1 (that is, if C is a covariance matrix
rather than a correlation matrix), mvtrnd rescales C to transform it to a
correlation matrix before generating the random numbers.
R = mvtrnd(C,df) returns a single random number from the
multivariate t distribution.
Examples SIGMA = [1 0.8;0.8 1];
R = mvtrnd(SIGMA,3,100);
plot(R(:,1),R(:,2),'+')
18-829
mvtrnd
See Also mvtpdf, mvtcdf
“Multivariate t Distribution” on page B-64
18-830
cvpartition.N property
Purpose Number of observations (including observations with missing group
values)
Description Number of observations (including observations with missing group
values).
18-831
NaiveBayes class
Purpose Naive Bayes classifier
Description A NaiveBayes object defines a Naive Bayes classifier. A Naive Bayes
classifier assigns a new observation to the most probable class,
assuming the features are conditionally independent given the class
value.
Construction NaiveBayes Create NaiveBayes object
Methods disp Display NaiveBayes classifier
object
display Display NaiveBayes classifier
object
fit Create Naive Bayes classifier
object by fitting training data
posterior Compute posterior probability of
each class for test data
predict Predict class label for test data
subsasgn Subscripted reference for
NaiveBayes object
subsref Subscripted reference for
NaiveBayes object
Properties CIsNonEmpty Flag for non-empty classes
CLevels Class levels
CNames Class names
CPrior Class priors
18-832
NaiveBayes class
Dist Distribution names
NClasses Number of classes
NDims Number of dimensions
Params Parameter estimates
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
Examples Predict the class label using the Naive Bayes classifier
load fisheriris
Use the default Gaussian distribution and a confusion matrix:
O1 = NaiveBayes.fit(meas,species);
C1 = O1.predict(meas);
cMat1 = confusionmat(species,C1)
This returns:
cMat1 =
50 0 0
0 47 3
0 3 47
Use the Gaussian distribution for features 1 and 3 and use the kernel
density estimation for features 2 and 4:
O2 = NaiveBayes.fit(meas,species,'dist',...
{'normal','kernel','normal','kernel'});
C2 = O2.predict(meas);
cMat2 = confusionmat(species,C2)
This returns:
18-833
NaiveBayes class
cMat2 =
50 0 0
0 47 3
0 3 47
References [1] Mitchell, T. (1997) Machine Learning, McGraw Hill.
[2] Vangelis M., Ion A., and Geogios P. Spam Filtering with Naive
Bayes - Which Naive Bayes? (2006) Third Conference on Email and
Anti-Spam.
[3] George H. John and Pat Langley. Estimating continuous
distributions in bayesian classifiers (1995) the Eleventh Conference on
Uncertainty in Artificial Intelligence.
See Also “Naive Bayes Classification” on page 12-6
“Grouped Data” on page 2-34
18-834
NaiveBayes
Purpose Create NaiveBayes object
Description You cannot create a NaiveBayes classifier by calling the constructor.
Use NaiveBayes.fit to create a NaiveBayes classifier by fitting the
object to training data.
See Also fit
18-835
nancov
Purpose Covariance ignoring NaN values
Syntax Y = nancov(X)
Y = nancov(X1,X2)
Y = nancov(...,1)
Y = nancov(...,'pairwise')
Description Y = nancov(X) is the covariance cov of X, computed after removing
observations with NaN values.
For vectors x, nancov(x) is the sample variance of the remaining
elements, once NaN values are removed. For matrices X, nancov(X) is
the sample covariance of the remaining observations, once observations
(rows) containing any NaN values are removed.
Y = nancov(X1,X2), where X1 and X2 are matrices with the same
number of elements, is equivalent to nancov(X), where X = [X1(:)
X2(:)].
nancov removes the mean from each variable (column for matrix X)
before calculating Y. If n is the number of remaining observations after
removing observations with NaN values, nancov normalizes Y by either
n – 1 or n , depending on whether n > 1 or n = 1, respectively. To specify
normalization by n, use Y = nancov(...,1).
Y = nancov(...,'pairwise') computes Y(i,j) using rows with no
NaN values in columns i or j. The result Y may not be a positive definite
matrix.
Examples Generate random data for two variables (columns) with random missing
values:
X = rand(10,2);
p = randperm(numel(X));
X(p(1:5)) = NaN
X =
0.8147 0.1576
NaN NaN
0.1270 0.9572
18-836
nancov
0.9134 NaN
0.6324 NaN
0.0975 0.1419
0.2785 0.4218
0.5469 0.9157
0.9575 0.7922
0.9649 NaN
Establish a correlation between a third variable and the other two
variables:
X(:,3) = sum(X,2)
X =
0.8147 0.1576 0.9723
NaN NaN NaN
0.1270 0.9572 1.0842
0.9134 NaN NaN
0.6324 NaN NaN
0.0975 0.1419 0.2394
0.2785 0.4218 0.7003
0.5469 0.9157 1.4626
0.9575 0.7922 1.7497
0.9649 NaN NaN
Compute the covariance matrix for the three variables after removing
observations (rows) with NaN values:
Y = nancov(X)
Y =
0.1311 0.0096 0.1407
0.0096 0.1388 0.1483
0.1407 0.1483 0.2890
See Also NaN, cov, var, nanvar
18-837
nanmax
Purpose Maximum ignoring NaN values
Syntax y = nanmax(X)
Y = nanmax(X1,X2)
y = nanmax(X,[],dim)
[y,indices] = nanmax(...)
Description y = nanmax(X) is the maximum max of X, computed after removing
NaN values.
For vectors x, nanmax(x) is the maximum of the remaining elements,
once NaN values are removed. For matrices X, nanmax(X) is a row vector
of column maxima, once NaN values are removed. For multidimensional
arrays X, nanmax operates along the first nonsingleton dimension.
Y = nanmax(X1,X2) returns an array Y the same size as X1 and X2 with
Y(i,j) = nanmax(X1(i,j),X2(i,j)). Scalar inputs are expanded to
an array of the same size as the other input.
y = nanmax(X,[],dim) operates along the dimension dim of X.
[y,indices] = nanmax(...) also returns the row indices of the
maximum values for each column in the vector indices.
Examples Find column maxima and their indices for data with missing values:
X = magic(3);
X([1 6:9]) = repmat(NaN,1,5)
X =
NaN 1 NaN
3 5 NaN
4 NaN NaN
[y,indices] = nanmax(X)
y =
4 5 NaN
indices =
3 2 1
See Also NaN, max, nanmin
18-838
nanmean
Purpose Mean ignoring NaN values
Syntax y = nanmean(X)
y = nanmean(X,dim)
Description y = nanmean(X) is the mean of X, computed after removing NaN values.
For vectors x, nanmean(x) is the mean of the remaining elements, once
NaN values are removed. For matrices X, nanmean(X) is a row vector of
column means, once NaN values are removed. For multidimensional
arrays X, nanmean operates along the first nonsingleton dimension.
y = nanmean(X,dim) takes the mean along dimension dim of X.
Note If X contains a vector of all NaN values along some dimension, the
vector is empty once the NaN values are removed, so the sum of the
remaining elements is 0. Since the mean involves division by 0, its
value is NaN. The output NaN is not a mean of NaN values.
Examples Find column means for data with missing values:
X = magic(3);
X([1 6:9]) = repmat(NaN,1,5)
X =
NaN 1 NaN
3 5 NaN
4 NaN NaN
y = nanmean(X)
y =
3.5000 3.0000 NaN
See Also NaN, mean, nanmedian
18-839
nanmedian
Purpose Median ignoring NaN values
Syntax y = nanmedian(X)
y = nanmedian(X,dim)
Description y = nanmedian(X) is the median of X, computed after removing NaN
values.
For vectors x, nanmedian(x) is the median of the remaining elements,
once NaN values are removed. For matrices X, nanmedian(X) is a
row vector of column medians, once NaN values are removed. For
multidimensional arrays X, nanmedian operates along the first
nonsingleton dimension.
y = nanmedian(X,dim) takes the mean along dimension dim of X.
Examples Find column medians for data with missing values:
X = magic(3);
X([1 6:9]) = repmat(NaN,1,5)
X =
NaN 1 NaN
3 5 NaN
4 NaN NaN
y = nanmedian(X)
y =
3.5000 3.0000 NaN
See Also NaN, median, nanmean
18-840
nanmin
Purpose Minimum ignoring NaN values
Syntax y = nanmin(X)
Y = nanmin(X1,X2)
y = nanmin(X,[],dim)
[y,indices] = nanmin(...)
Description y = nanmin(X) is the minimum min of X, computed after removing
NaN values.
For vectors x, nanmin(x) is the minimum of the remaining elements,
once NaN values are removed. For matrices X, nanmin(X) is a row vector
of column minima, once NaN values are removed. For multidimensional
arrays X, nanmin operates along the first nonsingleton dimension.
Y = nanmin(X1,X2) returns an array Y the same size as X1 and X2 with
Y(i,j) = nanmin(X1(i,j),X2(i,j)). Scalar inputs are expanded to
an array of the same size as the other input.
y = nanmin(X,[],dim) operates along the dimension dim of X.
[y,indices] = nanmin(...) also returns the row indices of the
minimum values for each column in the vector indices.
Examples Find column minima and their indices for data with missing values:
X = magic(3);
X([1 6:9]) = repmat(NaN,1,5)
X =
NaN 1 NaN
3 5 NaN
4 NaN NaN
[y,indices] = nanmin(X)
y =
3 1 NaN
indices =
2 1 1
See Also NaN, min, nanmax
18-841
nanstd
Purpose Standard deviation ignoring NaN values
Syntax y = nanstd(X)
y = nanstd(X,1)
y = nanstd(X,flag,dim)
Description y = nanstd(X) is the standard deviation std of X, computed after
removing NaN values.
For vectors x, nanstd(x) is the sample standard deviation of the
remaining elements, once NaN values are removed. For matrices X,
nanstd(X) is a row vector of column sample standard deviations,
once NaN values are removed. For multidimensional arrays X, nanstd
operates along the first nonsingleton dimension.
If n is the number of remaining observations after removing
observations with NaN values, nanstd normalizes y by n–1. To specify
normalization by n, use y = nanstd(X,1).
y = nanstd(X,flag,dim) takes the standard deviation along the
dimension dim of X. The flag is 0 or 1 to specify normalization by n –
1 or n, respectively, where n is the number of remaining observations
after removing observations with NaN values.
Examples Find column standard deviations for data with missing values:
X = magic(3);
X([1 6:9]) = repmat(NaN,1,5)
X =
NaN 1 NaN
3 5 NaN
4 NaN NaN
y = nanstd(X)
y =
0.7071 2.8284 NaN
See Also NaN, std, nanvar, nanmean
18-842
nansum
Purpose Sum ignoring NaN values
Syntax y = nansum(X)
y = nansum(X,dim)
Description y = nansum(X) is the sum of X, computed after removing NaN values.
For vectors x, nansum(x) is the sum of the remaining elements, once
NaN values are removed. For matrices X, nansum(X) is a row vector
of column sums, once NaN values are removed. For multidimensional
arrays X, nansum operates along the first nonsingleton dimension.
y = nansum(X,dim) takes the sum along dimension dim of X.
Note If X contains a vector of all NaN values along some dimension, the
vector is empty once the NaN values are removed, so the sum of the
remaining elements is 0. The output 0 is not a sum of NaN values.
Examples Find column sums for data with missing values:
X = magic(3);
X([1 6:9]) = repmat(NaN,1,5)
X =
NaN 1 NaN
3 5 NaN
4 NaN NaN
y = nansum(X)
y =
7 6 0
See Also NaN, sum
18-843
nanvar
Purpose Variance, ignoring NaN values
Syntax y = nanvar(X)
y = nanvar(X,1)
y = nanvar(X,w)
y = nanvar(X,w,dim)
Description y = nanvar(X) is the variance var of X, computed after removing NaN
values.
For vectors x, nanvar(x) is the sample variance of the remaining
elements, once NaN values are removed. For matrices X, nanvar(X)
is a row vector of column sample variances, once NaN values are
removed. For multidimensional arrays X, nanvar operates along the
first nonsingleton dimension.
nancov removes the mean from each variable (column for matrix X)
before calculating Y. If n is the number of remaining observations after
removing observations with NaN values, nanvar normalizes y by either
n – 1 or n , depending on whether n > 1 or n = 1, respectively. To specify
normalization by n, use y = nanvar(X,1).
y = nanvar(X,w) computes the variance using the weight vector w.
The length of w must equal the length of the dimension over which
nanvar operates, and its elements must be nonnegative. Elements of X
corresponding to NaN values of w are ignored.
y = nanvar(X,w,dim) takes the variance along the dimension dim of X.
Set w to [] to use the default normalization by n – 1.
Examples Find column standard deviations for data with missing values:
X = magic(3);
X([1 6:9]) = repmat(NaN,1,5)
X =
NaN 1 NaN
3 5 NaN
4 NaN NaN
y = nanvar(X)
18-844
nanvar
y =
0.5000 8.0000 NaN
See Also NaN, var, nanstd, nanmean
18-845
nbincdf
Purpose Negative binomial cumulative distribution function
Syntax Y = nbincdf(X,R,P)
Description Y = nbincdf(X,R,P) computes the negative binomial cdf at each of
the values in X using the corresponding number of successes, R and
probability of success in a single trial, P. X, R, and P can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of Y. A scalar input for X, R, or P is expanded to a constant
array with the same dimensions as the other inputs.
The negative binomial cdf is
x
⎛ r + i − 1⎞ r i
y = F ( x | r, p) = ∑ ⎜⎝ i
⎟ p q I(0,1,...) (i)
⎠
i =0
The simplest motivation for the negative binomial is the case of
successive random trials, each having a constant probability P of
success. The number of extra trials you must perform in order to
observe a given number R of successes has a negative binomial
distribution. However, consistent with a more general interpretation
of the negative binomial, nbincdf allows R to be any positive value,
including nonintegers. When R is noninteger, the binomial coefficient in
the definition of the cdf is replaced by the equivalent expression
Γ(r + i)
Γ(r)Γ(i + 1)
Examples x = (0:15);
p = nbincdf(x,3,0.5);
stairs(x,p)
18-846
nbincdf
See Also cdf, nbinpdf, nbininv, nbinstat, nbinfit, nbinrnd
“Negative Binomial Distribution” on page B-72
18-847
nbinfit
Purpose Negative binomial parameter estimates
Syntax parmhat = nbinfit(data)
[parmhat,parmci] = nbinfit(data,alpha)
[...] = nbinfit(data,alpha,options)
Description parmhat = nbinfit(data) returns the maximum likelihood estimates
(MLEs) of the parameters of the negative binomial distribution given
the data in the vector data.
[parmhat,parmci] = nbinfit(data,alpha) returns MLEs and
100(1-alpha) percent confidence intervals. By default, alpha = 0.05,
which corresponds to 95% confidence intervals.
[...] = nbinfit(data,alpha,options) accepts a structure,
options, that specifies control parameters for the iterative algorithm
the function uses to compute maximum likelihood estimates. The
negative binomial fit function accepts an options structure which you
can create using the function statset. Enter statset('nbinfit')
to see the names and default values of the parameters that nbinfit
accepts in the options structure. See the reference page for statset
for more information about these options.
Note The variance of a negative binomial distribution is greater than
its mean. If the sample variance of the data in data is less than its
sample mean, nbinfit cannot compute MLEs. You should use the
poissfit function instead.
See Also nbincdf, nbininv, nbinpdf, nbinrnd, nbinstat, mle, statset
“Negative Binomial Distribution” on page B-72
18-848
nbininv
Purpose Negative binomial inverse cumulative distribution function
Syntax X = nbininv(Y,R,P)
Description X = nbininv(Y,R,P) returns the inverse of the negative binomial
cdf with corresponding number of successes, R and probability of
success in a single trial, P. Since the binomial distribution is discrete,
nbininv returns the least integer X such that the negative binomial cdf
evaluated at X equals or exceeds Y. Y, R, and P can be vectors, matrices,
or multidimensional arrays that all have the same size, which is also
the size of X. A scalar input for Y, R, or P is expanded to a constant array
with the same dimensions as the other inputs.
The simplest motivation for the negative binomial is the case of
successive random trials, each having a constant probability P of
success. The number of extra trials you must perform in order to
observe a given number R of successes has a negative binomial
distribution. However, consistent with a more general interpretation
of the negative binomial, nbininv allows R to be any positive value,
including nonintegers.
Examples How many times would you need to flip a fair coin to have a 99%
probability of having observed 10 heads?
flips = nbininv(0.99,10,0.5) + 10
flips =
33
Note that you have to flip at least 10 times to get 10 heads. That is why
the second term on the right side of the equals sign is a 10.
See Also icdf, nbincdf, nbinpdf, nbinstat, nbinfit, nbinrnd
“Negative Binomial Distribution” on page B-72
18-849
nbinpdf
Purpose Negative binomial probability density function
Syntax Y = nbinpdf(X,R,P)
Description Y = nbinpdf(X,R,P) returns the negative binomial pdf at each of
the values in X using the corresponding number of successes, R and
probability of success in a single trial, P. X, R, and P can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of Y. A scalar input for X, R, or P is expanded to a constant
array with the same dimensions as the other inputs. Note that the
density function is zero unless the values in X are integers.
The negative binomial pdf is
⎛ r + x − 1⎞ r x
y = f ( x | r, p) = ⎜ ⎟ p q I(0,1,...) ( x)
⎝ x ⎠
The simplest motivation for the negative binomial is the case of
successive random trials, each having a constant probability P of
success. The number of extra trials you must perform in order to
observe a given number R of successes has a negative binomial
distribution. However, consistent with a more general interpretation
of the negative binomial, nbinpdf allows R to be any positive value,
including nonintegers. When R is noninteger, the binomial coefficient in
the definition of the pdf is replaced by the equivalent expression
Γ(r + x)
Γ(r)Γ( x + 1)
Examples x = (0:10);
y = nbinpdf(x,3,0.5);
plot(x,y,'+')
set(gca,'Xlim',[-0.5,10.5])
18-850
nbinpdf
See Also pdf, nbincdf, nbininv, nbinstat, nbinfit, nbinrnd
“Negative Binomial Distribution” on page B-72
18-851
nbinrnd
Purpose Negative binomial random numbers
Syntax RND = nbinrnd(R,P)
RND = nbinrnd(R,P,m)
RND = nbinrnd(R,P,m,n)
Description RND = nbinrnd(R,P) is a matrix of random numbers chosen from a
negative binomial distribution with corresponding number of successes,
R and probability of success in a single trial, P. R and P can be vectors,
matrices, or multidimensional arrays that have the same size, which is
also the size of RND. A scalar input for R or P is expanded to a constant
array with the same dimensions as the other input.
RND = nbinrnd(R,P,m) generates random numbers with parameters R
and P, where v is a row vector. If v is a 1-by-2 vector, R is a matrix with
v(1) rows and v(2) columns. If v is 1-by-n, R is an n-dimensional array.
RND = nbinrnd(R,P,m,n) generates random numbers with parameters
R and P, where scalars m and n are the row and column dimensions
of RND.
The simplest motivation for the negative binomial is the case of
successive random trials, each having a constant probability P of
success. The number of extra trials you must perform in order to
observe a given number R of successes has a negative binomial
distribution. However, consistent with a more general interpretation
of the negative binomial, nbinrnd allows R to be any positive value,
including nonintegers.
Examples Suppose you want to simulate a process that has a defect probability of
0.01. How many units might Quality Assurance inspect before finding
three defective items?
r = nbinrnd(3,0.01,1,6)+3
r =
496 142 420 396 851 178
See Also random, nbinpdf, nbincdf, nbininv, nbinstat, nbinfit
18-852
nbinrnd
“Negative Binomial Distribution” on page B-72
18-853
nbinstat
Purpose Negative binomial mean and variance
Syntax [M,V] = nbinstat(R,P)
Description [M,V] = nbinstat(R,P) returns the mean of and variance for the
negative binomial distribution with corresponding number of successes,
R and probability of success in a single trial, P. R and P can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of M and V. A scalar input for R or P is expanded to a
constant array with the same dimensions as the other input.
The mean of the negative binomial distribution with parameters r and p
is rq / p, where q = 1 – p. The variance is rq / p2.
The simplest motivation for the negative binomial is the case of
successive random trials, each having a constant probability P of
success. The number of extra trials you must perform in order to
observe a given number R of successes has a negative binomial
distribution. However, consistent with a more general interpretation
of the negative binomial, nbinstat allows R to be any positive value,
including nonintegers.
Examples p = 0.1:0.2:0.9;
r = 1:5;
[R,P] = meshgrid(r,p);
[M,V] = nbinstat(R,P)
M =
9.0000 18.0000 27.0000 36.0000 45.0000
2.3333 4.6667 7.0000 9.3333 11.6667
1.0000 2.0000 3.0000 4.0000 5.0000
0.4286 0.8571 1.2857 1.7143 2.1429
0.1111 0.2222 0.3333 0.4444 0.5556
V =
90.0000 180.0000 270.0000 360.0000 450.0000
7.7778 15.5556 23.3333 31.1111 38.8889
2.0000 4.0000 6.0000 8.0000 10.0000
18-854
nbinstat
0.6122 1.2245 1.8367 2.4490 3.0612
0.1235 0.2469 0.3704 0.4938 0.6173
See Also nbinpdf, nbincdf, nbininv, nbinfit, nbinrnd
“Negative Binomial Distribution” on page B-72
18-855
ncfcdf
Purpose Noncentral F cumulative distribution function
Syntax P = ncfcdf(X,NU1,NU2,DELTA)
Description P = ncfcdf(X,NU1,NU2,DELTA) computes the noncentral F cdf at
each of the values in X using the corresponding numerator degrees of
freedom in NU1, denominator degrees of freedom in NU2, and positive
noncentrality parameters in DELTA. NU1, NU2, and DELTA can be vectors,
matrices, or multidimensional arrays that have the same size, which is
also the size of P. A scalar input for X, NU1, NU2, or DELTA is expanded to
a constant array with the same dimensions as the other inputs.
The noncentral F cdf is
⎛⎛1 ⎞j ⎞
∞ ⎜⎜ ⎟ − ⎟
F ( x | 1 , 2 , ) = ∑ ⎜⎜ ⎝
2 ⎠ ⎟I ⎛ 1 ⋅ x 1 + j, 2 ⎞
j!
e2 ⎟ ⎜⎜ + ⋅ x 2 ⎟
2 ⎟⎠
j =0 ⎜ ⎟ ⎝ 2 1
⎜ ⎟
⎝ ⎠
where I(x|a,b) is the incomplete beta function with parameters a and b.
Examples Compare the noncentral F cdf with δ = 10 to the F cdf with the same
number of numerator and denominator degrees of freedom (5 and 20
respectively).
x = (0.01:0.1:10.01)';
p1 = ncfcdf(x,5,20,10);
p = fcdf(x,5,20);
plot(x,p,'-',x,p1,'-')
18-856
ncfcdf
References [1] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 189–200.
See Also cdf, ncfpdf, ncfinv, ncfstat, ncfrnd
“Noncentral F Distribution” on page B-78
18-857
ncfinv
Purpose Noncentral F inverse cumulative distribution function
Syntax X = ncfinv(P,NU1,NU2,DELTA)
Description X = ncfinv(P,NU1,NU2,DELTA) returns the inverse of the noncentral
F cdf with numerator degrees of freedom NU1, denominator degrees
of freedom NU2, and positive noncentrality parameter DELTA for the
corresponding probabilities in P. P, NU1, NU2, and DELTA can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of X. A scalar input for P, NU1, NU2, or DELTA is expanded
to a constant array with the same dimensions as the other inputs.
Examples One hypothesis test for comparing two sample variances is to take
their ratio and compare it to an F distribution. If the numerator and
denominator degrees of freedom are 5 and 20 respectively, then you
reject the hypothesis that the first variance is equal to the second
variance if their ratio is less than that computed below.
critical = finv(0.95,5,20)
critical =
2.7109
Suppose the truth is that the first variance is twice as big as the second
variance. How likely is it that you would detect this difference?
prob = 1 - ncfcdf(critical,5,20,2)
prob =
0.1297
If the true ratio of variances is 2, what is the typical (median) value you
would expect for the F statistic?
ncfinv(0.5,5,20,2)
ans =
1.2786
18-858
ncfinv
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
Hoboken, NJ: Wiley-Interscience, 2000.
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 189–200.
See Also icdf, ncfcdf, ncfpdf, ncfstat, ncfrnd
“Noncentral F Distribution” on page B-78
18-859
ncfpdf
Purpose Noncentral F probability density function
Syntax Y = ncfpdf(X,NU1,NU2,DELTA)
Description Y = ncfpdf(X,NU1,NU2,DELTA) computes the noncentral F pdf at
each of the values in X using the corresponding numerator degrees of
freedom in NU1, denominator degrees of freedom in NU2, and positive
noncentrality parameters in DELTA. X, NU1, N2, and B can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of Y. A scalar input for P, NU1, NU2, or DELTA is expanded
to a constant array with the same dimensions as the other inputs.
The F distribution is a special case of the noncentral F where δ = 0. As δ
increases, the distribution flattens like the plot in the example.
Examples Compare the noncentral F pdf with δ = 10 to the F pdf with the same
number of numerator and denominator degrees of freedom (5 and 20
respectively).
x = (0.01:0.1:10.01)';
p1 = ncfpdf(x,5,20,10);
p = fpdf(x,5,20);
plot(x,p,'-',x,p1,'-')
References [1] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 189–200.
See Also pdf, ncfcdf, ncfinv, ncfstat, ncfrnd
18-860
ncfpdf
“Noncentral F Distribution” on page B-78
18-861
ncfrnd
Purpose Noncentral F random numbers
Syntax R = ncfrnd(NU1,NU2,DELTA)
R = ncfrnd(NU1,NU2,DELTA,v)
R = ncfrnd(NU1,NU2,DELTA,m,n)
Description R = ncfrnd(NU1,NU2,DELTA) returns a matrix of random numbers
chosen from the noncentral F distribution with corresponding
numerator degrees of freedom in NU1, denominator degrees of freedom
in NU2, and positive noncentrality parameters in DELTA. NU1, NU2, and
DELTA can be vectors, matrices, or multidimensional arrays that have
the same size, which is also the size of R. A scalar input for NU1, NU2,
or DELTA is expanded to a constant matrix with the same dimensions
as the other inputs.
R = ncfrnd(NU1,NU2,DELTA,v) returns a matrix of random numbers
with parameters NU1, NU2, and DELTA, where v is a row vector. If v is a
1-by-2 vector, R is a matrix with v(1) rows and v(2) columns. If v is
1-by-n, R is an n-dimensional array.
R = ncfrnd(NU1,NU2,DELTA,m,n) generates random numbers with
parameters NU1, NU2, and DELTA, where scalars m and n are the row
and column dimensions of R.
Examples Compute six random numbers from a noncentral F distribution with 10
numerator degrees of freedom, 100 denominator degrees of freedom and
a noncentrality parameter, δ, of 4.0. Compare this to the F distribution
with the same degrees of freedom.
r = ncfrnd(10,100,4,1,6)
r =
2.5995 0.8824 0.8220 1.4485 1.4415 1.4864
r1 = frnd(10,100,1,6)
r1 =
0.9826 0.5911 1.0967 0.9681 2.0096 0.6598
18-862
ncfrnd
References [1] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 189–200.
See Also random, ncfpdf, ncfcdf, ncfinv, ncfstat
“Noncentral F Distribution” on page B-78
18-863
ncfstat
Purpose Noncentral F mean and variance
Syntax [M,V] = ncfstat(NU1,NU2,DELTA)
Description [M,V] = ncfstat(NU1,NU2,DELTA) returns the mean of and variance
for the noncentral F pdf with corresponding numerator degrees of
freedom in NU1, denominator degrees of freedom in NU2, and positive
noncentrality parameters in DELTA. NU1, NU2, and DELTA can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of M and V. A scalar input for NU1, NU2, or DELTA is
expanded to a constant array with the same dimensions as the other
input.
The mean of the noncentral F distribution with parameters ν1, ν2, and
δ is
2 ( + 1 )
1 ( 2 − 2)
where ν2 > 2.
The variance is
2
⎛ ⎞ ⎡ ( + )2 + (2 + )( − 2) ⎤
2⎜ 2 ⎟ ⎢ 1 1 2
⎥
⎝ 1 ⎠ ⎢⎣ ( 2 − 2)2 ( 2 − 4) ⎥⎦
where ν2 > 4.
Examples [m,v]= ncfstat(10,100,4)
m =
1.4286
v =
0.4252
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 73–74.
18-864
ncfstat
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 189–200.
See Also ncfpdf, ncfcdf, ncfinv, ncfrnd
“Noncentral F Distribution” on page B-78
18-865
NaiveBayes.NClasses property
Purpose Number of classes
Description The NClasses property specifies the number of classes in the grouping
variable used to create the Naive Bayes classifier.
18-866
gmdistribution.NComponents property
Purpose Number k of mixture components
Description The number k of mixture components.
18-867
nctcdf
Purpose Noncentral t cumulative distribution function
Syntax P = nctcdf(X,NU,DELTA)
Description P = nctcdf(X,NU,DELTA) computes the noncentral t cdf at each of
the values in X using the corresponding degrees of freedom in NU and
noncentrality parameters in DELTA. X, NU, and DELTA can be vectors,
matrices, or multidimensional arrays that have the same size, which
is also the size of P. A scalar input for X, NU, or DELTA is expanded to a
constant array with the same dimensions as the other inputs.
Examples Compare the noncentral t cdf with DELTA = 1 to the t cdf with the same
number of degrees of freedom (10).
x = (-5:0.1:5)';
p1 = nctcdf(x,10,1);
p = tcdf(x,10);
plot(x,p,'-',x,p1,'-')
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 147–148.
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 201–219.
See Also cdf, nctpdf, nctinv, nctstat, nctrnd
“Noncentral t Distribution” on page B-80
18-868
nctinv
Purpose Noncentral t inverse cumulative distribution function
Syntax X = nctinv(P,NU,DELTA)
Description X = nctinv(P,NU,DELTA) returns the inverse of the noncentral t cdf
with NU degrees of freedom and noncentrality parameter DELTA for
the corresponding probabilities in P. P, NU, and DELTA can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of X. A scalar input for P, NU, or DELTA is expanded to a
constant array with the same dimensions as the other inputs.
Examples x = nctinv([0.1 0.2],10,1)
x =
-0.2914 0.1618
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 147–148.
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 201–219.
See Also icdf, nctcdf, nctpdf, nctstat, nctrnd
“Noncentral t Distribution” on page B-80
18-869
nctpdf
Purpose Noncentral t probability density function
Syntax Y = nctpdf(X,V,DELTA)
Description Y = nctpdf(X,V,DELTA) computes the noncentral t pdf at each of
the values in X using the corresponding degrees of freedom in V and
noncentrality parameters in DELTA. Vector or matrix inputs for X, V, and
DELTA must have the same size, which is also the size of Y. A scalar
input for X, V, or DELTA is expanded to a constant matrix with the same
dimensions as the other inputs.
Examples Compare the noncentral t pdf with DELTA = 1 to the t pdf with the same
number of degrees of freedom (10):
x = (-5:0.1:5)';
nct = nctpdf(x,10,1);
t = tpdf(x,10);
plot(x,nct,'b-','LineWidth',2)
hold on
plot(x,t,'g--','LineWidth',2)
legend('nct','t')
18-870
nctpdf
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 147–148.
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 201–219.
See Also pdf, nctcdf, nctinv, nctstat, nctrnd
“Noncentral t Distribution” on page B-80
18-871
nctrnd
Purpose Noncentral t random numbers
Syntax R = nctrnd(V,DELTA)
R = nctrnd(V,DELTA,v)
R = nctrnd(V,DELTA,m,n)
Description R = nctrnd(V,DELTA) returns a matrix of random numbers chosen
from the noncentral T distribution using the corresponding degrees of
freedom in V and noncentrality parameters in DELTA. V and DELTA can
be vectors, matrices, or multidimensional arrays. A scalar input for V
or DELTA is expanded to a constant array with the same dimensions
as the other input.
R = nctrnd(V,DELTA,v) returns a matrix of random numbers with
parameters V and DELTA, where v is a row vector. If v is a 1-by-2 vector,
R is a matrix with v(1) rows and v(2) columns. If v is 1-by-n, R is an
n-dimensional array.
R = nctrnd(V,DELTA,m,n) generates random numbers with
parameters V and DELTA, where scalars m and n are the row and column
dimensions of R.
Examples nctrnd(10,1,5,1)
ans =
1.6576
1.0617
1.4491
0.2930
3.6297
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 147–148.
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 201–219.
18-872
nctrnd
See Also random, nctpdf, nctcdf, nctinv, nctstat
“Noncentral t Distribution” on page B-80
18-873
nctstat
Purpose Noncentral t mean and variance
Syntax [M,V] = nctstat(NU,DELTA)
Description [M,V] = nctstat(NU,DELTA) returns the mean of and variance for the
noncentral t pdf with NU degrees of freedom and noncentrality parameter
DELTA. NU and DELTA can be vectors, matrices, or multidimensional
arrays that all have the same size, which is also the size of M and V. A
scalar input for NU or DELTA is expanded to a constant array with the
same dimensions as the other input.
The mean of the noncentral t distribution with parameters ν and δ is
( / 2)1 / 2 Γ(( − 1) / 2)
Γ( / 2)
where ν > 1.
The variance is
2
⎡ Γ(( − 1) / 2) ⎤
(1 + 2 ) − 2 ⎢
( − 2 ) 2 ⎣ Γ( / 2) ⎥⎦
where ν > 2.
Examples [m,v] = nctstat(10,1)
m =
1.0837
v =
1.3255
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 147–148.
18-874
nctstat
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 201–219.
See Also nctpdf, nctcdf, nctinv, nctrnd
“Noncentral t Distribution” on page B-80
18-875
ncx2cdf
Purpose Noncentral chi-square cumulative distribution function
Syntax P = ncx2cdf(X,V,DELTA)
Description P = ncx2cdf(X,V,DELTA) computes the noncentral chi-square cdf at
each of the values in X using the corresponding degrees of freedom in V
and positive noncentrality parameters in DELTA. X, V, and DELTA can be
vectors, matrices, or multidimensional arrays that all have the same
size, which is also the size of P. A scalar input for X, V, or DELTA is
expanded to a constant array with the same dimensions as the other
inputs.
Some texts refer to this distribution as the generalized Rayleigh,
Rayleigh-Rice, or Rice distribution.
The noncentral chi-square cdf is
⎛⎛1 ⎞j ⎞
∞ ⎜ ⎜ ⎟ − ⎟
F ( x | , ) = ∑ ⎜⎜ ⎝ ⎟ Pr ⎡ 2
2 ⎠ ⎤
j!
e2 ⎟ ⎣ +2 j ≤ x ⎦
j =0 ⎜ ⎟
⎜ ⎟
⎝ ⎠
Examples Compare the noncentral chi-square cdf with DELTA = 2 to the
chi-square cdf with the same number of degrees of freedom (4):
x = (0:0.1:10)';
ncx2 = ncx2cdf(x,4,2);
chi2 = chi2cdf(x,4);
plot(x,ncx2,'b-','LineWidth',2)
hold on
plot(x,chi2,'g--','LineWidth',2)
legend('ncx2','chi2','Location','NW')
18-876
ncx2cdf
References [1] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 130–148.
See Also cdf, ncx2pdf, ncx2inv, ncx2stat, ncx2rnd
“Noncentral Chi-Square Distribution” on page B-76
18-877
ncx2inv
Purpose Noncentral chi-square inverse cumulative distribution function
Syntax X = ncx2inv(P,V,DELTA)
Description X = ncx2inv(P,V,DELTA) returns the inverse of the noncentral
chi-square cdf using the corresponding degrees of freedom in V and
positive noncentrality parameters in DELTA, at the corresponding
probabilities in P. P, V, and DELTA can be vectors, matrices, or
multidimensional arrays that all have the same size, which is also the
size of X. A scalar input for P, V, or DELTA is expanded to a constant
array with the same dimensions as the other inputs.
Algorithm ncx2inv uses Newton’s method to converge to the solution.
Examples ncx2inv([0.01 0.05 0.1],4,2)
ans =
0.4858 1.1498 1.7066
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 50–52.
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 130–148.
See Also icdf, ncx2cdf, ncx2pdf, ncx2stat, ncx2rnd
“Noncentral Chi-Square Distribution” on page B-76
18-878
ncx2pdf
Purpose Noncentral chi-square probability density function
Syntax Y = ncx2pdf(X,V,DELTA)
Description Y = ncx2pdf(X,V,DELTA) computes the noncentral chi-square pdf at
each of the values in X using the corresponding degrees of freedom in V
and positive noncentrality parameters in DELTA. Vector or matrix inputs
for X, V, and DELTA must have the same size, which is also the size of Y.
A scalar input for X, V, or DELTA is expanded to a constant array with
the same dimensions as the other inputs.
Some texts refer to this distribution as the generalized Rayleigh,
Rayleigh-Rice, or Rice distribution.
Examples Compare the noncentral chi-square pdf with DELTA = 2 to the
chi-square pdf with the same number of degrees of freedom (4):
x = (0:0.1:10)';
ncx2 = ncx2pdf(x,4,2);
chi2 = chi2pdf(x,4);
plot(x,ncx2,'b-','LineWidth',2)
hold on
plot(x,chi2,'g--','LineWidth',2)
legend('ncx2','chi2')
18-879
ncx2pdf
References [1] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 130–148.
See Also pdf, ncx2cdf, ncx2inv, ncx2stat, ncx2rnd
“Noncentral Chi-Square Distribution” on page B-76
18-880
ncx2rnd
Purpose Noncentral chi-square random numbers
Syntax R = ncx2rnd(V,DELTA)
R = ncx2rnd(V,DELTA,v)
R = ncx2rnd(V,DELTA,m,n)
Description R = ncx2rnd(V,DELTA) returns a matrix of random numbers chosen
from the noncentral chi-square distribution using the corresponding
degrees of freedom in V and positive noncentrality parameters in DELTA.
V and DELTA can be vectors, matrices, or multidimensional arrays that
have the same size, which is also the size of R. A scalar input for V or
DELTA is expanded to a constant array with the same dimensions as
the other input.
R = ncx2rnd(V,DELTA,v) returns a matrix of random numbers with
parameters V and DELTA, where v is a row vector. If v is a 1-by-2 vector,
R is a matrix with v(1) rows and v(2) columns. If v is 1-by-n, R is an
n-dimensional array.
R = ncx2rnd(V,DELTA,m,n) generates random numbers with
parameters V and DELTA, where scalars m and n are the row and column
dimensions of R.
Examples ncx2rnd(4,2,6,3)
ans =
6.8552 5.9650 11.2961
5.2631 4.2640 5.9495
9.1939 6.7162 3.8315
10.3100 4.4828 7.1653
2.1142 1.9826 4.6400
3.8852 5.3999 0.9282
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 50–52.
18-881
ncx2rnd
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 130–148.
See Also random, ncx2pdf, ncx2cdf, ncx2inv, ncx2stat
“Noncentral Chi-Square Distribution” on page B-76
18-882
ncx2stat
Purpose Noncentral chi-square mean and variance
Syntax [M,V] = ncx2stat(NU,DELTA)
Description [M,V] = ncx2stat(NU,DELTA) returns the mean of and variance
for the noncentral chi-square pdf with NU degrees of freedom and
noncentrality parameter DELTA. NU and DELTA can be vectors, matrices,
or multidimensional arrays that all have the same size, which is also
the size of M and V. A scalar input for NU or DELTA is expanded to a
constant array with the same dimensions as the other input.
The mean of the noncentral chi-square distribution with parameters ν
and δ is ν+δ, and the variance is 2(ν+2δ).
Examples [m,v] = ncx2stat(4,2)
m =
6
v =
16
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 50–52.
[2] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc.,
1970, pp. 130–148.
See Also ncx2pdf, ncx2cdf, ncx2inv, ncx2rnd
“Noncentral Chi-Square Distribution” on page B-76
18-883
categorical.ndims
Purpose Number of dimensions of categorical array
Syntax n = ndims(A)
Description n = ndims(A) returns the number of dimensions in the categorical
array A. The number of dimensions in an array is always greater than
or equal to 2. Trailing singleton dimensions are ignored. Put simply,
ndims(A) is length(size(A)).
See Also size
18-884
gmdistribution.NDimensions property
Purpose Dimension d of multivariate Gaussian distributions
Description The dimension d of the multivariate Gaussian distributions.
18-885
dataset.ndims
Purpose Number of dimensions of dataset array
Syntax n = ndims(A)
Description n = ndims(A) returns the number of dimensions in the dataset A. The
number of dimensions in an array is always 2.
See Also size
18-886
qrandset.ndims
Purpose Number of dimensions in matrix
Syntax n = ndims(p)
Description n = ndims(p) returns the number of dimensions in the matrix that
is created by the syntax p(:,:). Since this is always a 2-D matrix, n
is always equal to 2.
See Also qrandset, size
18-887
NaiveBayes.NDims property
Purpose Number of dimensions
Description The NDims property specifies the number of dimensions, which is equal
to the number of features in the training data used to create the Naive
Bayes classifier.
18-888
qrandstream.ne
Purpose Not equal relation for handles
Syntax h1 ~= h2
Description Handles are equal if they are handles for the same object and are
unequal otherwise.
h1 ~= h2 performs element-wise comparisons between handle arrays
h1 and h2. h1 and h2 must be of the same dimensions unless one is a
scalar. The result is a logical array of the same dimensions, where each
element is an element-wise ~= result.
If one of h1 or h2 is scalar, scalar expansion is performed and the result
will match the dimensions of the array that is not scalar.
tf = ne(h1, h2) stores the result in a logical array of the same
dimensions.
See Also qrandstream, eq, ge, gt, le, lt
18-889
NeighborSearcher class
Purpose Nearest neighbor search object
Description NeighborSearcher is an abstract class used for nearest neighbor
search. You cannot create instances of this class directly. Instead,
create an instance of a derived class such as ExhaustiveSearcher or
KDTreeSearcher either by calling the derived class constructor or by
calling the function createns.
Construction NeighborSearcher is an abstract class. You cannot create instances of
this class directly. You can construct an object in a subclass, such as
KDTreeSearcher or ExhaustiveSearcher, either by calling the subclass
constructors or by using the createns function.
Properties X
A matrix used to create the object.
Distance
A string specifying a built-in distance metric (applies to both
ExhaustiveSearcher and KDTreeSearcher) or a function handle
(only applies to ExhaustiveSearcher) that you provide when you
create the object. This property is the default distance metric used
when you call the knnsearch method to find nearest neighbors
for future query points.
DistParameter
Specifies the additional parameter for the chosen distance metric.
The value is:
• If 'Distance' is 'minkowski': A positive scalar indicating
the exponent of the Minkowski distance. (Applies for both
ExhaustiveSearcher and KDTreeSearcher.)
• If 'Distance' is 'mahalanobis': A positive definite matrix
representing the covariance matrix used for computing the
Mahalanobis distance. (Only applies for ExhaustiveSearcher.)
18-890
NeighborSearcher class
• If 'Distance' is 'seuclidean': A vector representing the scale
value of the data when computing the 'seuclidean' distance.
(Only applies for ExhaustiveSearcher.)
• Otherwise: Empty.
See Also createns | KDTreeSearcher | ExhaustiveSearcher
18-891
qrandset.net
Purpose Generate quasi-random point set
Syntax X = net(p,n)
Description X = net(p,n) returns the first n points X from the point set p of the
qrandset class. X is n-by-d, where d is the dimension of the point set.
Objects p of the @qrandset class encapsulate properties of a specified
quasi-random sequence. Values of the point set are not generated and
stored in memory until p is accessed using net or parenthesis indexing.
Examples Use haltonset to generate a 3-D Halton point set, skip the first 1000
values, and then retain every 101st point:
p = haltonset(3,'Skip',1e3,'Leap',1e2)
p =
Halton point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : none
Use scramble to apply reverse-radix scrambling:
p = scramble(p,'RR2')
p =
Halton point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : RR2
Use net to generate the first four points:
X0 = net(p,4)
X0 =
0.0928 0.6950 0.0029
0.6958 0.2958 0.8269
18-892
qrandset.net
0.3013 0.6497 0.4141
0.9087 0.7883 0.2166
Use parenthesis indexing to generate every third point, up to the 11th
point:
X = p(1:3:11,:)
X =
0.0928 0.6950 0.0029
0.9087 0.7883 0.2166
0.3843 0.9840 0.9878
0.6831 0.7357 0.7923
See Also haltonset, sobolset, qrandstream
18-893
nlinfit
Purpose Nonlinear regression
Syntax beta = nlinfit(X,y,fun,beta0)
[beta,r,J,COVB,mse] = nlinfit(X,y,fun,beta0)
[...] = nlinfit(X,y,fun,beta0,options)
Description beta = nlinfit(X,y,fun,beta0) returns a vector beta of coefficient
estimates for a nonlinear regression of the responses in y on the
predictors in X using the model specified by fun. X is an n-by-p matrix of
p predictors at each of n observations. y is an n-by-1 vector of observed
responses. fun is a function handle to a function of the form:
yhat = modelfun(b,X)
where b is a coefficient vector. beta0 is a vector containing initial
values for the coefficients. beta is the same length as beta0.
[beta,r,J,COVB,mse] = nlinfit(X,y,fun,beta0) returns the fitted
coefficients beta, the residuals r, the Jacobian J of fun, the estimated
covariance matrix COVB for the fitted coefficients, and an estimate mse
of the variance of the error term. You can use these outputs with
nlpredci to produce error estimates on predictions, and with nlparci
to produce error estimates on the estimated coefficients. If you use the
robust fitting option (see below), you must use COVB and may need mse
as input to nlpredci or nlparci to insure that the confidence intervals
take the robust fit properly into account.
[...] = nlinfit(X,y,fun,beta0,options) specifies control
parameters for the algorithm used in nlinfit. options is a structure
created by a call to statset. Applicable statset parameters are:
• 'MaxIter' — Maximum number of iterations allowed. The default
is 100.
• 'TolFun' — Termination tolerance on the residual sum of squares.
The defaults is 1e-8.
• 'TolX' — Termination tolerance on the estimated coefficients beta.
The default is 1e-8.
18-894
nlinfit
• 'Display' — Level of display output during estimation. The choices
are
- 'off' (the default)
- 'iter'
- 'final'
• 'DerivStep' — Relative difference used in finite difference gradient
calculation. May be a scalar, or the same size as the parameter
vector b0. The default is eps^(1/3).
• 'FunValCheck' — Check for invalid values, such as NaN or Inf, from
the objective function. Values are 'off' or 'on' (the default).
• 'Robust' — Invoke robust fitting option. Values are 'off' (the
default) or 'on'.
• 'WgtFun' — A weight function for robust fitting. Valid only when
'Robust' is 'on'. It can be 'bisquare' (the default), 'andrews',
'cauchy', 'fair', 'huber', 'logistic', 'talwar', or 'welsch'. It
can also be a function handle that accepts a normalized residual as
input and returns the robust weights as output.
• 'Tune' — The tuning constant used in robust fitting to normalize the
residuals before applying the weight function. The value is a positive
scalar, with the default value dependent on the weight function. This
parameter is required if the weight function is specified as a function
handle.
nlinfit treats NaNs in y or modelfun(beta0,X) as missing data and
ignores the corresponding rows.
nlintool is a graphical user interface to nlinfit.
Algorithm nlinfit uses the Levenberg-Marquardt algorithm [1] for nonlinear
least squares to compute non-robust fits.
For robust fits, nlinfit uses an algorithm [2] that iteratively refits a
weighted nonlinear regression, where the weights at each iteration
are based on each observation’s residual from the previous iteration.
18-895
nlinfit
These weights serve to downweight points that are outliers so that
their influence on the fit is decreased. Iterations continue until the
weights converge. The computation of weights is identical to the
iterative reweighting scheme used by robustfit for fitting a robust
linear model, and in particular, the possible weighting functions and
their tuning parameters as the same. The nonlinear regression at each
iteration is a weighted version of the L-M algorithm that nlinfit uses
for non-robust fits.
Examples The data in reaction.mat are partial pressures of three chemical
reactants and the corresponding reaction rates. The function hougen
implements the nonlinear Hougen-Watson model for reaction rates. The
following fits the model to the data:
load reaction
beta = nlinfit(reactants,rate,@hougen,beta)
beta =
1.2526
0.0628
0.0400
0.1124
1.1914
References [1] Seber, G. A. F., and C. J. Wild. Nonlinear Regression. Hoboken, NJ:
Wiley-Interscience, 2003.
[2] DuMouchel, W. H., and F. L. O’Brien. “Integrating a Robust Option
into a Multiple Regression Computing Environment.” Computer Science
and Statistics: Proceedings of the 21st Symposium on the Interface.
Alexandria, VA: American Statistical Association, 1989.
[3] Holland, P. W., and R. E. Welsch. “Robust Regression Using
Iteratively Reweighted Least-Squares.” Communications in Statistics:
Theory and Methods, A6, 1977, pp. 813–827.
See Also nlparci, nlpredci, nlintool
18-896
nlinfit
For information on weighted nonlinear regression, type showdemo
wnlsdemo.
18-897
nlintool
Purpose Interactive nonlinear regression
Syntax nlintool(X,y,fun,beta0)
nlintool(X,y,fun,beta0,alpha)
nlintool(X,y,fun,beta0,alpha,'xname','yname')
Description nlintool(X,y,fun,beta0) is a graphical user interface to the nlinfit
function, and uses the same input arguments. The interface displays
plots of the fitted response against each predictor, with the other
predictors held fixed. The fixed values are in the text boxes below each
predictor axis. Change the fixed values by typing in a new value or
by dragging the vertical lines in the plots to new positions. When you
change the value of a predictor, all plots update to display the model
at the new point in predictor space. Dashed red curves show 95%
simultaneous confidence bands for the function.
nlintool(X,y,fun,beta0,alpha) shows 100(1-alpha)% confidence
bands. These are simultaneous confidence bounds for the function value.
Using the Boundser pmenu you can switch between simultaneous and
non-simultaneous bounds, and between bounds on the function and
bounds for predicting a new observation.
nlintool(X,y,fun,beta0,alpha,'xname','yname') labels the plots
using the string matrix 'xname' for the predictors and the string
'yname' for the response.
Examples The data in reaction.mat are partial pressures of three chemical
reactants and the corresponding reaction rates. The function hougen
implements the nonlinear Hougen-Watson model for reaction rates. The
following fits the model to the data:
load reaction
nlintool(reactants,rate,@hougen,beta,0.01,xn,yn)
18-898
nlintool
See Also nlinfit, polytool, rstool
18-899
nlmefit
Purpose Nonlinear mixed-effects estimation
Syntax beta = nlmefit(X,y,group,V,fun,beta0)
[beta,PSI] = nlmefit(X,y,group,V,fun,beta0)
[beta,PSI,stats] = nlmefit(X,y,group,V,fun,beta0)
[beta,PSI,stats,B] = nlmefit(X,y,group,V,fun,beta0)
[beta,PSI,stats,B] = nlmefit(X,y,group,V,fun,beta0,'Name',
value)
Description beta = nlmefit(X,y,group,V,fun,beta0) fits a nonlinear
mixed-effects regression model and returns estimates of the fixed effects
in beta. By default, nlmefit fits a model in which each parameter
is the sum of a fixed and a random effect, and the random effects are
uncorrelated (their covariance matrix is diagonal).
X is an n-by-h matrix of n observations on h predictors.
y is an n-by-1 vector of responses.
group is a grouping variable indicating m groups in the observations.
group is a categorical variable, a numeric vector, a character matrix
with rows for group names, or a cell array of strings.
V is an m-by-g matrix or cell array of g group-specific predictors. These
are predictors that take the same value for all observations in a group.
The rows of V are assigned to groups using grp2idx, according to the
order specified by grp2idx(group). Use a cell array for V if group
predictors vary in size across groups. Use [] for V if there are no
group-specific predictors.
fun is a handle to a function that accepts predictor values and model
parameters and returns fitted values. fun has the form
yfit = modelfun(PHI,XFUN,VFUN)
The arguments are:
• PHI — A 1-by-p vector of model parameters.
• XFUN — A k-by-h array of predictors, where:
18-900
nlmefit
- k = 1 if XFUN is a single row of X.
- k = ni if XFUN contains the rows of X for a single group of size ni.
- k = n if XFUN contains all rows of X.
• VFUN — Group-specific predictors given by one of:
- A 1-by-g vector corresponding to a single group and a single row
of V.
- An n-by-g array, where the jth row is V(I,:) if the jth observation
is in group I.
If V is empty, nlmefit calls modelfun with only two inputs.
• yfit — A k-by-1 vector of fitted values
When either PHI or VFUN contains a single row, it corresponds to all
rows in the other two input arguments.
Note If modelfun can compute yfit for more than one vector of model
parameters per call, use the 'Vectorization' parameter (described
later) for improved performance.
beta0 is a q-by-1 vector with initial estimates for q fixed effects. By
default, q is the number of model parameters p.
nlmefit fits the model by maximizing an approximation to the marginal
likelihood with random effects integrated out, assuming that:
• Random effects are multivariate normally distributed and
independent between groups.
• Observation errors are independent, identically normally distributed,
and independent of the random effects.
18-901
nlmefit
[beta,PSI] = nlmefit(X,y,group,V,fun,beta0) also returns PSI, an
r-by-r estimated covariance matrix for the random effects. By default, r
is equal to the number of model parameters p.
[beta,PSI,stats] = nlmefit(X,y,group,V,fun,beta0) also returns
stats, a structure with fields:
• logl — The maximized log-likelihood for the fitted model
• mse — The estimated error variance for the fitted model
• aic — The Akaike information criterion for the fitted model
• bic — The Bayesian information criterion for the fitted model
• sebeta — The standard errors for beta
• dfe — The error degrees of freedom for the model
[beta,PSI,stats,B] = nlmefit(X,y,group,V,fun,beta0) also
returns B, an r-by-m matrix of estimated random effects for the m
groups. By default, r is equal to the number of model parameters p.
[beta,PSI,stats,B] =
nlmefit(X,y,group,V,fun,beta0,'Name',value) specifies one
or more optional parameter name/value pairs. Specify Name
inside single quotes.
Use the following parameters to fit a model different from the default.
(The default model is obtained by setting both FEConstDesign and
REConstDesign to eye(p), or by setting both FEParamsSelect and
REParamsSelect to 1:p.) Use at most one parameter with an 'FE'
prefix and one parameter with an 'RE' prefix. The nlmefit function
requires you to specify at least one fixed effect and one random effect.
18-902
nlmefit
Parameter Value
FEParamsSelect A vector specifying which elements of
the parameter vector PHI include a fixed
effect, given as a numeric vector of indices
between 1 and p or as a 1-by-p logical
vector. If q is the specified number of
elements, then the model includes q fixed
effects.
FEConstDesign A p-by-q design matrix ADESIGN, where
ADESIGN*beta are the fixed components of
the p elements of PHI.
FEGroupDesign A p-by-q-by-m array specifying a different
p-by-q fixed-effects design matrix for each
of the m groups.
FEObsDesign A p-by-q-by-n array specifying a different
p-by-q fixed-effects design matrix for each
of the n observations.
REParamsSelect A vector specifying which elements of the
parameter vector PHI include a random
effect, given as a numeric vector of indices
between 1 and p or as a 1-by-p logical
vector. The model includes r random
effects, where r is the specified number of
elements.
REConstDesign A p-by-r design matrix BDESIGN, where
BDESIGN*B are the random components of
the p elements of PHI.
REGroupDesign A p-by-r-by-m array specifying a different
p-by-r random-effects design matrix for
each of m groups.
REObsDesign A p-by-r-by-n array specifying a different
p-by-r random-effects design matrix for
each of n observations.
18-903
nlmefit
Use the following parameters to control the iterative algorithm for
maximizing the likelihood:
Parameter Value
RefineBeta0 Determines whether nlmefit makes an
initial refinement of beta0 by first fitting
modelfun without random effects and
replacing beta0 with beta. Choices are
'on' and 'off'. The default value is 'on'.
ApproximationType The method used to approximate the
likelihood of the model. Choices are:
• 'LME' — Use the likelihood for the
linear mixed-effects model at the
current conditional estimates of beta
and B. This is the default.
• 'RELME' — Use the restricted likelihood
for the linear mixed-effects model at the
current conditional estimates of beta
and B.
• 'FO' — First-order Laplacian
approximation without random effects.
• 'FOCE' — First-order Laplacian
approximation at the conditional
estimates of B.
18-904
nlmefit
Parameter Value
Vectorization Indicates acceptable sizes for the PHI,
XFUN, and VFUN input arguments to
modelfun. Choices are:
• 'SinglePhi' — modelfun can only
accept a single set of model parameters
at a time, so PHI must be a single
row vector in each call. nlmefit calls
modelfun in a loop, if necessary, with
a single PHI vector and with XFUN
containing rows for a single observation
or group at a time. VFUN may be a single
row that applies to all rows of XFUN, or a
matrix with rows corresponding to rows
in XFUN. This is the default.
• 'SingleGroup' — modelfun can only
accept inputs corresponding to a single
group in the data, so XFUN must contain
rows of X from a single group in each
call. Depending on the model, PHI is
a single row that applies to the entire
group or a matrix with one row for each
observation. VFUN is a single row.
• 'Full' — modelfun can accept inputs
for multiple parameter vectors and
multiple groups in the data. Either PHI
or VFUN may be a single row that applies
to all rows of XFUN or a matrix with
rows corresponding to rows in XFUN.
This option can improve performance
by reducing the number of calls to
modelfun, but may require modelfun to
perform singleton expansion on PHI or V.
18-905
nlmefit
Parameter Value
CovParameterization Specifies the parameterization used
internally for the scaled covariance matrix.
Choices are 'chol' for the Cholesky
factorization or 'logm' the matrix
logarithm. The default is 'logm'.
CovPattern Specifies an r-by-r logical or numeric
matrix P that defines the pattern of the
random-effects covariance matrix PSI.
nlmefit estimates the variances along
the diagonal of PSI and the covariances
specified by nonzeroes in the off-diagonal
elements of P. Covariances corresponding
to zero off-diagonal elements in P are
constrained to be zero. If P does not specify
a row-column permutation of a block
diagonal matrix, nlmefit adds nonzero
elements to P as needed. The default
value of P is eye(r), corresponding to
uncorrelated random effects.
Alternatively, P may be a 1-by-r vector
containing values in 1:r, with equal values
specifying groups of random effects. In this
case, nlmefit estimates covariances only
within groups, and constrains covariances
across groups to be zero.
18-906
nlmefit
Parameter Value
ParamTransform A vector of P values specifying a
transformation function f() for each of the
P parameters:
XB = ADESIGN*BETA + BDESIGN*B
PHI = f(XB)
Each element of the vector must be one
of the following integer codes specifying
the transformation for the corresponding
value of PHI:
• 0: PHI = XB (default for all parameters)
• 1: log(PHI) = XB
• 2: probit(PHI) = XB
• 3: logit(PHI) = XB
Options A structure of the form returned by
statset. nlmefit uses the following
statset parameters:
• 'TolX' — Termination tolerance on the
estimated fixed and random effects. The
default is 1e-4.
• 'TolFun' — Termination tolerance on
the log-likelihood function. The default
is 1e-4.
• 'MaxIter' — Maximum number of
iterations allowed. The default is 200.
• 'Display' — Level of iterative display
during estimation. Choices are:
- 'off' — Displays no information.
This is the default.
18-907
nlmefit
Parameter Value
- 'iter' — Displays iterative output
to the command window.
- 'final' — Displays the final output.
• 'FunValCheck' — Check for invalid
values, such as NaN or Inf, from
modelfun. Choices are 'on' and 'off'.
The default is 'on'.
• 'OutputFcn' — Function handle
specified using @, a cell array with
function handles or an empty array
(default). The solver calls all output
functions after each iteration.
OptimFun Specifies the optimization function used
in maximizing the likelihood. Choices
are 'fminsearch' to use fminsearch or
'fminunc' to use fminunc. The default
is 'fminsearch'. You can only specify
'fminunc' if Optimization Toolbox
software is installed.
Examples Display data on the growth of five orange trees:
CIRC = [30 58 87 115 120 142 145;
33 69 111 156 172 203 203;
30 51 75 108 115 139 140;
32 62 112 167 179 209 214;
30 49 81 125 142 174 177];
time = [118 484 664 1004 1231 1372 1582];
h = plot(time,CIRC','o','LineWidth',2);
xlabel('Time (days)')
ylabel('Circumference (mm)')
18-908
nlmefit
title('{\bf Orange Tree Growth}')
legend([repmat('Tree ',5,1),num2str((1:5)')],...
'Location','NW')
grid on
hold on
Use an anonymous function to specify a logistic growth model:
model=@(PHI,t)(PHI(:,1))./(1+exp(-(t-PHI(:,2))./PHI(:,3)));
18-909
nlmefit
Fit the model using nlmefit with default settings (that is, assuming
each parameter is the sum of a fixed and a random effect, with no
correlation among the random effects):
TIME = repmat(time,5,1);
NUMS = repmat((1:5)',size(time));
beta0 = [100 100 100];
[beta1,PSI1,stats1] = nlmefit(TIME(:),CIRC(:),NUMS(:),...
[],model,beta0)
beta1 =
191.3189
723.7609
346.2518
PSI1 =
962.1533 0 0
0 0.0000 0
0 0 297.9931
stats1 =
logl: -131.5457
mse: 59.7881
aic: 277.0913
bic: 287.9788
sebeta: NaN
dfe: 28
The negligible variance of the second random effect, PSI1(2,2),
suggests that it can be removed to simplify the model:
[beta2,PSI2,stats2,b2] = nlmefit(TIME(:),CIRC(:),...
NUMS(:),[],model,beta0,'REParamsSelect',[1 3])
beta2 =
191.3190
723.7610
346.2527
PSI2 =
962.0491 0
18-910
nlmefit
0 298.1869
stats2 =
logl: -131.5457
mse: 59.7881
aic: 275.0913
bic: 284.4234
sebeta: NaN
dfe: 29
b2 =
-28.5254 31.6061 -36.5071 39.0738 -5.6475
10.0034 -0.7633 6.0080 -9.4630 -5.7853
The log-likelihood logl is unaffected, and both the Akaike and Bayesian
information criteria (aic and bic) are reduced, supporting the decision
to drop the second random effect from the model.
Use the estimated fixed effects in beta2 and the estimated random
effects for each tree in b2 to plot the model through the data:
PHI = repmat(beta2,1,5) + ... % Fixed effects
[b2(1,:);zeros(1,5);b2(2,:)]; % Random effects
colors = get(h,'Color');
tplot = 0:0.1:1600;
for I = 1:5
fitted_model=@(t)(PHI(1,I))./(1+exp(-(t-PHI(2,I))./ ...
PHI(3,I)));
plot(tplot,fitted_model(tplot),'Color',colors{I}, ...
'LineWidth',2)
end
18-911
nlmefit
References [1] Lindstrom, M. J., and D. M. Bates. “Nonlinear mixed-effects models
for repeated measures data.” Biometrics. Vol. 46, 1990, pp. 673–687.
[2] Davidian, M., and D. M. Giltinan. Nonlinear Models for Repeated
Measurements Data. New York: Chapman & Hall, 1995.
[3] Pinheiro, J. C., and D. M. Bates. “Approximations to the
log-likelihood function in the nonlinear mixed-effects model.” Journal of
Computational and Graphical Statistics. Vol. 4, 1995, pp. 12–35.
18-912
nlmefit
[4] Demidenko, E. Mixed Models: Theory and Applications. Hoboken,
NJ: John Wiley & Sons, Inc., 2004.
See Also nlinfit | nlpredci | nlmefitsa
How To • “Grouped Data” on page 2-34
18-913
nlmefitsa
Purpose Fit nonlinear mixed effects model with stochastic EM algorithm
Syntax [BETA,PSI,STATS,B] = nlmefitsa(X,Y,GROUP,V,MODELFUN,BETA0)
[BETA,PSI,STATS,B] = nlmefitsa(X,Y,GROUP,V,MODELFUN,BETA0,
'Name',Value)
Description [BETA,PSI,STATS,B] = nlmefitsa(X,Y,GROUP,V,MODELFUN,BETA0)
fits a nonlinear mixed-effects regression model and returns estimates
of the fixed effects in BETA. By default, nlmefitsa fits a model where
each model parameter is the sum of a corresponding fixed and random
effect, and the covariance matrix of the random effects is diagonal, i.e.,
uncorrelated random effects.
The BETA, PSI, and other values this function returns are the result
of a random (Monte Carlo) simulation designed to converge to the
maximum likelihood estimates of the parameters. Because the results
are random, it is advisable to examine the plot of simulation to results
to be sure that the simulation has converged. It may also be helpful to
run the function multiple times, using multiple starting values, or use
the 'Replicates' parameter to perform multiple simulations.
[BETA,PSI,STATS,B] =
nlmefitsa(X,Y,GROUP,V,MODELFUN,BETA0,'Name',Value)
accepts one or more comma-separated parameter name/value
pairs. Specify Name inside single quotes.
Input Definitions:
Arguments In the following list of arguments, the following variable definitions
apply:
• n — number of observations
• h — number of predictor variables
• m — number of groups
• g — number of group-specific predictor variables
• p — number of parameters
18-914
nlmefitsa
• f — number of fixed effects
X
An n-by-h matrix of n observations on h predictor variables.
Y
An n-by-1 vector of responses.
GROUP
A grouping variable indicating which of m groups each observation
belongs to. GROUP can be a categorical variable, a numeric vector,
a character matrix with rows for group names, or a cell array
of strings.
V
An m-by-g matrix of g group-specific predictor variables for each
of the m groups in the data. These are predictor values that take
on the same value for all observations in a group. Rows of V are
ordered according to GRP2IDX(GROUP). Use an m-by-g cell array
for V if any of the group-specific predictor values vary in size
across groups. Specify [] for V if there are no group predictors.
MODELFUN
A handle to a function that accepts predictor values and model
parameters, and returns fitted values. MODELFUN has the form
YFIT = MODELFUN(PHI,XFUN,VFUN) with input arguments
• PHI — A 1-by-p vector of model parameters.
• XFUN — An l-by-h array of predictor variables where
- l is 1 if XFUN is a single row of X
- l is ni if XFUN contains the rows of X for a single group of size ni
- l is n if XFUN contains all rows of X.
• VFUN — Either
18-915
nlmefitsa
- A 1-by-g vector of group-specific predictors for a single group,
corresponding to a single row of V
- An n-by-g matrix, where if the k-th observation is in group i,
then the k-th row of VFUN is V(i,:).
If V is empty, nlmefitsa calls MODELFUN with only two inputs.
MODELFUN returns an l-by-1 vector of fitted values YFIT. When
either PHI or VFUN contains a single row, that one row corresponds
to all rows in the other two input arguments. For improved
performance, use the 'Vectorization' parameter name/value
pair (described below) if MODELFUN can compute YFIT for more
than one vector of model parameters in one call.
BETA0
An f-by-1 vector with initial estimates for the f fixed effects. By
default, f is equal to the number of model parameters p. BETA0
can also be an f-by-REPS matrix, and the estimation is repeated
REPS times using each column of BETA0 as a set of starting values.
Name/Value Pairs
By default, nlmefitsa fits a model where each model parameter is the
sum of a corresponding fixed and random effect. Use the following
parameter name/value pairs to fit a model with a different number of
or dependence on fixed or random effects. Use at most one parameter
name with an 'FE' prefix and one parameter name with an 'RE' prefix.
Note that some choices change the way nlmefitsa calls MODELFUN, as
described further below.
FEParamsSelect
A vector specifying which elements of the model parameter vector
PHI include a fixed effect, as a numeric vector with elements in
1:p, or as a 1-by-p logical vector. The model will include f fixed
effects, where f is the specified number of elements.
FEConstDesign
18-916
nlmefitsa
A p-by-f design matrix ADESIGN, where ADESIGN*BETA are the
fixed components of the p elements of PHI.
FEGroupDesign
A p-by-f-by-m array specifying a different p-by-f fixed effects
design matrix for each of the m groups.
REParamsSelect
A vector specifying which elements of the model parameter vector
PHI include a random effect, as a numeric vector with elements in
1:p, or as a 1-by-p logical vector. The model will include r random
effects, where r is the specified number of elements.
REConstDesign
A p-by-r design matrix BDESIGN, where BDESIGN*B are the random
components of the p elements of PHI. This matrix must consist of
0s and 1s, with at most one 1 per row.
The default model is equivalent to setting both FEConstDesign and
REConstDesign to eye(p), or to setting both FEParamsSelect and
REParamsSelect to 1:p.
Additional optional parameter name/value pairs control the iterative
algorithm used to maximize the likelihood:
CovPattern
Specifies an r-by-r logical or numeric matrix PAT that defines the
pattern of the random effects covariance matrix PSI. nlmefitsa
computes estimates for the variances along the diagonal of
PSI as well as covariances that correspond to non-zeroes in
the off-diagonal of PAT. nlmefitsa constrains the remaining
covariances, i.e., those corresponding to off-diagonal zeroes in
PAT, to be zero. PAT must be a row-column permutation of a block
diagonal matrix, and nlmefitsa adds non-zero elements to PAT
as needed to produce such a pattern. The default value of PAT is
eye(r), corresponding to uncorrelated random effects.
18-917
nlmefitsa
Alternatively, specify PAT as a 1-by-r vector containing values in
1:r. In this case, elements of PAT with equal values define groups
of random effects, nlmefitsa estimates covariances only within
groups, and constrains covariances across groups to be zero.
Cov0
Initial value for the covariance matrix PSI. Must be an r-by-r
positive definite matrix. If empty, the default value depends on
the values of BETA0.
ComputeStdErrors
true to compute standard errors for the coefficient estimates and
store them in the output STATS structure, or false (default) to
omit this computation.
ErrorModel
A string specifying the form of the error term. Default is
'constant'. Each model defines the error using a standard
normal (Gaussian) variable e, the function value f, and one or two
parameters a and b. Choices are
• 'constant' — y = f + a*e
• 'proportional' — y = f + b*f*e
• 'combined' — y = f + (a+b*f)*e
• 'exponential' — y = f*exp(a*e), or equivalently log(y) = log(f)
+ a*e
If this parameter is given, the output STATS.rmse field has the
value
• a for 'constant' and 'exponential'
• b for 'proportional'
• [a b] for 'combined'
18-918
nlmefitsa
ErrorParameters
A scalar or two-element vector specifying starting values for
parameters of the error model. This specifies the a, b, or [a b]
values depending on the ErrorModel parameter.
LogLikMethod
Specifies the method for approximating the log likelihood. Choices
are:
• 'is' — Importance sampling
• 'gq' — Gaussian quadrature
• 'lin' — Linearization
• 'none' — Omit the log likelihood approximation (default)
NBurnIn
Number of initial burn-in iterations during which the parameter
estimates are not recomputed. Default is 5.
NChains
Number c of "chains" simulated. Default is 1. Setting c>1 causes
c simulated coefficient vectors to be computed for each group
during each iteration. Default depends on the data, and is chosen
to provide about 100 groups across all chains.
NIterations
Number of iterations. This can be a scalar or a three-element
vector. Controls how many iterations are performed for each of
three phases of the algorithm:
1 simulated annealing
2 full step size
3 reduced step size
18-919
nlmefitsa
Default is [150 150 100]. A scalar is distributed across the three
phases in the same proportions as the default.
NMCMCIterations
Number of Markov Chain Monte Carlo (MCMC) iterations. This
can be a scalar or a three-element vector. Controls how many
of three different types of MCMC updates are performed during
each phase of the main iteration:
1 full multivariate update
2 single coordinate update
3 multiple coordinate update
Default is [2 2 2]. If you input a scalar value, it is applied to
all three types.
OptimFun
Either 'fminsearch' or 'fminunc', specifying the optimization
function to be used during the estimation process. Default is
'fminsearch'. Use of 'fminunc' requires Optimization Toolbox.
Options
A structure created by a call to statset. nlmefitsa uses the
following statset parameters:
• Display — Level of display during estimation.
- 'off' — (the default) displays no information
- 'final' — displays information after the final iteration of
the estimation algorithm
- 'iter' — displays information at each iteration
• FunValCheck
- 'on' (the default) — to check for invalid values (such as NaN
or Inf) from MODELFUN
18-920
nlmefitsa
- 'off' — to skip this check
• OutputFcn — Function handle specified using @, a cell
array with function handles or an empty array. nlmefitsa
calls all output functions after each iteration. See
nlmefitsaoutputfcn.m (the default output function for
nlmefitsa) for an example of an output function.
ParamTransform
A vector of p values specifying a transformation function f() for
each of the p parameters:
XB = ADESIGN*BETA + BDESIGN*B
PHI = f(XB)
Each element of the vector must be one of the following integer
codes specifying the transformation for the corresponding value
of PHI:
• 0: PHI = XB (default for all parameters)
• 1: log(PHI) = XB
• 2: probit(PHI) = XB
• 3: logit(PHI) = XB
Replicates
Number REPS of estimations to perform starting from the starting
values in the vector BETA0. If BETA0 is a matrix, REPS must
match the number of columns in BETA0. Default is the number of
columns in BETA0.
Vectorization
Determines the possible sizes of the PHI, XFUN, and VFUN input
arguments to MODELFUN. Possible values are:
18-921
nlmefitsa
• 'SinglePhi' — MODELFUN is a function (such as an ODE
solver) that can only compute YFIT for a single set of model
parameters at a time, i.e., PHI must be a single row vector in
each call. nlmefitsa calls MODELFUN in a loop if necessary using
a single PHI vector and with XFUN containing rows for a single
observation or group at a time. VFUN may be a single row that
applies to all rows of XFUN, or a matrix with rows corresponding
to rows in XFUN.
• 'SingleGroup' — MODELFUN can only accept inputs
corresponding to a single group in the data, i.e., XFUN must
contain rows of X from a single group in each call. Depending
on the model, PHI is a single row that applies to the entire
group, or a matrix with one row for each observation. VFUN is
a single row.
• 'Full' — MODELFUN can accept inputs for multiple parameter
vectors and multiple groups in the data. Either PHI or VFUN
may be a single row that applies to all rows of XFUN, or a matrix
with rows corresponding to rows in XFUN. Using this option
can improve performance by reducing the number of calls
to MODELFUN, but may require MODELFUN to perform singleton
expansion on PHI or V.
The default for 'Vectorization' is 'SinglePhi'. In all cases, if
V is empty, nlmefitsa calls MODELFUN with only two inputs.
Output BETA
Arguments Estimates of the fixed effects
PSI
An r-by-r estimated covariance matrix for the random effects. By
default, r is equal to the number of model parameters p.
STATS
A structure with the following fields:
18-922
nlmefitsa
• logl — The maximized log-likelihood for the fitted model;
empty if the LogLikMethod parameter has its default value
of 'none'
• rmse — The square root of the estimated error variance
• aic — The Akaike information criterion (empty if logl is
empty)
• bic — The Bayesian information criterion (empty if logl is
empty)
• sebeta — The standard errors for BETA (empty if the
ComputeStdErrors parameter has its default value of false)
• covb — The estimated covariance of the parameter estimates
(empty if ComputeStdErrors is false)
• dfe — The error degrees of freedom
Examples Fit a model to data on concentrations of the drug indomethacin in the
bloodstream of six subjects over eight hours:
load indomethacin
model = @(phi,t)(phi(:,1).*exp(-phi(:,2).*t)+phi(:,3).*exp(-phi(:,4).*t));
phi0 = [1 1 1 1];
% log transform for 2nd and 4th parameters
xform = [0 1 0 1];
[beta,PSI,stats,br] = nlmefitsa(time,concentration,...
subject,[],model,phi0,'ParamTransform',xform)
18-923
nlmefitsa
% Plot the data along with an overall "population" fit
clf
phi = [beta(1), exp(beta(2)), beta(3), exp(beta(4))];
h = gscatter(time,concentration,subject);
xlabel('Time (hours)')
ylabel('Concentration (mcg/ml)')
title('{\bf Indomethacin Elimination}')
xx = linspace(0,8);
line(xx,model(phi,xx),'linewidth',2,'color','k')
18-924
nlmefitsa
% Plot individual curves based on random effect estimates
for j=1:6
phir = [beta(1)+br(1,j), exp(beta(2)+br(2,j)), ...
beta(3)+br(3,j), exp(beta(4)+br(4,j))];
line(xx,model(phir,xx),'color',get(h(j),'color'))
end
Algorithm In order to estimate the parameters of a nonlinear mixed effects
model, we would like to choose the parameter values that maximize a
18-925
nlmefitsa
likelihood function. These values are called the maximum likelihood
estimates. The likelihood function can be written in the form
( ) ∫ p ( y| , b, 2 ) p ( b|Σ ) db
p y| , 2 , Σ =
where
• y is the response data
• β is the vector of population coefficients
• σ2 is the residual variance
• ∑ is the covariance matrix for the random effects
• b is the set of unobserved random effects
Each p() function on the right-hand-side is a normal (Gaussian)
likelihood function that may depend on covariates.
Since the integral does not have a closed form, it is difficult to
find parameters that maximize it. Delyon, Lavielle, and Moulines
[1] proposed to find the maximum likelihood estimates using an
Expectation-Maximization (EM) algorithm in which the E step is
replaced by a stochastic procedure. They called their algorithm
SAEM, for Stochastic Approximation EM. They demonstrated that this
algorithm has desirable theoretical properties, including convergence
under practical conditions and convergence to a local maximum of the
likelihood function. Their proposal involves three steps:
1 Simulation: Generate simulated values of the random effects b from
the posterior density p(b|Σ) given the current parameter estimates.
2 Stochastic approximation: Update the expected value of the log
likelihood function by taking its value from the previous step, and
moving part way toward the average value of the log likelihood
calculated from the simulated random effects.
18-926
nlmefitsa
3 Maximization step: Choose new parameter estimates to maximize
the log likelihood function given the simulated values of the random
effects.
References [1] Delyon, B., M. Lavielle, and E. Moulines, Convergence of a stochastic
approximation version of the EM algorithm, Annals of Statistics, 27,
94-128, 1999.
[2] Mentré, France, and Marc Lavielle, Stochastic EM algorithms in
population PKPD analyses, 2008.
See Also nlinfit | nlmefit
18-927
gmdistribution.NlogL property
Purpose Negative of log-likelihood
Description The negative of the log-likelihood of the data.
Note This property applies only to gmdistribution objects constructed
with fit.
18-928
ProbDistParametric.NLogL property
Purpose Read-only value specifying negative log likelihood for input data to
ProbDistParametric object
Description NLogL is a read-only property of the ProbDistParametric class. NLogL
is a value specifying the negative log likelihood for input data used to fit
a distribution represented by a ProbDistParametric object.
Values The value is a numeric scalar for a distribution fit to input data, that
is, a distribution created using the fitdist function. This property is
empty for distributions created without fitting to data, that is, by using
the ProbDistUnivParam.ProbDistUnivParam constructor. Use this
information to view and compare the negative log likelihood for input
data supplied to create distributions.
18-929
ProbDistUnivKernel.NLogL property
Purpose Read-only value specifying negative log likelihood for input data to
ProbDistUnivKernel object
Description NLogL is a read-only property of the ProbDistUnivKernel class. NLogL
is a value specifying the negative log likelihood for input data used to fit
a distribution represented by a ProbDistUnivKernel object.
Values The value is a numeric scalar for a distribution fit to input data, that is,
a distribution created using the fitdist function. Use this information
to view and compare the negative log likelihood for input data used to
create distributions.
18-930
nlparci
Purpose Nonlinear regression parameter confidence intervals
Syntax ci = nlparci(beta,resid,'covar',sigma)
ci = nlparci(beta,resid,'jacobian',J)
ci = nlparci(...,'alpha',alpha)
Description ci = nlparci(beta,resid,'covar',sigma) returns the 95%
confidence intervals ci for the nonlinear least squares parameter
estimates beta. Before calling nlparci, use nlinfit to fit a nonlinear
regression model and get the coefficient estimates beta, residuals
resid, and estimated coefficient covariance matrix sigma.
ci = nlparci(beta,resid,'jacobian',J) is an alternative syntax
that also computes 95% confidence intervals. J is the Jacobian
computed by nlinfit. If the 'robust' option is used with nlinfit,
use the 'covar' input rather than the 'jacobian' input so that the
required sigma parameter takes the robust fitting into account.
ci = nlparci(...,'alpha',alpha) returns 100(1-alpha)%
confidence intervals.
nlparci treats NaNs in resid or J as missing values, and ignores the
corresponding observations.
The confidence interval calculation is valid for systems where the length
of resid exceeds the length of beta and J has full column rank. When J
is ill-conditioned, confidence intervals may be inaccurate.
Examples Continuing the example from nlinfit:
load reaction
[beta,resid,J,Sigma] = ...
nlinfit(reactants,rate,@hougen,beta);
ci = nlparci(beta,resid,'jacobian',J)
ci =
-0.7467 3.2519
-0.0377 0.1632
18-931
nlparci
-0.0312 0.1113
-0.0609 0.2857
-0.7381 3.1208
See Also nlinfit, nlpredci
18-932
nlpredci
Purpose Nonlinear regression prediction confidence intervals
Syntax [ypred,delta] = nlpredci(modelfun,x,beta,resid,'covar',sigma)
[ypred,delta] = nlpredci(modelfun,x,beta,resid,'jacobian',J)
[...] = nlpredci(...,param1,val1,param2,val2,...)
Description [ypred,delta] =
nlpredci(modelfun,x,beta,resid,'covar',sigma) returns
predictions, ypred, and 95% confidence interval half-widths, delta, for
the nonlinear regression model defined by modelfun, at input values x.
modelfun is a function handle, specified using @, that accepts two
arguments—a coefficient vector and the array x—and returns a vector
of fitted y values. Before calling nlpredci, use nlinfit to fit modelfun
by nonlinear least squares and get estimated coefficient values beta,
residuals resid, and estimated coefficient covariance matrix sigma.
[ypred,delta] =
nlpredci(modelfun,x,beta,resid,'jacobian',J) is an alternative
syntax that also computes 95% confidence intervals. J is the Jacobian
computed by nlinfit. If the 'robust' option is used with nlinfit, use
the 'covar' input rather than the 'jacobian' input so that the
required sigma parameter takes the robust fitting into account.
[...] = nlpredci(...,param1,val1,param2,val2,...) accepts
optional parameter name/value pairs.
Parameter Value
'alpha' A value between 0 and 1 that specifies the confidence
level as 100(1-alpha)%. Default is 0.05.
'mse' The mean squared error returned by nlinfit. This is
required to predict new observations (see 'predopt') if
the robust option is used with nlinfit; otherwise, the
'mse' is computed from the residuals and does not take
the robust fitting into account.
18-933
nlpredci
Parameter Value
'predopt' Either 'curve' (the default) to compute confidence
intervals for the estimated curve (function value) at
x, or 'observation' for prediction intervals for a
new observation at x. If 'observation'is specified
after using a robust option with nlinfit, the 'mse'
parameter must be supplied to specify the robust
estimate of the mean squared error.
'simopt' Either 'on' for simultaneous bounds, or 'off' (the
default) for nonsimultaneous bounds.
nlpredci treats NaNs in resid or J as missing values, and ignores the
corresponding observations.
The confidence interval calculation is valid for systems where the
length of resid exceeds the length of beta and J has full column rank
at beta. When J is ill-conditioned, predictions and confidence intervals
may be inaccurate.
Examples Continuing the example from nlinfit, you can determine the predicted
function value at the value newX and the half-width of a confidence
interval for it.
load reaction;
[beta,resid,J,Sigma] = nlinfit(reactants,rate,@hougen,...
beta);
newX = reactants(1:2,:);
[ypred, delta] = nlpredci(@hougen,newX,beta,resid,...
'Covar',Sigma);
ypred =
8.4179
3.9542
delta =
0.2805
18-934
nlpredci
0.2474
See Also nlinfit, nlparci
18-935
nnmf
Purpose Nonnegative matrix factorization
Syntax [W,H] = nnmf(A,k)
[W,H] = nnmf(A,k,param1,val1,param2,val2,...)
[W,H,D] = nnmf(...)
Description [W,H] = nnmf(A,k) factors the nonnegative n-by-m matrix A into
nonnegative factors W (n-by-k) and H (k-by-m). The factorization is not
exact; W*H is a lower-rank approximation to A. The factors W and H are
chosen to minimize the root-mean-squared residual D between A and
W*H:
D = sqrt(norm(A-W*H,'fro')/(N*M))
The factorization uses an iterative method starting with random initial
values for W and H. Because the root-mean-squared residual D may
have local minima, repeated factorizations may yield different W and H.
Sometimes the algorithm converges to a solution of lower rank than k,
which may indicate that the result is not optimal.
W and H are normalized so that the rows of H have unit length. The
columns of W are ordered by decreasing length.
[W,H] = nnmf(A,k,param1,val1,param2,val2,...) specifies optional
parameter name/value pairs from the following table.
Parameter Value
'algorithm' Either 'als' (the default) to use an alternating
least-squares algorithm, or 'mult' to use a
multiplicative update algorithm.
In general, the 'als' algorithm converges faster
and more consistently. The 'mult' algorithm is
more sensitive to initial values, which makes it a
good choice when using 'replicates' to find W
and H from multiple random starting values.
18-936
nnmf
Parameter Value
'w0' An n-by-k matrix to be used as the initial value
for W.
'h0' A k-by-m matrix to be used as the initial value
for H.
'options' An options structure as created by the statset
function. nnmf uses the following fields of the
options structure: Display, TolX, TolFun,
and MaxIter. Unlike in optimization settings,
reaching MaxIter iterations is treated as
convergence.
'replicates' The number of times to repeat the factorization,
using new random starting values for W and H,
except at the first replication if 'w0' and 'h0'
are given. This is most beneficial with the 'mult'
algorithm. The default is 1.
[W,H,D] = nnmf(...) also returns D, the root mean square residual.
Examples Example 1
Compute a nonnegative rank-two approximation of the measurements
of the four variables in Fisher’s iris data:
load fisheriris
[W,H] = nnmf(meas,2);
H
H =
0.6852 0.2719 0.6357 0.2288
0.8011 0.5740 0.1694 0.0087
The first and third variables in meas (sepal length and petal length,
with coefficients 0.6852 and 0.6357, respectively) provide relatively
strong weights to the first column of W. The first and second variables
in meas (sepal length and sepal width, with coefficients 0.8011and
0.5740) provide relatively strong weights to the second column of W.
18-937
nnmf
Create a biplot of the data and the variables in meas in the column
space of W:
biplot(H','scores',W,'varlabels',{'sl','sw','pl','pw'});
axis([0 1.1 0 1.1])
xlabel('Column 1')
ylabel('Column 2')
Example 2
Starting from a random array X with rank 20, try a few iterations at
several replicates using the multiplicative algorithm:
X = rand(100,20)*rand(20,50);
opt = statset('MaxIter',5,'Display','final');
18-938
nnmf
[W0,H0] = nnmf(X,5,'replicates',10,...
'options',opt,...
'algorithm','mult');
rep iteration rms resid |delta x|
1 5 0.560887 0.0245182
2 5 0.66418 0.0364471
3 5 0.609125 0.0358355
4 5 0.608894 0.0415491
5 5 0.619291 0.0455135
6 5 0.621549 0.0299965
7 5 0.640549 0.0438758
8 5 0.673015 0.0366856
9 5 0.606835 0.0318931
10 5 0.633526 0.0319591
Final root mean square residual = 0.560887
Continue with more iterations from the best of these results using
alternating least squares:
opt = statset('Maxiter',1000,'Display','final');
[W,H] = nnmf(X,5,'w0',W0,'h0',H0,...
'options',opt,...
'algorithm','als');
rep iteration rms resid |delta x|
1 80 0.256914 9.78625e-005
Final root mean square residual = 0.256914
References [1] Berry, M. W., et al. “Algorithms and Applications for Approximate
Nonnegative Matrix Factorization.” Computational Statistics and Data
Analysis. Vol. 52, No. 1, 2007, pp. 155–173.
See Also princomp, factoran, statset
18-939
classregtree.nodeerr
Purpose Return vector of node errors
Syntax e = nodeerr(t)
e = nodeerr(t,nodes)
Description e = nodeerr(t) returns an n-element vector e of the errors of the nodes
in the tree t, where n is the number of nodes. For a regression tree, the
error e(i) for node i is the variance of the observations assigned to
node i. For a classification tree, e(i) is the misclassification probability
for node i.
e = nodeerr(t,nodes) takes a vector nodes of node numbers and
returns the errors for the specified nodes.
The error e is the so-called resubstitution error computed by applying
the tree to the same data used to create the tree. This error is likely to
under estimate the error you would find if you applied the tree to new
data. The test function provides options to compute the error (or cost)
using cross-validation or a test sample.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
18-940
classregtree.nodeerr
view(t)
e = nodeerr(t)
e =
0.6667
0
0.5000
0.0926
0.0217
0.0208
18-941
classregtree.nodeerr
0.3333
0
0
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, numnodes, test
18-942
classregtree.nodeprob
Purpose Node probabilities
Syntax p = nodeprob(t)
p = nodeprob(t,nodes)
Description p = nodeprob(t) returns an n-element vector p of the probabilities of
the nodes in the tree t, where n is the number of nodes. The probability
of a node is computed as the proportion of observations from the original
data that satisfy the conditions for the node. For a classification tree,
this proportion is adjusted for any prior probabilities assigned to each
class.
p = nodeprob(t,nodes) takes a vector nodes of node numbers and
returns the probabilities for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-943
classregtree.nodeprob
p = nodeprob(t)
p =
1.0000
0.3333
0.6667
0.3600
0.3067
0.3200
0.0400
0.3133
18-944
classregtree.nodeprob
0.0067
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, nodesize, numnodes
18-945
classregtree.nodesize
Purpose Return node size
Syntax sizes = nodesize(t)
sizes = nodesize(t,nodes)
Description sizes = nodesize(t) returns an n-element vector sizes of the sizes
of the nodes in the tree t, where n is the number of nodes. The size of
a node is defined as the number of observations from the data used to
create the tree that satisfy the conditions for the node.
sizes = nodesize(t,nodes) takes a vector nodes of node numbers
and returns the sizes for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-946
classregtree.nodesize
sizes = nodesize(t)
sizes =
150
50
100
54
46
48
6
47
18-947
classregtree.nodesize
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, numnodes
18-948
qrandstream.notify
Purpose Notify listeners of event
Syntax notify(h,'eventname')
notify(h,'eventname',data)
Description notify(h,'eventname') notifies listeners added to the event named
eventname on handle object array h that the event is taking place. h
is the array of handles to objects triggering the event, and eventname
must be a string.
notify(h,'eventname',data) provides a way of encapsulating
information about an event which can then be accessed by each
registered listener. data must belong to the event.eventdata class.
See Also addlistener, event.EventData, events, qrandstream
18-949
nominal class
Superclasses categorical
Purpose Arrays for nominal categorical data
Description Nominal arrays are used to store discrete values that are not
numeric and that do not have an ordering. A nominal array provides
efficient storage and convenient manipulation of such data, while also
maintaining meaningful labels for the values. Nominal arrays are often
used as grouping variables.
You can subscript, concatenate, reshape, etc. nominal arrays much
like ordinary numeric arrays. You can test equality between elements
of two nominal arrays, or between a nominal array and a single string
representing a nominal value.
Construction Use the nominal constructor to create a nominal array from a numeric,
logical, or character array, or from a cell array of strings.
nominal Construct nominal categorical
array
Methods Each nominal array carries along a list of possible values that it
can store, known as its levels. The list is created when you create a
nominal array, and you can access it using the getlevels method, or
modify it using the addlevels, mergelevels, or droplevels methods.
Assignment to the array will also add new levels automatically if the
values assigned are not already levels of the array.
You can change the order of the list of levels for a nominal array using
the reorderlevels method, however, that order has no significance for
the values in the array. The order is used only for display purposes, or
when you convert the nominal array to numeric values using methods
such as double or subsindex, or compare two arrays using isequal. If
you need to work with values that have a mathematical ordering, you
should use an ordinal array instead.
18-950
nominal class
Inherited Methods
Methods in the following table are inherited from categorical.
addlevels Add levels to categorical array
cat Concatenate categorical arrays
cellstr Convert categorical array to cell
array of strings
char Convert categorical array to
character array
circshift Shift categorical array circularly
ctranspose Transpose categorical matrix
disp Display categorical array
display Display categorical array
double Convert categorical array to
double array
droplevels Drop levels
end Last index in indexing expression
for categorical array
flipdim Flip categorical array along
specified dimension
fliplr Flip categorical matrix in
left/right direction
flipud Flip categorical matrix in
up/down direction
getlabels Access categorical array labels
getlevels Get categorical array levels
hist Plot histogram of categorical data
18-951
nominal class
horzcat Horizontal concatenation for
categorical arrays
int16 Convert categorical array to
signed 16-bit integer array
int32 Convert categorical array to
signed 32-bit integer array
int64 Convert categorical array to
signed 64-bit integer array
int8 Convert categorical array to
signed 8-bit integer array
intersect Set intersection for categorical
arrays
ipermute Inverse permute dimensions of
categorical array
isempty True for empty categorical array
isequal True if categorical arrays are
equal
islevel Test for levels
ismember True for elements of categorical
array in set
isscalar True if categorical array is scalar
isundefined Test for undefined elements
isvector True if categorical array is vector
length Length of categorical array
levelcounts Element counts by level
ndims Number of dimensions of
categorical array
numel Number of elements in categorical
array
18-952
nominal class
permute Permute dimensions of
categorical array
reorderlevels Reorder levels
repmat Replicate and tile categorical
array
reshape Resize categorical array
rot90 Rotate categorical matrix 90
degrees
setdiff Set difference for categorical
arrays
setlabels Label levels
setxor Set exclusive-or for categorical
arrays
shiftdim Shift dimensions of categorical
array
single Convert categorical array to
single array
size Size of categorical array
squeeze Squeeze singleton dimensions
from categorical array
subsasgn Subscripted assignment for
categorical array
subsindex Subscript index for categorical
array
subsref Subscripted reference for
categorical array
summary Summary statistics for categorical
array
times Product of categorical arrays
18-953
nominal class
transpose Transpose categorical matrix
uint16 Convert categorical array to
unsigned 16-bit integers
uint32 Convert categorical array to
unsigned 32-bit integers
uint64 Convert categorical array to
unsigned 64-bit integers
uint8 Convert categorical array to
unsigned 8-bit integers
union Set union for categorical arrays
unique Unique values in categorical
array
vertcat Vertical concatenation for
categorical arrays
Properties Inherited Properties
Properties in the following table are inherited from categorical.
labels Text labels for levels
undeflabel Text label for undefined levels
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
Examples Create a nominal array from string data in a cell array:
colors = nominal({'r' 'b' 'g';'g' 'r' 'b';'b' 'r' 'g'},...
{'blue' 'green' 'red'})
% Find elements meeting a criterion
colors == 'red'
18-954
nominal class
ismember(colors,{'red' 'blue'})
% Compare two nominal arrays
colors2 = fliplr(colors)
colors == colors2
See Also histc | ordinal
18-955
nominal
Purpose Construct nominal categorical array
Syntax B = nominal(A)
B = nominal(A,labels)
B = nominal(A,labels,levels)
B = nominal(A,labels,[],edges)
Description B = nominal(A) creates a nominal array B from the array A. A can
be numeric, logical, character, categorical, or a cell array of strings.
nominal creates the levels of B from the sorted unique values in A, and
creates default labels for them.
B = nominal(A,labels) labels the levels in B using the character array
or cell array of strings labels. nominal assigns labels to levels in B in
order according to the sorted unique values in A.
B = nominal(A,labels,levels) creates a nominal array with possible
levels defined by levels. levels is a vector whose values can be
compared to those in A using the equality operator. nominal assigns
labels to each level from the corresponding elements of labels.
If A contains any values not present in levels, the levels of the
corresponding elements of B are undefined.
B = nominal(A,labels,[],edges) creates a nominal array by binning
the numeric array A with bin edges given by the numeric vector
edges. The uppermost bin includes values equal to the right-most
edge. nominal assigns labels to each level in B from the corresponding
elements of labels. edges must have one more element than labels.
By default, an element of B is undefined if the corresponding element
of A is NaN (when A is numeric), an empty string (when A is character),
or undefined (when A is categorical). nominal treats such elements as
“undefined” or “missing” and does not include entries for them among
the possible levels for B. To create an explicit level for such elements
instead of treating them as undefined, you must use the levels input,
and include NaN, the empty string, or an undefined element.
You may include duplicate labels in labels in order to merge multiple
values in A into a single level in B.
18-956
nominal
Examples Create a nominal array from Fisher’s iris data:
load fisheriris
species = nominal(species);
summary(species)
setosa versicolor virginica
50 50 50
Create a nominal array from characters, and provide explicit labels:
colors1 = nominal({'r' 'b' 'g'; 'g' 'r' 'b'; 'b' 'r' 'g'},...
{'blue' 'green' 'red'})
Create a nominal array from characters, and provide both explicit labels
and an explicit order for display:
colors2 = nominal({'r' 'b' 'g'; 'g' 'r' 'b'; 'b' 'r' 'g'}, ...
{'red' 'green' 'blue'},{'r' 'g' 'b'})
Create a nominal array from integer data, merging odd and even values
into only two nominal levels. Provide explicit labels:
toss = nominal(randi([1 4],5,2),{'odd' 'even' 'odd' 'even'},1:4)
1 Load patient data from the CSV file hospital.dat and store the
information in a dataset array with observation names given by the
first column in the data (patient identification):
patients = dataset('file','hospital.dat',...
'delimiter',',',...
'ReadObsNames',true);
18-957
nominal
2 Make the {0,1}-valued variable smoke nominal, and change the labels
to 'No' and 'Yes':
patients.smoke = nominal(patients.smoke,{'No','Yes'});
3 Add new levels to smoke as placeholders for more detailed histories
of smokers:
patients.smoke = addlevels(patients.smoke,...
{'0-5 Years','5-10 Years','LongTerm'});
4 Assuming the nonsmokers have never smoked, relabel the 'No' level:
patients.smoke = setlabels(patients.smoke,'Never','No');
5 Drop the undifferentiated 'Yes' level from smoke:
patients.smoke = droplevels(patients.smoke,'Yes');
Warning: OLDLEVELS contains categorical levels that
were present in A, caused some array elements to have
undefined levels.
Note that smokers now have an undefined level.
6 Set each smoker to one of the new levels, by observation name:
patients.smoke('YPL-320') = '5-10 Years';
See Also histc, ordinal
18-958
normcdf
Purpose Normal cumulative distribution function
Syntax P = normcdf(X,mu,sigma)
[P,PLO,PUP] = normcdf(X,mu,sigma,pcov,alpha)
Description P = normcdf(X,mu,sigma) computes the normal cdf at each of the
values in X using the corresponding mean mu and standard deviation
sigma. X, mu, and sigma can be vectors, matrices, or multidimensional
arrays that all have the same size. A scalar input is expanded to a
constant array with the same dimensions as the other inputs. The
parameters in sigma must be positive.
[P,PLO,PUP] = normcdf(X,mu,sigma,pcov,alpha) produces
confidence bounds for P when the input parameters mu and sigma are
estimates. pcov is the covariance matrix of the estimated parameters.
alpha specifies 100(1 - alpha)% confidence bounds. The default value of
alpha is 0.05. PLO and PUP are arrays of the same size as P containing
the lower and upper confidence bounds.
The function normdf computes confidence bounds for P using a normal
approximation to the distribution of the estimate
X − ˆ
ˆ
and then transforming those bounds to the scale of the output P. The
computed bounds give approximately the desired confidence level when
you estimate mu, sigma, and pcov from large samples, but in smaller
samples other methods of computing the confidence bounds might be
more accurate.
The normal cdf is
−(t − ) 2
1 x
p = F ( x| , ) = ∫−∞ e 2 dt
2
2
The result, p, is the probability that a single observation from a normal
distribution with parameters µ and σ will fall in the interval (-∞ x].
18-959
normcdf
The standard normal distribution has µ = 0 and σ = 1.
Examples What is the probability that an observation from a standard normal
distribution will fall on the interval [-1 1]?
p = normcdf([-1 1]);
p(2)-p(1)
ans =
0.6827
More generally, about 68% of the observations from a normal
distribution fall within one standard deviation, σ, of the mean, µ.
See Also cdf, normpdf, norminv, normstat, normfit, normlike, normrnd
“Normal Distribution” on page B-83
18-960
normfit
Purpose Normal parameter estimates
Syntax [muhat,sigmahat] = normfit(data)
[muhat,sigmahat,muci,sigmaci] = normfit(data)
[muhat,sigmahat,muci,sigmaci] = normfit(data,alpha)
[...] = normfit(data,alpha,censoring)
[...] = normfit(data,alpha,censoring,freq)
[...] = normfit(data,alpha,censoring,freq,options)
Description [muhat,sigmahat] = normfit(data) returns estimates of the mean,
µ, and standard deviation, σ, of the normal distribution given the data
in data.
[muhat,sigmahat,muci,sigmaci] = normfit(data) returns 95%
confidence intervals for the parameter estimates on the mean and
standard deviation in the arrays muci and sigmaci, respectively. The
first row of muci contains the lower bounds of the confidence intervals
for µ the second row contains the upper bounds. The first row of
sigmaci contains the lower bounds of the confidence intervals for σ, and
the second row contains the upper bounds.
[muhat,sigmahat,muci,sigmaci] = normfit(data,alpha) returns
100(1 - alpha) % confidence intervals for the parameter estimates,
where alpha is a value in the range [0 1] specifying the width of the
confidence intervals. By default, alpha is 0.05, which corresponds to
95% confidence intervals.
[...] = normfit(data,alpha,censoring) accepts a Boolean vector,
censoring, of the same size as data, which is 1 for observations that
are right-censored and 0 for observations that are observed exactly.
data must be a vector in order to pass in the argument censoring.
[...] = normfit(data,alpha,censoring,freq) accepts a frequency
vector, freq, of the same size as data. Typically, freq contains integer
frequencies for the corresponding elements in data, but can contain
any nonnegative values. Pass in [] for alpha, censoring, or freq to
use their default values.
18-961
normfit
[...] = normfit(data,alpha,censoring,freq,options) accepts a
structure, options, that specifies control parameters for the iterative
algorithm the function uses to compute maximum likelihood estimates
when there is censoring. The normal fit function accepts an options
structure which you can create using the function statset. Enter
statset('normfit') to see the names and default values of the
parameters that normfit accepts in the options structure. See the
reference page for statset for more information about these options.
Examples In this example the data is a two-column random normal matrix. Both
columns have µ = 10 and σ = 2. Note that the confidence intervals below
contain the "true values."
data = normrnd(10,2,100,2);
[mu,sigma,muci,sigmaci] = normfit(data)
mu =
10.1455 10.0527
sigma =
1.9072 2.1256
muci =
9.7652 9.6288
10.5258 10.4766
sigmaci =
1.6745 1.8663
2.2155 2.4693
See Also mle, normlike, normpdf, normcdf, norminv, normstat, normrnd
“Normal Distribution” on page B-83
18-962
norminv
Purpose Normal inverse cumulative distribution function
Syntax X = norminv(P,mu,sigma)
[X,XLO,XUP] = norminv(P,mu,sigma,pcov,alpha)
Description X = norminv(P,mu,sigma) computes the inverse of the normal cdf
using the corresponding mean mu and standard deviation sigma at
the corresponding probabilities in P. P, mu, and sigma can be vectors,
matrices, or multidimensional arrays that all have the same size. A
scalar input is expanded to a constant array with the same dimensions
as the other inputs. The parameters in sigma must be positive, and the
values in P must lie in the interval [0 1].
[X,XLO,XUP] = norminv(P,mu,sigma,pcov,alpha) produces
confidence bounds for X when the input parameters mu and sigma are
estimates. pcov is the covariance matrix of the estimated parameters.
alpha specifies 100(1 - alpha)% confidence bounds. The default value of
alpha is 0.05. XLO and XUP are arrays of the same size as X containing
the lower and upper confidence bounds.
The function norminv computes confidence bounds for P using a normal
approximation to the distribution of the estimate
ˆ + ˆ q
where q is the Pth quantile from a normal distribution with mean 0 and
standard deviation 1. The computed bounds give approximately the
desired confidence level when you estimate mu, sigma, and pcov from
large samples, but in smaller samples other methods of computing the
confidence bounds may be more accurate.
The normal inverse function is defined in terms of the normal cdf as
x = F −1 ( p| , ) = { x : F ( x | , ) = p}
where
18-963
norminv
−(t − ) 2
1 x
p = F ( x| , ) = ∫−∞ e 2 dt
2
2
The result, x, is the solution of the integral equation above where you
supply the desired probability, p.
Examples Find an interval that contains 95% of the values from a standard
normal distribution.
x = norminv([0.025 0.975],0,1)
x =
-1.9600 1.9600
Note that the interval x is not the only such interval, but it is the
shortest.
xl = norminv([0.01 0.96],0,1)
xl =
-2.3263 1.7507
The interval xl also contains 95% of the probability, but it is longer
than x.
See Also icdf, normcdf, normpdf, normstat, normfit, normlike, normrnd
“Normal Distribution” on page B-83
18-964
normlike
Purpose Normal negative log-likelihood
Syntax nlogL = normlike(params,data)
[nlogL,AVAR] = normlike(params,data)
[...] = normlike(param,data,censoring)
[...] = normlike(param,data,censoring,freq)
Description nlogL = normlike(params,data) returns the negative of the normal
log-likelihood function. params(1) is the mean, mu, and params(2)
is the standard deviation, sigma.
[nlogL,AVAR] = normlike(params,data) also returns the inverse of
Fisher’s information matrix, AVAR. If the input parameter values in
params are the maximum likelihood estimates, the diagonal elements
of AVAR are their asymptotic variances. AVAR is based on the observed
Fisher’s information, not the expected information.
[...] = normlike(param,data,censoring) accepts a Boolean
vector, censoring, of the same size as data, which is 1 for observations
that are right-censored and 0 for observations that are observed exactly.
[...] = normlike(param,data,censoring,freq) accepts a
frequency vector, freq, of the same size as data. The vector freq
typically contains integer frequencies for the corresponding elements in
data, but can contain any nonnegative values. Pass in [] for censoring
to use its default value.
normlike is a utility function for maximum likelihood estimation.
See Also normfit, normpdf, normcdf, norminv, normstat, normrnd
“Normal Distribution” on page B-83
18-965
normpdf
Purpose Normal probability density function
Syntax Y = normpdf(X,mu,sigma)
Description Y = normpdf(X,mu,sigma) computes the pdf at each of the values in
X using the normal distribution with mean mu and standard deviation
sigma. X, mu, and sigma can be vectors, matrices, or multidimensional
arrays that all have the same size. A scalar input is expanded to a
constant array with the same dimensions as the other inputs. The
parameters in sigma must be positive.
The normal pdf is
−( x − )2
1
y = f ( x| , ) = e 2
2
2
The likelihood function is the pdf viewed as a function of the parameters.
Maximum likelihood estimators (MLEs) are the values of the
parameters that maximize the likelihood function for a fixed value of x.
The standard normal distribution has µ = 0 and σ = 1.
If x is standard normal, then xσ + µ is also normal with mean µ and
standard deviation σ. Conversely, if y is normal with mean µ and
standard deviation σ, then x = (y-µ) / σ is standard normal.
Examples mu = [0:0.1:2];
[y i] = max(normpdf(1.5,mu,1));
MLE = mu(i)
MLE =
1.5000
See Also pdf, normcdf, norminv, normstat, normfit, normlike, normrnd, mvnpdf
“Normal Distribution” on page B-83
18-966
normplot
Purpose Normal probability plot
Syntax h = normplot(X)
Description h = normplot(X) displays a normal probability plot of the data in X.
For matrix X, normplot displays a line for each column of X. h is a handle
to the plotted lines.
The plot has the sample data displayed with the plot symbol '+'.
Superimposed on the plot is a line joining the first and third quartiles
of each column of X (a robust linear fit of the sample order statistics.)
This line is extrapolated out to the ends of the sample to help evaluate
the linearity of the data.
The purpose of a normal probability plot is to graphically assess whether
the data in X could come from a normal distribution. If the data are
normal the plot will be linear. Other distribution types will introduce
curvature in the plot. normplot uses midpoint probability plotting
positions. Use probplot when the data included censored observations.
Examples Generate a normal sample and a normal probability plot of the data.
x = normrnd(10,1,25,1);
normplot(x)
18-967
normplot
See Also cdfplot, wblplot, probplot, hist, normfit, norminv, normpdf,
normspec, normstat, normcdf, normrnd, normlike
“Normal Distribution” on page B-83
18-968
normrnd
Purpose Normal random numbers
Syntax R = normrnd(mu,sigma)
R = normrnd(mu,sigma,v)
R = normrnd(mu,sigma,m,n)
Description R = normrnd(mu,sigma) generates random numbers from the normal
distribution with mean parameter mu and standard deviation parameter
sigma. mu and sigma can be vectors, matrices, or multidimensional
arrays that have the same size, which is also the size of R. A scalar
input for mu or sigma is expanded to a constant array with the same
dimensions as the other input.
R = normrnd(mu,sigma,v) generates random numbers from the
normal distribution with mean parameter mu and standard deviation
parameter sigma, where v is a row vector. If v is a 1-by-2 vector, R
is a matrix with v(1) rows and v(2) columns. If v is 1-by-n, R is an
n-dimensional array.
R = normrnd(mu,sigma,m,n) generates random numbers from the
normal distribution with mean parameter mu and standard deviation
parameter sigma, where scalars m and n are the row and column
dimensions of R.
Examples n1 = normrnd(1:6,1./(1:6))
n1 =
2.1650 2.3134 3.0250 4.0879 4.8607 6.2827
n2 = normrnd(0,1,[1 5])
n2 =
0.0591 1.7971 0.2641 0.8717 -1.4462
n3 = normrnd([1 2 3;4 5 6],0.1,2,3)
n3 =
0.9299 1.9361 2.9640
4.1246 5.0577 5.9864
18-969
normrnd
See Also random, normpdf, normcdf, norminv, normstat, normfit, normlike,
mvnrnd
“Normal Distribution” on page B-83
18-970
normspec
Purpose Normal density plot between specifications
Syntax normspec(specs)
normspec(specs,mu,sigma)
normspec(specs,mu,sigma,region)
p = normspec(...)
[p,h] = normspec(...)
Description normspec(specs) plots the standard normal density, shading the
portion inside the specification limits given by the two-element vector
specs. Set specs(1) to -Inf if there is no lower limit; set specs(2)
to Inf if there is no upper limit.
normspec(specs,mu,sigma) shades the portion inside the specification
limits of a normal density with parameters mu and sigma. The defaults
are mu = 0 and sigma = 1.
normspec(specs,mu,sigma,region) shades the region either
'inside' or 'outside' the specification limits. The default is
'inside'.
p = normspec(...) also returns the probability, p, of the shaded area.
[p,h] = normspec(...) also returns a handle h to the line objects.
Examples A production process fills cans of paint. The average amount of paint in
any can is 1 gallon, but variability in the process produces a standard
deviation of 2 ounces (2/128 gallons). What is the probability that cans
will be filled under specification by 3 or more ounces?
p = normspec([1-3/128,Inf],1,2/128,'outside')
p =
0.0668
18-971
normspec
See Also capaplot, histfit
“Normal Distribution” on page B-83
18-972
normstat
Purpose Normal mean and variance
Syntax [M,V] = normstat(mu,sigma)
Description [M,V] = normstat(mu,sigma) returns the mean of and variance
for the normal distribution using the corresponding mean mu and
standard deviation sigma. mu and sigma can be vectors, matrices, or
multidimensional arrays that all have the same size, which is also the
size of M and V. A scalar input for mu or sigma is expanded to a constant
array with the same dimensions as the other input.
The mean of the normal distribution with parameters µ and σ is µ, and
the variance is σ2.
Examples n = 1:5;
[m,v] = normstat(n'*n,n'*n)
m =
1 2 3 4 5
2 4 6 8 10
3 6 9 12 15
4 8 12 16 20
5 10 15 20 25
v =
1 4 9 16 25
4 16 36 64 100
9 36 81 144 225
16 64 144 256 400
25 100 225 400 625
See Also normpdf, normcdf, norminv, normfit, normlike, normrnd
“Normal Distribution” on page B-83
18-973
piecewisedistribution.nsegments
Purpose Number of segments
Syntax n = nsegments(obj)
Description n = nsegments(obj) returns the number of segments n in the
piecewise distribution object obj.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
n = nsegments(obj)
n =
3
See Also paretotails, boundary, segment
18-974
CompactTreeBagger.NTrees property
Purpose Number of decision trees in ensemble
Description The NTrees property is a scalar equal to the number of decision trees
in the ensemble.
See Also Trees
18-975
TreeBagger.NTrees property
Purpose Number of decision trees in ensemble
Description The NTrees property is a scalar equal to the number of decision trees
in the ensemble.
See Also Trees
18-976
ProbDistParametric.NumParams property
Purpose Read-only value specifying number of parameters of ProbDistParametric
object
Description NumParams is a read-only property of the ProbDistParametric class.
NumParams is a value specifying the number of parameters of a
distribution represented by a ProbDistParametric object.
Values This value is an integer that counts both the specified parameters and
parameters that are fit to the data. Use this information to view and
compare the number of parameters supplied to create distributions.
18-977
dataset.numel
Purpose Number of elements in dataset array
Syntax n = numel(A)
n = numel(A, varargin)
Description n = numel(A) returns 1. To find the number of elements, n, in the
dataset array A, use prod(size(A)) or numel(A,':',':').
n = numel(A, varargin) returns the number of subscripted elements,
n, in A(index1, index2, ..., indexn), where varargin is a cell
array whose elements are index1, index2, ... indexn.
See Also length, size
18-978
categorical.numel
Purpose Number of elements in categorical array
Syntax n = numel(A)
n = numel(A, varargin)
Description n = numel(A) returns the number of elements in the categorical array
A.
n = numel(A, varargin) returns the number of subscripted elements,
n, in A(index1, index2, ..., indexN), where varargin is a cell
array whose elements are index1, index2, ... indexN.
See Also size
18-979
classregtree.numnodes
Purpose Number of nodes
Syntax n = numnodes(t)
Description n = numnodes(t) returns the number of nodes n in the tree t.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t=
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-980
classregtree.numnodes
n = numnodes(t)
n =
9
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree
18-981
cvpartition.NumTestSets property
Purpose Number of test sets
Description Value is the number of folds in partitions of type 'kfold' and
'leaveout'.
Value is 1 in partitions of type 'holdout' and 'resubstitution'.
18-982
TreeBagger.NVarToSample property
Purpose Number of variables for random feature selection
Description The NVarToSample property specifies the number of predictor or feature
variables to select at random for each decision split. By default, it is set
to the square root of the total number of variables for classification and
one third of the total number of variables for regression. Setting this
argument to any valid value except 'all' invokes Breiman’s "random
forest" algorithm.
See Also classregtree
18-983
dataset.ObsNames property
Purpose Cell array of nonempty, distinct strings giving names of observations
in data set
Description A cell array of nonempty, distinct strings giving the names of the
observations in the data set. This property may be empty, but if not
empty, the number of strings must equal the number of observations.
18-984
TreeBagger.oobError
Purpose Out-of-bag error
Syntax err = oobError(B)
err = oobError(B,'param1',val1,'param2',val2,...)
Description err = oobError(B) computes the misclassification probability (for
classification trees) or mean squared error (for regression trees) for
out-of-bag observations in the training data, using the trained bagger B.
err is a vector of length NTrees, where NTrees is the number of trees
in the ensemble.
err = oobError(B,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'mode' String indicating how oobError computes errors.
If set to 'cumulative' (default), the method
computes cumulative errors and err is a vector of
length NTrees, where the first element gives error
from trees(1), second element gives error from
trees(1:2) etc, up to trees(1:NTrees). If set to
'individual', err is a vector of length NTrees,
where each element is an error from each tree
in the ensemble. If set to 'ensemble', err is a
scalar showing the cumulative error for the entire
ensemble.
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements
in the input vector, and a scalar for 'ensemble'
mode. For example, in the 'cumulative' mode, the
first element gives error from trees(1), the second
element gives error from trees(1:2) etc.
18-985
TreeBagger.oobError
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. oobError uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple nonweighted majority vote. You cannot use
this argument in the 'individual' mode.
See Also CompactTreeBagger.error
18-986
TreeBagger.OOBIndices property
Purpose Indicator matrix for out-of-bag observations
Description The OOBIndices property is a logical array of size Nobs-by-NTrees
where Nobs is the number of observations in the training data and
NTrees is the number of trees in the ensemble. The (I,J) element is
true if observation I is out-of-bag for tree J and false otherwise. In
other words, a true value means observation I was not selected for the
training data used to grow tree J.
See Also classregtree
18-987
TreeBagger.OOBInstanceWeight property
Purpose Count of out-of-bag trees for each observation
Description The OOBInstanceWeight property is a numeric array of size Nobs-by-1
containing the number of trees used for computing out-of-bag response
for each observation. Nobs is the number of observations in the training
data used to create the ensemble.
18-988
TreeBagger.oobMargin
Purpose Out-of-bag margins
Syntax mar = oobMargin(B)
mar = oobMargin(B,'param1',val1,'param2',val2,...)
Description mar = oobMargin(B) computes an Nobs-by-NTrees matrix of
classification margins for out-of-bag observations in the training data,
using the trained bagger B.
mar = oobMargin(B,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'mode' String indicating how oobMargin computes
errors. If set to 'cumulative' (default), the
method computes cumulative margins and mar
is an Nobs-by-NTrees matrix, where the first
column gives margins from trees(1), second
column gives margins from trees(1:2) etc, up to
trees(1:NTrees). If set to 'individual', mar is
an Nobs-by-NTrees matrix, where each column
gives margins from each tree in the ensemble. If
set to 'ensemble', mar is a single column of length
Nobs showing the cumulative margins for the
entire ensemble.
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is
set to 'all' and the method uses all trees. If
’trees’ is a numeric vector, the method returns
an Nobs-by-NTrees matrix for 'cumulative' and
'individual' modes, where NTrees is the number
of elements in the input vector, and a single
column for 'ensemble' mode. For example, in
the 'cumulative' mode, the first column gives
margins from trees(1), the second column gives
margins from trees(1:2) etc.
18-989
TreeBagger.oobMargin
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. oobMargin uses
these weights to combine output from the specified
trees by taking a weighted average instead of the
simple nonweighted majority vote. You cannot use
this argument in the 'individual' mode.
See Also CompactTreeBagger.margin
18-990
TreeBagger.oobMeanMargin
Purpose Out-of-bag mean margins
Syntax mar = oobMeanMargin(B)
mar = oobMeanMargin(B,'param1',val1,'param2',val2,...)
Description mar = oobMeanMargin(B) computes average classification margins for
out-of-bag observations in the training data, using the trained bagger B.
oobMeanMargin averages the margins over all out-of-bag observations.
mar is a row-vector of length NTrees, where NTrees is the number
of trees in the ensemble.
mar = oobMeanMargin(B,'param1',val1,'param2',val2,...)
specifies optional parameter name/value pairs:
'mode' String indicating how oobMargin computes errors. If
set to 'cumulative' (default), is a vector of length
NTrees where the first element gives mean margin
from trees(1), second column gives mean margins
from trees(1:2) etc, up to trees(1:NTrees). If set
to 'individual', mar is a vector of length NTrees,
where each element is a mean margin from each tree
in the ensemble . If set to 'ensemble', mar is a scalar
showing the cumulative mean margin for the entire
ensemble .
'trees' Vector of indices indicating what trees to include
in this calculation. By default, this argument is set
to 'all' and the method uses all trees. If 'trees'
is a numeric vector, the method returns a vector of
length NTrees for 'cumulative' and 'individual'
modes, where NTrees is the number of elements in the
input vector, and a scalar for 'ensemble' mode. For
example, in the 'cumulative' mode, the first element
gives mean margin from trees(1), the second
element gives mean margin from trees(1:2) etc.
'treeweights' Vector of tree weights. This vector must have the
same length as the 'trees' vector. oobMeanMargin
uses these weights to combine output from the
specified trees by taking a weighted average instead
of the simple nonweighted majority vote. You cannot
use this argument in the 'individual' mode.
18-991
TreeBagger.oobMeanMargin
See Also CompactTreeBagger.meanMargin
18-992
TreeBagger.OOBPermutedVarCountRaiseMargin
property
Purpose Variable importance for raising margin
Description The OOBPermutedVarCountRaiseMargin property is a numeric array of
size 1-by-Nvars containing a measure of variable importance for each
predictor. For any variable, the measure is the difference between the
number of raised margins and the number of lowered margins if the
values of that variable are permuted across the out-of-bag observations.
This measure is computed for every tree, then averaged over the
entire ensemble and divided by the standard deviation over the entire
ensemble. This property is empty for regression trees.
18-993
TreeBagger.OOBPermutedVarDeltaError property
Purpose Variable importance for classification error
Description The OOBPermutedVarDeltaError property is a numeric array of size
1-by-Nvars containing a measure of importance for each predictor
variable (feature). For any variable, the measure is the increase in
classification if the values of that variable are permuted across the
out-of-bag observations. This measure is computed for every tree,
then averaged over the entire ensemble and divided by the standard
deviation over the entire ensemble.
18-994
TreeBagger.OOBPermutedVarDeltaMeanMargin
property
Purpose Variable importance for classification margin
Description The OOBPermutedVarDeltaMeanMargin property is a numeric array of
size 1-by-Nvars containing a measure of importance for each predictor
variable (feature). For any variable, the measure is the decrease in
the classification margin if the values of that variable are permuted
across the out-of-bag observations. This measure is computed for
every tree, then averaged over the entire ensemble and divided by the
standard deviation over the entire ensemble. This property is empty
for regression trees.
18-995
TreeBagger.oobPredict
Purpose Ensemble predictions for out-of-bag observations
Syntax Y = oobPredict(B)
Y = oobPredict(B,'param1',val1,'param2',val2,...)
Description Y = oobPredict(B) computes predicted responses using the trained
bagger B for out-of-bag observations in the training data. The output
has one prediction for each observation in the training data. The
returned Y is a cell array of strings for classification and a numeric
array for regression.
Y = oobPredict(B,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
'trees' Array of tree indices to use for computation of
responses. Default is 'all'.
'treeweights' Array of NTrees weights for weighting votes from
the specified trees.
See Also CompactTreeBagger.predict, OOBIndices
18-996
ordinal class
Superclasses categorical
Purpose Arrays for ordinal categorical data
Description Ordinal arrays are used to store discrete values that have an ordering
but are not numeric. An ordinal array provides efficient storage
and convenient manipulation of such data, while also maintaining
meaningful labels for the values. Ordinal arrays are often used as
grouping variables.
Like a numerical array, an ordinal array can have any size or
dimension. You can subscript, concatenate, reshape, sort, etc. ordinal
arrays, much like ordinary numeric arrays. You can make comparisons
between elements of two ordinal arrays, or between an ordinal array
and a single string representing a ordinal value.
Construction Use the ordinal constructor to create an ordinal array from a numeric,
logical, or character array, or from a cell array of strings.
ordinal Construct ordinal categorical
array
Methods Each ordinal array carries along a list of possible values that it can
store, known as its levels. The list is created when you create an
ordinal array, and you can access it using the getlevels method, or
modify it using the addlevels, mergelevels, or droplevels methods.
Assignment to the array will also add new levels automatically if the
values assigned are not already levels of the array. The ordering on
values stored in an ordinal array is defined by the order of the list of
levels. You can change that order using the reorderlevels method.
The following table lists operations available for ordinal arrays.
ismember Test for membership
mergelevels Merge levels
18-997
ordinal class
sort Sort elements of ordinal array
sortrows Sort rows
Inherited Methods
Methods in the following table are inherited from categorical.
addlevels Add levels to categorical array
cat Concatenate categorical arrays
cellstr Convert categorical array to cell
array of strings
char Convert categorical array to
character array
circshift Shift categorical array circularly
ctranspose Transpose categorical matrix
disp Display categorical array
display Display categorical array
double Convert categorical array to
double array
droplevels Drop levels
end Last index in indexing expression
for categorical array
flipdim Flip categorical array along
specified dimension
fliplr Flip categorical matrix in
left/right direction
flipud Flip categorical matrix in
up/down direction
getlabels Access categorical array labels
getlevels Get categorical array levels
18-998
ordinal class
hist Plot histogram of categorical data
horzcat Horizontal concatenation for
categorical arrays
int16 Convert categorical array to
signed 16-bit integer array
int32 Convert categorical array to
signed 32-bit integer array
int64 Convert categorical array to
signed 64-bit integer array
int8 Convert categorical array to
signed 8-bit integer array
intersect Set intersection for categorical
arrays
ipermute Inverse permute dimensions of
categorical array
isempty True for empty categorical array
isequal True if categorical arrays are
equal
islevel Test for levels
ismember True for elements of categorical
array in set
isscalar True if categorical array is scalar
isundefined Test for undefined elements
isvector True if categorical array is vector
length Length of categorical array
levelcounts Element counts by level
ndims Number of dimensions of
categorical array
18-999
ordinal class
numel Number of elements in categorical
array
permute Permute dimensions of
categorical array
reorderlevels Reorder levels
repmat Replicate and tile categorical
array
reshape Resize categorical array
rot90 Rotate categorical matrix 90
degrees
setdiff Set difference for categorical
arrays
setlabels Label levels
setxor Set exclusive-or for categorical
arrays
shiftdim Shift dimensions of categorical
array
single Convert categorical array to
single array
size Size of categorical array
squeeze Squeeze singleton dimensions
from categorical array
subsasgn Subscripted assignment for
categorical array
subsindex Subscript index for categorical
array
subsref Subscripted reference for
categorical array
18-1000
ordinal class
summary Summary statistics for categorical
array
times Product of categorical arrays
transpose Transpose categorical matrix
uint16 Convert categorical array to
unsigned 16-bit integers
uint32 Convert categorical array to
unsigned 32-bit integers
uint64 Convert categorical array to
unsigned 64-bit integers
uint8 Convert categorical array to
unsigned 8-bit integers
union Set union for categorical arrays
unique Unique values in categorical
array
vertcat Vertical concatenation for
categorical arrays
Properties Inherited Properties
Properties in the following table are inherited from categorical.
labels Text labels for levels
undeflabel Text label for undefined levels
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
Examples Create an ordinal array from integer data:
quality = ordinal([1 2 3; 3 2 1; 2 1 3],{'low' 'medium' 'high'})
18-1001
ordinal class
% Find elements meeting a criterion
quality >= 'medium'
ismember(quality,{'low' 'high'})
% Compare two ordinal arrays
quality2 = fliplr(quality)
quality == quality2
References [1] Johnson, N. L., S. Kotz, and A. W. Kemp, Univariate Discrete
Distributions, 2nd edition, Wiley, 1992, pp. 124-130.
See Also histc, nominal
18-1002
ordinal
Purpose Construct ordinal categorical array
Syntax B = ordinal(A)
B = ordinal(A,labels)
B = ordinal(A,labels,levels)
B = ordinal(A,labels,[],edges)
Description B = ordinal(A) creates an ordinal array B from the array A. A can
be numeric, logical, character, categorical, or a cell array of strings.
ordinal creates the levels of B from the sorted unique values in A, and
creates default labels for them.
B = ordinal(A,labels) labels the levels in B using the character array
or cell array of strings labels. ordinal assigns labels to levels in B in
order according to the sorted unique values in A.
B = ordinal(A,labels,levels) creates an ordinal array with
possible levels defined by levels. levels is a vector whose values
can be compared to those in A using the equality operator. ordinal
assigns labels to each level from the corresponding elements of labels.
If A contains any values not present in levels, the levels of the
corresponding elements of B are undefined. Use [] for labels to allow
ordinal to create default labels.
B = ordinal(A,labels,[],edges) creates an ordinal array by binning
the numeric array A, with bin edges given by the numeric vector
edges. The uppermost bin includes values equal to the right-most
edge. ordinal assigns labels to each level in B from the corresponding
elements of labels. edges must have one more element than labels.
By default, an element of B is undefined if the corresponding element
of A is NaN (when A is numeric), an empty string (when A is character),
or undefined (when A is categorical). ordinal treats such elements as
“undefined” or “missing” and does not include entries for them among
the possible levels for B. To create an explicit level for such elements
instead of treating them as undefined, you must use the levels input,
and include NaN, the empty string, or an undefined element.
18-1003
ordinal
You may include duplicate labels in labels in order to merge multiple
values in A into a single level in B.
Examples Create an ordinal array from integer data, and provide explicit labels:
quality1 = ordinal([1 2 3; 3 2 1; 2 1 3],...
{'low' 'medium' 'high'})
Create an ordinal array from integer data, and provide both explicit
labels and an explicit order:
quality2 = ordinal([1 2 3; 3 2 1; 2 1 3],...
{'high' 'medium' 'low'},[3 2 1])
Create an ordinal array by binning floating point values:
size = ordinal(rand(5,2),{'small' 'medium' 'large'},...
[],[0 1/3 2/3 1])
Create an ordinal array from the measurements in Fisher’s iris data,
ignoring decimal lengths:
load fisheriris
m = floor(min(meas(:)));
M = floor(max(meas(:)));
labels = num2str((m:M)');
edges = m:M+1;
cms = ordinal(meas,labels,[],edges)
meas(1:5,:)
ans =
5.1000 3.5000 1.4000 0.2000
4.9000 3.0000 1.4000 0.2000
18-1004
ordinal
4.7000 3.2000 1.3000 0.2000
4.6000 3.1000 1.5000 0.2000
5.0000 3.6000 1.4000 0.2000
cms(1:5,:)
ans =
5 3 1 0
4 3 1 0
4 3 1 0
4 3 1 0
5 3 1 0
Create an age group ordinal array from the data in hospital.mat:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
AgeGroup = ordinal(hospital.Age,labels,[],edges);
hospital.Age(1:5)
ans =
38
43
38
40
49
AgeGroup(1:5)
ans =
30s
40s
30s
40s
40s
See Also histc | nominal
18-1005
CompactTreeBagger.outlierMeasure
Purpose Outlier measure for data
Syntax out = outlierMeasure(B,X)
out = outlierMeasure(B,X,'param1',val1,'param2',val2,...)
Description out = outlierMeasure(B,X) computes outlier measures for predictors
X using trees in the ensemble B. The method computes the outlier
measure for a given observation by taking an inverse of the average
squared proximity between this observation and other observations.
outlierMeasure then normalizes these outlier measures by subtracting
the median of their distribution, taking the absolute value of this
difference, and dividing by the median absolute deviation. A high value
of the outlier measure indicates that this observation is an outlier.
You can supply the proximity matrix directly by using the 'data'
parameter.
out = outlierMeasure(B,X,'param1',val1,'param2',val2,...)
specifies optional parameter name/value pairs:
'data' Flag indicating how to treat the X input argument. If
set to 'predictors' (default), the method assumes X
is a matrix of predictors and uses it for computation
of the proximity matrix. If set to 'proximity', the
method treats X as a proximity matrix returned
by the proximity method. If you do not supply
the proximity matrix, outlierMeasure computes
it internally. If you use the proximity method to
compute a proximity matrix, supplying it as input to
outlierMeasure reduces computing time.
'labels' Vector of true class labels. True class labels can be
either a numeric vector, character matrix, or cell
array of strings. When you supply this parameter,
the method performs the outlier calculation for any
observations using only other observations from the
same class. This parameter must specify one label for
each observation (row) in X.
18-1006
CompactTreeBagger.outlierMeasure
See Also proximity
18-1007
TreeBagger.OutlierMeasure property
Purpose Measure for determining outliers
Description The OutlierMeasure property is a numeric array of size Nobs-by-1,
where Nobs is the number of observations in the training data,
containing outlier measures for each observation.
See Also CompactTreeBagger.outlierMeasure
18-1008
parallelcoords
Purpose Parallel coordinates plot
Syntax parallelcoords(X)
parallelcoords(X,...,'Standardize','on')
parallelcoords(X,...,'Standardize','PCA')
parallelcoords(X,...,'Standardize','PCAStd')
parallelcoords(X,...,'Quantile',alpha)
parallelcoords(X,...,'Group',group)
parallelcoords(X,...,'Labels',labels)
parallelcoords(X,...,PropertyName,PropertyValue,...)
h = parallelcoords(X,...)
parallelcoords(axes,...)
Description parallelcoords(X) creates a parallel coordinates plot of the
multivariate data in the n-by-p matrix X. Rows of X correspond to
observations, columns to variables. A parallel coordinates plot is a
tool for visualizing high dimensional data, where each observation is
represented by the sequence of its coordinate values plotted against
their coordinate indices. parallelcoords treats NaNs in X as missing
values and does not plot those coordinate values.
parallelcoords(X,...,'Standardize','on') scales each column of X
to have mean 0 and standard deviation 1 before making the plot.
parallelcoords(X,...,'Standardize','PCA') creates a parallel
coordinates plot from the principal component scores of X, in order of
decreasing eigenvalues. parallelcoords removes rows of X containing
missing values (NaNs) for principal components analysis (PCA)
standardization.
parallelcoords(X,...,'Standardize','PCAStd') creates a parallel
coordinates plot using the standardized principal component scores.
parallelcoords(X,...,'Quantile',alpha) plots only the median
and the alpha and 1-alpha quantiles of f (t) at each value of t. This is
useful if X contains many observations.
parallelcoords(X,...,'Group',group) plots the data in different
groups with different colors. Groups are defined by group, a numeric
18-1009
parallelcoords
array containing a group index for each observation. (See “Grouped
Data” on page 2-34.) group can also be a categorical variable, character
matrix, or cell array of strings, containing a group name for each
observation.
parallelcoords(X,...,'Labels',labels) labels the coordinate tick
marks along the horizontal axis using labels, a character array or
cell array of strings.
parallelcoords(X,...,PropertyName,PropertyValue,...) sets
properties to the specified property values for all line graphics objects
created by parallelcoords.
h = parallelcoords(X,...) returns a column vector of handles to the
line objects created by parallelcoords, one handle per row of X. If you
use the 'Quantile' input argument, h contains one handle for each of
the three lines objects created. If you use both the 'Quantile' and the
'Group' input arguments, h contains three handles for each group.
parallelcoords(axes,...) plots into the axes with handle axes.
Examples % Make a grouped plot of the raw data
load fisheriris
labels = {'Sepal Length','Sepal Width',...
'Petal Length','Petal Width'};
parallelcoords(meas,'group',species,'labels',labels);
% Plot only the median and quartiles of each group
parallelcoords(meas,'group',species,'labels',labels,...
'quantile',.25);
See Also “Grouped Data” on page 2-34
andrewsplot, glyphplot
18-1010
ProbDistUnivParam.paramci
Purpose Return parameter confidence intervals of ProbDistUnivParam object
Syntax CI = paramci(PD)
CI = paramci(PD, Alpha)
Description CI = paramci(PD) returns CI, a 2-by-N array containing 95%
confidence intervals for the parameters of the ProbDistUnivParam
object PD. N is the number of parameters in the distribution.
When you create PD by specifying parameters (such as using the
ProbDistUnivParam.ProbDistUnivParam constructor or using the
fitdist function and specifying a 'binomial' or 'generalized
pareto' distribution) rather than by fitting to data, the confidence
intervals have a width of 0 because the parameters are viewed as
estimates of an unknown parameter.
CI = paramci(PD, Alpha) returns 100*(1 - Alpha)% confidence
intervals. Default Alpha is 0.05, which specifies 95% confidence
intervals.
Note If you create PD with a distribution that does not support
confidence intervals, then CI contains NaN values.
Input PD An object of the class ProbDistUnivParam.
Arguments
Alpha A value between 0 and 1 that specifies a
confidence interval. Default is 0.05, which
specifies 95% confidence intervals.
Output CI A 2-by-N array containing 100*(1 - Alpha)%
Arguments confidence intervals for the parameters of
the ProbDistUnivParam object PD. N is the
number of parameters in the distribution.
18-1011
ProbDistUnivParam.paramci
See Also fitdist
18-1012
ProbDistParametric.ParamCov property
Purpose Read-only covariance matrix of parameter estimates of
ProbDistParametric object
Description ParamCov is a read-only property of the ProbDistParametric class.
ParamCov is a covariance matrix containing the parameter estimates of
a distribution represented by a ProbDistParametric object. ParamCov
has a size of NumParams-by-NumParams.
Values This covariance matrix includes estimates for both the specified
parameters and parameters that are fit to the data. For specified
parameters, the covariance is 0, indicating the parameter is known
exactly. Use this information to view and compare the descriptions of
parameters supplied to create distributions.
18-1013
ProbDistParametric.ParamDescription property
Purpose Read-only cell array specifying descriptions of parameters of
ProbDistParametric object
Description ParamDescription is a read-only property of the ProbDistParametric
class. ParamDescription is a cell array of strings specifying the
descriptions or meanings of the parameters of a distribution represented
by a ProbDistParametric object. ParamDescription has a length of
NumParams.
Values This cell array includes a brief description of the meaning of both
the specified parameters and parameters that are fit to the data.
The description is the same as the parameter name when no further
description information is available. Use this information to view and
compare the descriptions of parameters used to create distributions.
18-1014
ProbDistParametric.ParamIsFixed property
Purpose Read-only logical array specifying fixed parameters of
ProbDistParametric object
Description ParamIsFixed is a read-only property of the ProbDistParametric class.
ParamIsFixed is a logical array specifying the fixed parameters of a
distribution represented by a ProbDistParametric object. ParamIsFixed
has a length of NumParams.
Values This array specifies a 1 (true) for fixed parameters, and a 0 (false)
for parameters that are estimated from the input data. Use this
information to view and compare the fixed parameters used to create
distributions.
18-1015
ProbDistParametric.ParamNames property
Purpose Read-only cell array specifying names of parameters of
ProbDistParametric object
Description ParamNames is a read-only property of the ProbDistParametric class.
ParamNames is a cell array of strings specifying the names of the
parameters of a distribution represented by a ProbDistParametric
object. ParamNames has a length of NumParams.
Values This cell array includes the names of both the specified parameters and
parameters that are fit to the data. Use this information to view and
compare the names of parameters used to create distributions.
18-1016
NaiveBayes.Params property
Purpose Parameter estimates
Description The Params property is an NClasses-by-NDims cell array containing
the parameter estimates, excluding the class priors. Params(i,j)
contains the parameter estimates for the jth feature in the ith class.
Params(i,j) is an empty cell if the ith class is empty.
The entry in Params(i,j) depends on the distribution type used for
the jth feature, as follows:
'normal' A vector of length two. The first element is the
mean, and the second element is standard deviation.
'kernel' A ProbDistUnivKernel object
'mvmn' A vector containing the probability for each possible
value of the jth feature in the ith class. The order
of the probabilities is decided by the sorted order of
all the unique values of the jth feature.
'mn' A scalar representing the probability the jth
token appearing in the ith class, Prob(token j
| class i). It is estimated as (1 + the number
of occurrence of token J in class I)/(NDims
+ the total number of token occurrence in
class I).
18-1017
ProbDistParametric.Params property
Purpose Read-only array specifying values of parameters of ProbDistParametric
object
Description Params is a read-only property of the ProbDistParametric class.
Params is an array of values specifying the values of the parameters of
a distribution represented by a ProbDistParametric object. Params
has a length of NumParams.
Values This array includes the values of both the specified parameters and
parameters that are fit to the data. Use this information to view and
compare the values of parameters used to create distributions.
18-1018
classregtree.parent
Purpose Parent node
Syntax p = parent(t)
p = parent(t,nodes)
Description p = parent(t) returns an n-element vector p containing the number of
the parent node for each node in the tree t, where n is the number of
nodes. The parent of the root node is 0.
p = parent(t,nodes) takes a vector nodes of node numbers and
returns the parent nodes for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 else node 3
2 class = setosa
3 if PW<1.75 then node 4 else node 5
4 if PL<4.95 then node 6 else node 7
5 class = virginica
6 if PW<1.65 then node 8 else node 9
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-1019
classregtree.parent
p = parent(t)
p =
0
1
1
3
3
4
4
6
18-1020
classregtree.parent
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, children, numnodes
18-1021
pareto
Purpose Pareto chart
Syntax pareto(y,names)
[h,ax] = pareto(...)
Description pareto(y,names) displays a Pareto chart where the values in the vector
y are drawn as bars in descending order. Each bar is labeled with the
associated value in the string matrix or cell array, names. pareto(y)
labels each bar with the index of the corresponding element in y.
The line above the bars shows the cumulative percentage.
[h,ax] = pareto(...) returns a combination of patch and line object
handles to the two axes created in ax.
Examples Create a Pareto chart from data measuring the number of manufactured
parts rejected for various types of defects.
defects = {'pits';'cracks';'holes';'dents'};
quantity = [5 3 19 25];
pareto(quantity,defects)
18-1022
pareto
See Also bar, hist
18-1023
paretotails class
Superclasses piecewisedistribution
Purpose Empirical distributions with Pareto tails
Construction paretotails Construct Pareto tails object
Methods lowerparams Lower Pareto tails parameters
upperparams Upper Pareto tails parameters
Inherited Methods
Methods in the following table are inherited from
piecewisedistribution.
boundary Piecewise distribution boundaries
cdf Cumulative distribution function
for piecewise distribution
disp Display piecewisedistribution
object
display Display piecewisedistribution
object
icdf Inverse cumulative distribution
function for piecewise distribution
nsegments Number of segments
pdf Probability density function for
piecewise distribution
random Random numbers from piecewise
distribution
segment Segments containing values
18-1024
paretotails class
Properties Objects of the paretotails class have no properties accessible by dot
indexing, get methods, or set methods. To obtain information about a
paretotails object, use the appropriate method.
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
See Also “Generalized Pareto Distribution” on page B-37
18-1025
paretotails
Purpose Construct Pareto tails object
Syntax obj = paretotails(x,pl,pu)
obj = paretotails(x,pl,pu,cdffun)
Description obj = paretotails(x,pl,pu) creates an object obj defining a
distribution consisting of the empirical distribution of x in the center
and Pareto distributions in the tails. x is a real-valued vector of
data values whose extreme observations are fit to generalized Pareto
distributions (GPDs). pl and pu identify the lower- and upper-tail
cumulative probabilities such that 100*pl and 100*(1-pu) percent
of the observations in x are, respectively, fit to a GPD by maximum
likelihood. If pl is 0, or if there are not at least two distinct observations
in the lower tail, then no lower Pareto tail is fit. If pu is 1, or if there are
not at least two distinct observations in the upper tail, then no upper
Pareto tail is fit.
obj = paretotails(x,pl,pu,cdffun) uses cdffun to estimate the cdf
of x between the lower and upper tail probabilities. cdffun may be
any of the following:
• 'ecdf' — Uses an interpolated empirical cdf, with data values as the
midpoints in the vertical steps in the empirical cdf, and computed by
linear interpolation between data values. This is the default.
• 'kernel' — Uses a kernel-smoothing estimate of the cdf.
• @fun — Uses a handle to a function of the form [p,xi] = fun(x)
that accepts the input data vector x and returns a vector p of cdf
values and a vector xi of evaluation points. Values in xi must be
sorted and distinct but need not equal the values in x.
cdffun is used to compute the quantiles corresponding to pl and pu
by inverse interpolation, and to define the fitted distribution between
these quantiles.
The output object obj is a Pareto tails object with methods to evaluate
the cdf, inverse cdf, and other functions of the fitted distribution. These
methods are well-suited to copula and other Monte Carlo simulations.
18-1026
paretotails
The pdf method in the tails is the GPD density, but in the center it is
computed as the slope of the interpolated cdf.
The paretotails class is a subclass of the piecewisedistribution
class, and many of its methods are derived from that class.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
[p,q] = boundary(obj);
x = linspace(-5,5);
plot(x,cdf(obj,x),'b-','LineWidth',2)
hold on
plot(x,tcdf(x,3),'r:','LineWidth',2)
plot(q,p,'bo','LineWidth',2,'MarkerSize',5)
legend('Pareto Tails Object','t Distribution',...
'Location','NW')
18-1027
paretotails
See Also cdf, ecdf, gpfit, icdf, ksdensity
18-1028
partialcorr
Purpose Linear or rank partial correlation coefficients
Syntax RHO = partialcorr(X,Z)
RHO = partialcorr(X,Y,Z)
[RHO,PVAL] = partialcorr(...)
[...] = partialcorr(...,param1,val1,param2,val2,...)
Description RHO = partialcorr(X,Z) returns the sample linear partial correlation
coefficients between pairs of variables in X controlling for the variables
in Z. X is an n-by-p matrix, and Z is an n-by-q matrix with rows
corresponding to observations, and columns corresponding to variables.
The output, RHO, is a symmetric p-by-p matrix.
RHO = partialcorr(X,Y,Z) returns the sample linear partial
correlation coefficients between pairs of variables between X and
Y, controlling for the variables in Z. X is an n-by-p1 matrix, Y an
n-by-p2 matrix, and Z is an n-by-q matrix, with rows corresponding to
observations, and columns corresponding to variables. RHO is a p1-by-p2
matrix, where the (i, j)th entry is the sample linear partial correlation
between the ith column in X and the jth column in Y.
If the covariance matrix of [X,Z] is
⎛ S11 S12 ⎞
S=⎜ ⎟
⎝ S12T S22 ⎠
then the partial correlation matrix of X, controlling for Z, can be defined
formally as a normalized version of the covariance matrix Sxy = S11
– (S12S22–1S12T)
[RHO,PVAL] = partialcorr(...) also returns PVAL, a matrix of
p-values for testing the hypothesis of no partial correlation against the
alternative that there is a nonzero partial correlation. Each element of
PVAL is the p value for the corresponding element of RHO. If PVAL(I,J)
is small, say less than 0.05, then the partial correlation, RHO(I,J), is
significantly different from zero.
18-1029
partialcorr
[...] = partialcorr(...,param1,val1,param2,val2,...)
specifies additional parameters and their values. Valid parameter/value
pars are listed in the following table.
Parameter Values
'type' • 'Pearson' — To compute Pearson (linear)
partial correlations. This is the default.
• 'Spearman' — To compute Spearman (rank)
partial correlations.
'rows' • 'all' — To use all rows regardless of missing
(NaN) values. This is the default.
• 'complete' — To use only rows with no missing
values.
• 'pairwise' — To compute RHO(I,J) using rows
with no missing values in column I or J.
'tail' • 'both' (the default) — the correlation is not zero.
• 'right' — the correlation is greater than zero.
The alternative
hypothesis • 'left' — the correlation is less than zero.
against which
to compute
p-values for
testing the
hypothesis
of no partial
correlation.
A 'pairwise' value for the rows parameter can produce a RHO that is
not positive definite. A 'complete' value always produces a positive
definite RHO, but when data is missing, the estimates will be based on
fewer observations, in general.
partialcorr computes p-values for linear and rank partial correlations
using a Student’s t distribution for a transformation of the correlation.
18-1030
partialcorr
This is exact for linear partial correlation when X and Z are normal, but
is a large-sample approximation otherwise.
See Also corr, tiedrank, corrcoef
18-1031
pcacov
Purpose Principal component analysis on covariance matrix
Syntax COEFF = pcacov(V)
[COEFF,latent] = pcacov(V)
[COEFF,latent,explained] = pcacov(V)
Description COEFF = pcacov(V) performs principal components analysis on the
p-by-p covariance matrix V and returns the principal component
coefficients, also known as loadings. COEFF is a p-by-p matrix, with
each column containing coefficients for one principal component. The
columns are in order of decreasing component variance.
pcacov does not standardize V to have unit variances. To perform
principal components analysis on standardized variables, use the
correlation matrix R = V./(SD*SD'), where SD = sqrt(diag(V)), in
place of V. To perform principal components analysis directly on the
data matrix, use princomp.
[COEFF,latent] = pcacov(V) returns latent, a vector containing the
principal component variances, that is, the eigenvalues of V.
[COEFF,latent,explained] = pcacov(V) returns explained, a vector
containing the percentage of the total variance explained by each
principal component.
Examples load hald
covx = cov(ingredients);
[COEFF,latent,explained] = pcacov(covx)
COEFF =
0.0678 -0.6460 0.5673 -0.5062
0.6785 -0.0200 -0.5440 -0.4933
-0.0290 0.7553 0.4036 -0.5156
-0.7309 -0.1085 -0.4684 -0.4844
latent =
517.7969
67.4964
12.4054
18-1032
pcacov
0.2372
explained =
86.5974
11.2882
2.0747
0.0397
References [1] Jackson, J. E. A User’s Guide to Principal Components. Hoboken,
NJ: John Wiley and Sons, 1991.
[2] Jolliffe, I. T. Principal Component Analysis. 2nd ed., New York:
Springer-Verlag, 2002.
[3] Krzanowski, W. J. Principles of Multivariate Analysis: A User’s
Perspective. New York: Oxford University Press, 1988.
[4] Seber, G. A. F., Multivariate Observations, Wiley, 1984.
See Also barttest, biplot, factoran, pcares, princomp , rotatefactors
18-1033
pcares
Purpose Residuals from principal component analysis
Syntax residuals = pcares(X,ndim)
[residuals,reconstructed] = pcares(X,ndim)
Description residuals = pcares(X,ndim) returns the residuals obtained by
retaining ndim principal components of the n-by-p matrix X. Rows of X
correspond to observations, columns to variables. ndim is a scalar and
must be less than or equal to p. residuals is a matrix of the same size
as X. Use the data matrix, not the covariance matrix, with this function.
pcares does not normalize the columns of X. To perform the principal
components analysis based on standardized variables, that is, based
on correlations, use pcares(zscore(X), ndim). You can perform
principal components analysis directly on a covariance or correlation
matrix, but without constructing residuals, by using pcacov.
[residuals,reconstructed] = pcares(X,ndim) returns the
reconstructed observations; that is, the approximation to X obtained by
retaining its first ndim principal components.
Examples This example shows the drop in the residuals from the first row of the
Hald data as the number of component dimensions increases from one
to three.
load hald
r1 = pcares(ingredients,1);
r2 = pcares(ingredients,2);
r3 = pcares(ingredients,3);
r11 = r1(1,:)
r11 =
2.0350 2.8304 -6.8378 3.0879
r21 = r2(1,:)
r21 =
-2.4037 2.6930 -1.6482 2.3425
18-1034
pcares
r31 = r3(1,:)
r31 =
0.2008 0.1957 0.2045 0.1921
References [1] Jackson, J. E., A User’s Guide to Principal Components, John Wiley
and Sons, 1991.
[2] Jolliffe, I. T., Principal Component Analysis, 2nd Edition, Springer,
2002.
[3] Krzanowski, W. J. Principles of Multivariate Analysis: A User’s
Perspective. New York: Oxford University Press, 1988.
[4] Seber, G. A. F. Multivariate Observations. Hoboken, NJ: John Wiley
& Sons, Inc., 1984.
See Also factoran, pcacov, princomp
18-1035
gmdistribution.PComponents property
Purpose Input vector of mixing proportions
Description Optional input vector of mixing proportions p, or its default value.
18-1036
pdf
Purpose Probability density functions
Syntax Y = pdf(name,X,A)
Y = pdf(name,X,A,B)
Y = pdf(name,X,A,B,C)
Description Y = pdf(name,X,A) computes the probability density function for the
one-parameter family of distributions specified by name. Parameter
values for the distribution are given in A. Densities are evaluated at
the values in X and returned in Y.
If X and A are arrays, they must be the same size. If X is a scalar, it is
expanded to a constant matrix the same size as A. If A is a scalar, it is
expanded to a constant matrix the same size as X.
Y is the common size of X and A after any necessary scalar expansion.
Y = pdf(name,X,A,B) computes the probability density function for
two-parameter families of distributions, where parameter values are
given in A and B.
If X, A, and B are arrays, they must be the same size. If X is a scalar, it is
expanded to a constant matrix the same size as A and B. If either A or B
are scalars, they are expanded to constant matrices the same size as X.
Y is the common size of X, A, and B after any necessary scalar expansion.
Y = pdf(name,X,A,B,C) computes the probability density function for
three-parameter families of distributions, where parameter values are
given in A, B, and C.
If X, A, B, and C are arrays, they must be the same size. If X is a scalar,
it is expanded to a constant matrix the same size as A, B, and C. If
any of A, B or C are scalars, they are expanded to constant matrices
the same size as X.
Y is the common size of X, A, B and C after any necessary scalar
expansion.
Acceptable strings for name are:
18-1037
pdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'beta' or “Beta a b —
'Beta' Distribution”
on page B-4
'bino' or “Binomial n: number p: —
'Binomial' Distribution” of trials probability
on page B-7 of success
for each
trial
'chi2' or “Chi-Square ν: degrees — —
'Chisquare' Distribution” of freedom
on page
B-12
'exp' or “Exponential : mean — —
Distribution”
'Exponential'
on page
B-16
'ev' or “Extreme : location σ: scale —
'Extreme Value parameter parameter
Value' Distribution”
on page
B-19
'f' or 'F' “F ν 1: ν 2: —
Distribution” numerator denominator
'gam' or on page
“Gamma degrees
a : shapeof degrees
b : scale of —
'Gamma' B-25
Distribution” freedom
parameter freedom
parameter
on page
B-27
18-1038
pdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'gev' or “Generalized K: shape : location σ: scale
'GeneralizedExtreme parameter parameter parameter
Extreme Value
Value' Distribution”
on page
B-32
'gp' or “Generalized k: tail index σ: scale : threshold
'GeneralizedPareto (shape) parameter (location)
Pareto' Distribution” parameter parameter
on page
B-37
'geo' or “Geometric p: — —
'Geometric' Distribution” probability
'hyge' or on page parameter
“Hypergeometric
M : size of the K: number n: number
B-41
Distribution”
'Hypergeometric' population of items of samples
on page with the drawn
B-43 desired
characteristic
in the
population
'logn' or “Lognormal σ —
'Lognormal' Distribution”
on page
B-51
'nbin' or “Negative r: number p: —
'Negative Binomial of successes probability
Binomial' Distribution” of success
on page in a single
B-72 trial
18-1039
pdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'ncf' or “Noncentral ν1: ν 2: δ:
'Noncentral F numerator denominator noncentrality
F' Distribution” degrees of degrees of parameter
on page freedom freedom
B-78
'nct' or “Noncentral ν: degrees δ: —
'Noncentral t of freedom noncentrality
t' Distribution” parameter
'ncx2' or on page
“Noncentral ν: degrees δ: —
'Noncentral B-80
Chi-Square of freedom noncentrality
Chi-square' Distribution” parameter
on page
B-76
'norm' or “Normal : mean σ: standard —
'Normal' Distribution” deviation
on page
B-83
'poiss' or “Poisson λ: mean — —
'Poisson' Distribution”
on page
B-89
'rayl' or “Rayleigh b: scale — —
'Rayleigh' Distribution” parameter
on page
B-91
't' or 'T' “Student’s t ν: degrees — —
Distribution” of freedom
on page
B-95
18-1040
pdf
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'unif' or “Uniform a: lower b: upper —
'Uniform' Distribution endpoint endpoint
(Continuous)” (minimum) (maximum)
on page
B-99
'unid' or “Uniform N: — —
'Discrete Distribution maximum
Uniform' (Discrete)” observable
on page value
B-101
'wbl' or “Weibull a: scale b: shape —
'Weibull' Distribution” parameter parameter
on page
B-103
Examples Compute the pdf of the normal distribution with mean 0 and standard
deviation 1 at inputs –2, –1, 0, 1, 2:
p1 = pdf('Normal',-2:2,0,1)
p1 =
0.0540 0.2420 0.3989 0.2420 0.0540
The order of the parameters is the same as for normpdf.
Compute the pdfs of Poisson distributions with rate parameters 0, 1, ...,
4 at inputs 1, 2, ..., 5, respectively:
p2 = pdf('Poisson',0:4,1:5)
p2 =
0.3679 0.2707 0.2240 0.1954 0.1755
The order of the parameters is the same as for poisspdf.
18-1041
pdf
See Also cdf, icdf, mle, random
18-1042
gmdistribution.pdf
Purpose Probability density function for Gaussian mixture distribution
Syntax y = pdf(obj,X)
Description y = pdf(obj,X) returns a vector y of length n containing the values
of the probability density function (pdf) for the gmdistribution object
obj, evaluated at the n-by-d data matrix X, where n is the number of
observations and d is the dimension of the data. obj is an object created
by gmdistribution or fit. y(I) is the cdf of observation I.
Examples Create a gmdistribution object defining a two-component mixture of
bivariate Gaussian distributions:
MU = [1 2;-3 -5];
SIGMA = cat(3,[2 0;0 .5],[1 0;0 1]);
p = ones(1,2)/2;
obj = gmdistribution(MU,SIGMA,p);
ezsurf(@(x,y)pdf(obj,[x y]),[-10 10],[-10 10])
18-1043
gmdistribution.pdf
See Also gmdistribution, fit, cdf, mvnpdf
18-1044
piecewisedistribution.pdf
Purpose Probability density function for piecewise distribution
Syntax P = pdf(obj,X)
Description P = pdf(obj,X) returns an array P of values of the probability density
function for the piecewise distribution object obj, evaluated at the
values in the array X.
Note For a Pareto tails object, the pdf is computed using the
generalized Pareto distribution in the tails. In the center, the pdf is
computed using the slopes of the cdf, which are interpolated between
a set of discrete values. Therefore the pdf in the center is piecewise
constant. It is noisy for a cdffun specified in paretotails via the
'ecdf' option, and somewhat smoother for the 'kernel' option, but
generally not a good estimate of the underlying density of the original
data.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
[p,q] = boundary(obj)
p =
0.1000
0.9000
q =
-1.7766
1.8432
pdf(obj,q)
ans =
0.2367
0.1960
18-1045
piecewisedistribution.pdf
See Also paretotails, cdf
18-1046
ProbDist.pdf
Purpose Return probability density function (PDF) for ProbDist object
Syntax Y = pdf(PD, X)
Description Y = pdf(PD, X) returns Y, an array containing the probability density
function (PDF) for the ProbDist object PD, evaluated at values in X.
Input PD An object of the class ProbDistUnivParam or
Arguments ProbDistUnivKernel.
X A numeric array of values where you want to
evaluate the PDF.
Output Y An array containing the probability density
Arguments function (PDF) for the ProbDist object PD.
See Also pdf
18-1047
pdist
Purpose Pairwise distance between pairs of objects
Syntax D = pdist(X)
D = pdist(X,distance)
D = pdist(X,'minkowski',P)
D = pdist(X,'mahalanobis',C)
Description D = pdist(X) computes the Euclidean distance between pairs of
objects in m-by-n data matrix X. Rows of X correspond to observations,
and columns correspond to variables. D is a row vector of length
m(m–1)/2, corresponding to pairs of observations in X. The distances
are arranged in the order (2,1), (3,1), ..., (m,1), (3,2), ..., (m,2), ...,
(m,m–1)). D is commonly used as a dissimilarity matrix in clustering or
multidimensional scaling.
To save space and computation time, D is formatted as a vector.
However, you can convert this vector into a square matrix using the
squareform function so that element i, j in the matrix, where i < j,
corresponds to the distance between objects i and j in the original data
set.
D = pdist(X,distance) computes the distance between objects in the
data matrix, X, using the method specified by distance, which can be
any of the following character strings.
Metric Description
'euclidean' Euclidean distance (default).
'seuclidean' Standardized Euclidean distance. Each
coordinate difference between rows in X is
scaled by dividing by the corresponding
element of the standard deviation
S=nanstd(X). To specify another value for
S, use D=pdist(X,'seuclidean',S).
'cityblock' City block metric.
18-1048
pdist
Metric Description
'minkowski' Minkowski distance. The default exponent is
2. To specify a different exponent, use D =
pdist(X,'minkowski',P), where P is a scalar
'chebychev' positive
Chebychevvalue of the (maximum
distance exponent. coordinate
difference).
'mahalanobis' Mahalanobis distance, using the sample
covariance of X as computed by nancov. To
compute the distance with a different covariance,
use D = pdist(X,'mahalanobis',C), where
the matrix C is symmetric and positive definite.
'cosine' One minus the cosine of the included angle
between points (treated as vectors).
'correlation' One minus the sample correlation between
points (treated as sequences of values).
'spearman' One minus the sample Spearman’s rank
correlation between observations (treated as
sequences of values).
'hamming' Hamming distance, which is the percentage of
coordinates that differ.
'jaccard' One minus the Jaccard coefficient, which is the
percentage of nonzero coordinates that differ.
custom distance A distance function specified using @:
function D = pdist(X,@distfun)
A distance function must be of form
d2 = distfun(XI,XJ)
taking as arguments a 1-by-n vector XI,
corresponding to a single row of X, and an
m2-by-n matrix XJ, corresponding to multiple
rows of X. distfun must accept a matrix XJ
18-1049
pdist
Metric Description
with an arbitrary number of rows. distfun
must return an m2-by-1 vector of distances d2,
whose kth element is the distance between XI
and XJ(k,:).
The output D is arranged in the order of ((2,1),(3,1),...,
(m,1),(3,2),...(m,2),.....(m,m–1)), i.e. the lower left triangle of the
full m-by-m distance matrix in column order. To get the distance
between the ith and jth observations (i < j), either use the formula
D((i–1)*(m–i/2)+j–i), or use the helper function Z = squareform(D),
which returns an m-by-m square symmetric matrix, with the (i,j) entry
equal to distance between observation i and observation j.
Metrics
Given an m-by-n data matrix X, which is treated as m (1-by-n) row
vectors x1, x2, ..., xm, the various distances between the vector xs and
xt are defined as follows:
• Euclidean distance
2
dst = ( xs − xt )( xs − xt )′
Notice that the Euclidean distance is a special case of the Minkowski
metric, where p = 2.
• Standardized Euclidean distance
2
dst = ( xs − xt )V −1 ( xs − xt )′
where V is the n-by-n diagonal matrix whose jth diagonal element is
S(j)2, where S is the vector of standard deviations.
• Mahalanobis distance
2
dst = ( xs − xt )C −1 ( xs − xt )′
18-1050
pdist
where C is the covariance matrix.
• City block metric
n
dst = ∑ xsj − xtj
j =1
Notice that the city block distance is a special case of the Minkowski
metric, where p=1.
• Minkowski metric
n p
dst = p ∑ xsj − xtj
j =1
Notice that for the special case of p = 1, the Minkowski metric gives
the city block metric, for the special case of p = 2, the Minkowski
metric gives the Euclidean distance, and for the special case of p = ∞,
the Minkowski metric gives the Chebychev distance.
• Chebychev distance
{
dst = max j xsj − xtj }
Notice that the Chebychev distance is a special case of the Minkowski
metric, where p = ∞.
• Cosine distance
xs xt′
dst = 1 −
( xs xs′ ) ( xt xt′ )
• Correlation distance
18-1051
pdist
( xs − xs ) ( xt − xt )′
dst = 1 −
( xs − xs ) ( xs − xs )′ ( xt − xt ) ( xt − xt )′
where
1 1
xs = ∑ xsj and xt = n ∑ xtj
n j j
• Hamming distance
dst = (#( xsj ≠ xtj ) / n)
• Jaccard distance
( ) ((
# ⎡ xsj ≠ xtj ∩ xsj ≠ 0 ∪ xtj ≠ 0 ⎤ ) ( ))
dst = ⎣ ⎦
( ) (
# ⎡ xsj ≠ 0 ∪ xtj ≠ 0 ⎤
⎣ ⎦ )
• Spearman distance
( rs − rs ) ( rt − rt )′
dst = 1 −
( rs − rs ) ( rs − rs )′ ( rt − rt ) ( rt − rt )′
where
- rsj is the rank of xsj taken over x1j, x2j, ...xmj, as computed by
tiedrank
- rs and rt are the coordinate-wise rank vectors of xs and xt, i.e.,
rs = (rs1, rs2, ... rsn)
1 ( n + 1)
- rs = ∑
n j
rsj =
2
18-1052
pdist
1 ( n + 1)
- rt = ∑
n j
rtj =
2
Examples Generate random data and find the unweighted Euclidean distance and
then find the weighted distance using two different methods:
% Compute the ordinary Euclidean distance.
X = randn(100, 5);
D = pdist(X,'euclidean'); % euclidean distance
% Compute the Euclidean distance with each coordinate
% difference scaled by the standard deviation.
Dstd = pdist(X,'seuclidean');
% Use a function handle to compute a distance that weights
% each coordinate contribution differently.
Wgts = [.1 .3 .3 .2 .1]; % coordinate weights
weuc = @(XI,XJ,W)(sqrt(bsxfun(@minus,XI,XJ).^2 * W'));
Dwgt = pdist(X, @(Xi,Xj) weuc(Xi,Xj,Wgts));
See Also cluster | clusterdata | cmdscale | cophenet | dendrogram |
inconsistent | linkage | pdist2 | silhouette | squareform
18-1053
pdist2
Purpose Pairwise distance between two sets of observations
Syntax D = pdist2(X,Y)
D = pdist2(X,Y,distance)
D = pdist2(X,Y,'minkowski',P)
D = pdist2(X,Y,'mahalanobis',C)
D = pdist2(X,Y,distance,'Smallest',K)
D = pdist2(X,Y,distance,'Largest',K)
[D,I] = pdist2(X,Y,distance,'Smallest',K)
[D,I] = pdist2(X,Y,distance,'Largest',K)
Description D = pdist2(X,Y) returns a matrix D containing the Euclidean
distances between each pair of observations in the mx-by-n data
matrix X and my-by-n data matrix Y. Rows of X and Y correspond to
observations, columns correspond to variables. D is an mx-by-my
matrix, with the (i,j) entry equal to distance between observation i in X
and observation j in Y. The (i,j) entry will be NaN if observation i in X
or observation j in Y contain NaNs.
D = pdist2(X,Y,distance) computes D using distance. Choices are:
Metric Description
'euclidean' Euclidean distance (default).
'seuclidean' Standardized Euclidean distance. Each coordinate
difference between rows in X and Y is scaled
by dividing by the corresponding element
of the standard deviation computed from X,
S=nanstd(X). To specify another value for S, use D =
PDIST2(X,Y,'seuclidean',S).
'cityblock' City block metric.
'minkowski' Minkowski distance. The default exponent is 2. To
compute the distance with a different exponent,
use D = pdist2(X,Y,'minkowski',P), where the
exponent P is a scalar positive value.
18-1054
pdist2
Metric Description
'chebychev' Chebychev distance (maximum coordinate
difference).
'mahalanobis' Mahalanobis distance, using the sample covariance
of X as computed by nancov. To compute the
distance with a different covariance, use D =
pdist2(X,Y,'mahalanobis',C) where the matrix C
is symmetric and positive definite.
'cosine' One minus the cosine of the included angle between
points (treated as vectors).
'correlation' One minus the sample correlation between points
(treated as sequences of values).
'spearman' One minus the sample Spearman’s rank correlation
between observations, treated as sequences of
values.
'hamming' Hamming distance, the percentage of coordinates
that differ.
'jaccard' One minus the Jaccard coefficient, the percentage of
nonzero coordinates that differ.
function A distance function specified using @:
D = pdist2(X,Y,@distfun).
A distance function must be of the form
function D2 = distfun(ZI, ZJ)
taking as arguments a 1-by-n vector ZI containing
a single observation from X or Y, an m2-by-n matrix
ZJ containing multiple observations from X or Y, and
returning an m2-by-1 vector of distances D2, whose
Jth element is the distance between the observations
ZI and ZJ(J,:).
18-1055
pdist2
Metric Description
If your data is not sparse, generally it is faster to use
a built-in distance than to use a function handle.
D = pdist2(X,Y,distance,'Smallest',K) returns a K-by-my matrix
D containing the K smallest pairwise distances to observations in X for
each observation in Y. pdist2 sorts the distances in each column of D in
ascending order. D = pdist2(X,Y,distance,'Largest',K) returns
the K largest pairwise distances sorted in descending order. If K is
greater than mx, pdist2 returns an mx-by-my distance matrix. For
each observation in Y, pdist2 finds the K smallest or largest distances
by computing and comparing the distance values to all the observations
in X.
[D,I] = pdist2(X,Y,distance,'Smallest',K) returns a
K-by-my matrix I containing indices of the observations in X
corresponding to the K smallest pairwise distances in D. [D,I] =
pdist2(X,Y,distance,'Largest',K) returns indices corresponding to
the K largest pairwise distances.
Metrics
Given an mx-by-n data matrix X, which is treated as mx (1-by-n) row
vectors x1, x2, ..., xmx, and my-by-n data matrix Y, which is treated as
my (1-by-n) row vectors y1, y2, ...,ymy, the various distances between the
vector xs and yt are defined as follows:
• Euclidean distance
2
dst = ( xs − yt )( xs − yt )′
Notice that the Euclidean distance is a special case of the Minkowski
metric, where p=2.
• Standardized Euclidean distance
2
dst = ( xs − yt )V −1 ( xs − yt )′
18-1056
pdist2
where V is the n-by-n diagonal matrix whose jth diagonal element is
S(j)2, where S is the vector of standard deviations.
• Mahalanobis distance
2
dst = ( xs − yt )C −1 ( xs − yt )′
where C is the covariance matrix.
• City block metric
n
dst = ∑ xsj − ytj
j =1
Notice that the city block distance is a special case of the Minkowski
metric, where p=1.
• Minkowski metric
n p
dst = p ∑ xsj − ytj
j =1
Notice that for the special case of p = 1, the Minkowski metric gives
the City Block metric, for the special case of p = 2, the Minkowski
metric gives the Euclidean distance, and for the special case of p=∞,
the Minkowski metric gives the Chebychev distance.
• Chebychev distance
{
dst = max j xsj − ytj }
Notice that the Chebychev distance is a special case of the Minkowski
metric, where p=∞.
• Cosine distance
18-1057
pdist2
⎛ xs yt′ ⎞
dst = ⎜ 1 − ⎟
⎜
⎝ ( xs xs′ ) ( yt yt′ ) ⎟⎠
• Correlation distance
( xs − xs ) ( yt − yt )′
dst = 1 −
( xs − xs ) ( xs − xs )′ ( yt − yt ) ( yt − yt )′
where
1
xs = ∑ xsj and
n j
1
yt = ∑ ytj
n j
• Hamming distance
dst = (#( xsj ≠ ytj ) / n)
• Jaccard distance
( ) ((
# ⎡ xsj ≠ ytj ∩ xsj ≠ 0 ∪ ytj ≠ 0 ⎤
⎣ ⎦) ( ))
dst =
⎡ (
# xsj ≠ 0 ∪ ytj ≠ 0
⎣ ) (
⎤
⎦ )
• Spearman distance
( rs − rs ) ( rt − rt )′
dst = 1 −
( rs − rs ) ( rs − rs )′ ( rt − rt ) ( rt − rt )′
where
18-1058
pdist2
- rsj is the rank of xsj taken over x1j, x2j, ...xmx,j, as computed by
tiedrank
- rtj is the rank of ytj taken over y1j, y2j, ...ymy,j, as computed by
tiedrank
- rs and rt are the coordinate-wise rank vectors of xs and yt, i.e. rs =
(rs1, rs2, ... rsn) and rt = (rt1, rt2, ... rtn)
1 ( n + 1)
- rs = ∑
n j
rsj =
2
1 ( n + 1)
- rt = ∑
n j
rtj =
2
Examples Generate random data and find the unweighted Euclidean distance,
then find the weighted distance using two different methods:
% Compute the ordinary Euclidean distance
X = randn(100, 5);
Y = randn(25, 5);
D = pdist2(X,Y,'euclidean'); % euclidean distance
% Compute the Euclidean distance with each coordinate
% difference scaled by the standard deviation
Dstd = pdist2(X,Y,'seuclidean');
% Use a function handle to compute a distance that weights
% each coordinate contribution differently.
Wgts = [.1 .3 .3 .2 .1];
weuc = @(XI,XJ,W)(sqrt(bsxfun(@minus,XI,XJ).^2 * W'));
Dwgt = pdist2(X,Y, @(Xi,Xj) weuc(Xi,Xj,Wgts));
See Also pdist | createns | knnsearch | KDTreeSearcher |
ExhaustiveSearcher
18-1059
pearsrnd
Purpose Pearson system random numbers
Syntax r = pearsrnd(mu,sigma,skew,kurt,m,n)
[r,type] = pearsrnd(...)
[r,type,coefs] = pearsrnd(...)
Description r = pearsrnd(mu,sigma,skew,kurt,m,n) returns an m-by-n matrix of
random numbers drawn from the distribution in the Pearson system
with mean mu, standard deviation sigma, skewness skew, and kurtosis
kurt. mu, sigma, skew, and kurt must be scalars.
Note Because r is a random sample, its sample moments, especially
the skewness and kurtosis, typically differ somewhat from the specified
distribution moments.
Some combinations of moments are not valid for any random variable,
and in particular, the kurtosis must be greater than the square of the
skewness plus 1. The kurtosis of the normal distribution is defined
to be 3.
r = pearsrnd(mu,sigma,skew,kurt) returns a scalar value.
r = pearsrnd(mu,sigma,skew,kurt,m,n,...) or r =
pearsrnd(mu,sigma,skew,kurt,[m,n,...]) returns an m-by-n-by-...
array.
[r,type] = pearsrnd(...) returns the type of the specified
distribution within the Pearson system. type is a scalar integer from
0 to 7. Set m and n to zero to identify the distribution type without
generating any random values.
The seven distribution types in the Pearson system correspond to the
following distributions:
• 0 — Normal distribution
• 1 — Four-parameter beta distribution
18-1060
pearsrnd
• 2 — Symmetric four-parameter beta distribution
• 3 — Three-parameter gamma distribution
• 4 — Not related to any standard distribution. The density is
proportional to:
(1+((x–a)/b)2)–c exp(–d arctan((x–a)/b)).
• 5 — Inverse gamma location-scale distribution
• 6 — F location-scale distribution
• 7 — Student’s t location-scale distribution
[r,type,coefs] = pearsrnd(...) returns the coefficients coefs
of the quadratic polynomial that defines the distribution via the
d(log(p(x))) -(a + x)
differential equation = .
dx (c(0) + c(1) ⋅ x + c(2) ⋅ x 2 )
Examples Generate random values from the standard normal distribution:
r = pearsrnd(0,1,0,3,100,1); % Equivalent to randn(100,1)
Determine the distribution type:
[r,type] = pearsrnd(0,1,1,4,0,0);
r =
[]
type =
1
References [1] Johnson, N.L., S. Kotz, and N. Balakrishnan (1994) Continuous
Univariate Distributions, Volume 1, Wiley-Interscience, Pg 15, Eqn
12.33.
See Also random, johnsrnd
18-1061
perfcurve
Purpose Compute Receiver Operating Characteristic (ROC) curve or other
performance curve for classifier output
Syntax [X,Y] = perfcurve(labels,scores,posclass)
[X,Y] = perfcurve(labels,scores,posclass,'Name', value)
[X,Y,T,AUC,OPTROCPT,SUBY,SUBYNAMES] = perfcurve(labels,scores,
posclass)
[X,Y,T,AUC] = perfcurve(labels,scores,posclass)
Description [X,Y] = perfcurve(labels,scores,posclass) computes a ROC
curve for a vector of classifier predictions scores given true class labels,
labels. labels can be a numeric vector, logical vector, character
matrix, cell array of strings or categorical vector. scores is a numeric
vector of scores returned by a classifier for some data. posclass is
the positive class label (scalar), either numeric (for numeric labels),
logical (for logical labels), or char. The returned values X and Y are
coordinates for the performance curve and can be visualized with
plot(X,Y). For more information on labels, scores, and posclass,
see “Input Arguments” on page 18-1064 . For more information on X
and Y, see “Output Arguments” on page 18-1068.
[X,Y] = perfcurve(labels,scores,posclass,'Name', value)
specifies one or more optional parameter name/value pairs, with Name
in single quotes. See “Input Arguments” on page 18-1064 for a list of
inputs, parameter name/value pairs, and respective explanations.
[X,Y,T,AUC,OPTROCPT,SUBY,SUBYNAMES] =
perfcurve(labels,scores,posclass) returns:
• An array of thresholds on classifier scores for the computed values
of X and Y (T).
• The area under curve (AUC) for the computed values of X and Y.
• The optimal operating point of the ROC curve (OPTROCPT).
• An array of Y values for negative subclasses (SUBY).
• A cell array of negative class names (SUBYNAMES).
18-1062
perfcurve
For more information on each output, see “Output Arguments” on page
18-1068.
[X,Y,T,AUC] = perfcurve(labels,scores,posclass) also returns
pointwise confidence bounds for the computed values X, Y, T, and AUC if
you supply cell arrays for labels and scores or set NBoot (see “Input
Arguments” on page 18-1064 ) to a positive integer. To compute the
confidence bounds, perfcurve uses either vertical averaging (VA) or
threshold averaging (TA). The returned values Y are an m-by-3 array
in which the 1st element in every row gives the mean value, the 2nd
element gives the lower bound and the 3rd element gives the upper
bound. The returned AUC is a row-vector with 3 elements following the
same convention. For VA, the returned values T are an m-by-3 array
and X is a column-vector. For TA, the returned values X are an m-by-3
matrix and T is a column-vector.
perfcurve computes confidence bounds using either cross-validation
or bootstrap. If you supply cell arrays for labels and scores,
perfcurve uses cross-validation and treats elements in the cell arrays
as cross-validation folds. labels can be a cell array of numeric vectors,
logical vectors, character matrices, cell arrays of strings or categorical
vectors. All elements in labels must have the same type. scores is a
cell array of numeric vectors. The cell arrays for labels and scores
must have the same number of elements, and the number of labels in
cell k must be equal to the number of scores in cell k for any k in the
range from 1 to the number of elements in scores.
If you set NBoot to a positive integer, perfcurve generates nboot
bootstrap replicas to compute pointwise confidence bounds. You cannot
supply cell arrays for labels and scores and set NBoot to a positive
integer at the same time.
perfcurve returns pointwise confidence bounds. It does not return a
simultaneous confidence band for the entire curve.
If you use 'XCrit' or 'YCrit' options described below to set the
criterion for X or Y to an anonymous function, perfcurve can only
compute confidence bounds by bootstrap.
18-1063
perfcurve
Input
Arguments labels labels can be a numeric vector, logical vector,
character matrix, cell array of strings or
categorical vector.
scores scores is a numeric vector of scores returned by a
classifier for some data. This vector must have as
many elements as labels does.
posclass posclass is the positive class label (scalar),
either numeric (for numeric labels) or char. The
specified positive class must be in the array of
input labels.
Name/Value Pairs
Name Value and Description
negClass List of negative classes. Can be either a numeric array or an
array of chars or a cell array of strings. By default, negClass
is set to 'all' and all classes found in the input array of labels
that are not the positive class are considered negative. If
negClass is a subset of the classes found in the input array of
labels, instances with labels that do not belong to either positive
or negative classes are discarded.
xCrit Criterion to compute for X. This criterion must be a monotone
function of the positive class score. perfcurve supports the
following criteria:
• TP — Number of true positive instances.
• FN — Number of false negative instances.
• FP — Number of false positive instances.
• TN — Number of true negative instances.
• TP+FP — Sum of TP and FP.
18-1064
perfcurve
Name Value and Description
• RPP — Rate of positive predictions.
RPP=(TP+FP)/(TP+FN+FP+TN)
• RNP — Rate of negative predictions.
RNP=(TN+FN)/(TP+FN+FP+TN)
• accu — Accuracy. accu = (TP+TN)/(TP+FN+FP+TN)
• TPR, sens, reca — True positive rate, sensitivity, recall. TPR,
sens, reca = TP/(TP+FN)
• FNR, miss — False negative rate, miss. FNR,miss=FN/(TP+FN)
• FPR, fall — False positive rate, fallout.
FPR,fall=FP/(TN+FP)
• TNR, spec — True negative rate, specificity.
TNR,spec=TN/(TN+FP)
• PPV, prec — Positive predictive value, precision.
PPV,prec=TP/(TP+FP)
• NPV — Negative predictive value. NPV=TN/(TN+FN)
• ecost — Expected cost.
ecost=(TP*COST(P|P)+FN*COST(N|P)+FP*
COST(P|N)+TN*COST(N|N))/(TP+FN+FP+TN)
In addition, you can define an arbitrary criterion by supplying
an anonymous function of three arguments, (C,scale,cost),
where C is a 2-by-2 confusion matrix, scale is a 2-by-1 array of
class scales, and cost is a 2-by-2 misclassification cost matrix.
Caution Some of these criteria return NaN values at one of the
two special thresholds, 'reject all' and 'accept all'.
18-1065
perfcurve
Name Value and Description
yCrit Criterion to compute for Y. perfcurve supports the same criteria
as for X. This criterion does not have to be a monotone function
of the positive class score.
XVals Values for the X criterion. The default value for xVals is 'all'
and perfcurve computes X and Y values for all scores. If the
value for xVals is not 'all', it must be a numeric array. In this
case, perfcurve computes X and Y only for the specified xVals.
TVals Thresholds for the positive class score. By default, TVals is
unset and perfcurve computes X, Y, and T values for all scores.
You can set TVals to either 'all' or a numeric array. If TVals is
set to 'all' or unset and XVals is unset, perfcurve returns X,
Y, and T values for all scores and computes pointwise confidence
bounds for Y and X using threshold averaging. If TVals is set to
a numeric array, perfcurve returns X, Y, and T values for the
specified thresholds and computes pointwise confidence bounds
for Y and X at these thresholds using threshold averaging. You
cannot set XVals and TVals at the same time.
UseNearest 'on' to use nearest values found in the data instead of the
specified numeric XVals or TVals and 'off' otherwise. If you
specify numeric XVals and set UseNearest to ’on’, perfcurve
returns nearest unique values X found in the data, as well
as corresponding values of Y and T. If you specify numeric
XVals and set UseNearest to 'off', perfcurve returns these
XVals sorted. By default this parameter is set to 'on'. If you
compute confidence bounds by cross-validation or bootstrap, this
parameter is always 'off'.
18-1066
perfcurve
Name Value and Description
ProcessNaN Specifies how perfcurve processes NaN scores. The default value
is 'ignore' and perfcurve removes observations with NaN
scores from the data. If you set the parameter to 'addtofalse',
perfcurve adds instances with NaN scores to false classification
counts in the respective class. That is, perfcurve always counts
instances from the positive class as false negative (FN), and
always counts instances from the negative class as false positive
(FP).
Prior Either string or array with two elements. It represents prior
probabilities for the positive and negative class, respectively.
Default is 'empirical', that is, perfcurve derives prior
probabilities from class frequencies. If set to 'uniform',
perfcurve sets all prior probabilities equal.
Cost A 2-by-2 matrix of misclassification costs [C(P|P) C(N|P);
C(P|N) C(N|N)] where C(I|J) is the cost of misclassifying class
J as class I. By default set to [0 0.5; 0.5 0].
Alpha A numeric value between 0 and 1. perfcurve returns
100*(1-Alpha) percent pointwise confidence bounds for X, Y, T
and AUC. By default set to 0.05 for 95% confidence interval.
Weights A numeric vector of non-negative observation weights. This
vector must have as many elements as scores or labels do. If
you supply cell arrays for scores and labels and you need to
supply weights, you must supply them as a cell array too. In
this case, every element in weights must be a numeric vector
with as many elements as the corresponding element in scores:
numel(weights{1})==numel(scores{1}) etc. To compute X,
Y and T or to compute confidence bounds by cross-validation,
perfcurve uses these observation weights instead of observation
counts. To compute confidence bounds by bootstrap, perfcurve
samples N out of N with replacement using these weights as
multinomial sampling probabilities.
18-1067
perfcurve
Name Value and Description
NBoot Number of bootstrap replicas for computation of confidence
bounds. Must be a positive integer. By default this parameter is
set to zero, and bootstrap confidence bounds are not computed.
If you supply cell arrays for labels and scores, this parameter
must be set to zero because perfcurve cannot use both
cross-validation and bootstrap to compute confidence bounds.
BootType Confidence interval type bootci uses to compute confidence
bounds. You can specify any type supported by bootci. By
default set to 'bca'.
BootArg Optional input arguments for bootci used to compute confidence
bounds. You can specify all arguments bootci supports. Empty
by default.
Output X x-coordinates for the performance curve. By default,
Arguments X is false positive rate, FPR, (equivalently, fallout, or
1–specificity). To change this output, use the 'xCrit'
name/value input. For accepted criterion, see 'xCrit'
in “Input Arguments” on page 18-1064 for more
information.
Y y-coordinates for the performance curve. By default,
Y is true positive rate, TPR, (equivalently, recall, or
sensitivity). To change this output, use the 'yCrit'
input. For accepted criterion, see 'xCrit' in “Input
Arguments” on page 18-1064 for more information.
18-1068
perfcurve
T An array of thresholds on classifier scores for the
computed values of X and Y. It has the same number
of rows as X and Y. For each threshold, TP is the count
of true positive observations with scores greater or
equal to this threshold, and FP is the count of false
positive observations with scores greater or equal to this
threshold. perfcurve defines negative counts, TN and
FN, in a similar way then sorts the thresholds in the
descending order which corresponds to the ascending
order of positive counts.
For the M distinct thresholds found in the array of
scores, perfcurve returns the X, Y and T arrays with M+1
rows. perfcurve sets elements T(2:M+1) to the distinct
thresholds, and T(1) replicates T(2). By convention,
T(1) represents the highest 'reject all' threshold
and perfcurve computes the corresponding values of X
and Y for TP=0 and FP=0. T(end) is the lowest 'accept
all' threshold for which TN=0 and FN=0.
AUC The area under curve (AUC) for the computed values of X
and Y. If you set xVals to 'all' (the default), perfcurve
computes AUC using the returned X and Y values. If
xVals is a numeric array, perfcurve computes AUC
using X and Y values found from all distinct scores in the
interval specified by the smallest and largest elements
of xVals. More precisely, perfcurve finds X values for
all distinct thresholds as if xVals were set to 'all',
then uses a subset of these (with corresponding Y values)
between min(xVals) and max(xVals) to compute AUC.
The function uses trapezoidal approximation to estimate
the area. If the first or last value of X or Y are NaNs,
perfcurve removes them to allow calculation of AUC.
This takes care of criteria that produce NaNs for the
special 'reject all' or 'accept all' thresholds, for
example, positive predictive value (PPV) or negative
predictive value (NPV).
18-1069
perfcurve
OPTROCPT The optimal operating point of the ROC curve as an
array of size 1-by-2 with FPR and TPR values for the
optimal ROC operating point. perfcurve computes
optrocpt only for the standard ROC curve and sets to
NaNs otherwise. To obtain the optimal operating point for
the ROC curve, perfcurve first finds the slope, S, using
cos t( P | N ) − cos t( N | N ) N
S= *
cos t( N | P)is− the
where cost(I|J) cos tcost
(P | P
of) assigning
P an instance
of class J to class I, and P=TP+FN and N=TN+FP are the
total instance counts in the positive and negative class,
respectively. perfcurve then finds the optimal operating
point by moving the straight line with slope S from the
upper left corner of the ROC plot (FPR=0, TPR=1) down
and to the right until it intersects the ROC curve.
SUBY An array of Y values for negative subclasses. If you
only specify one negative class, SUBY is identical to Y.
Otherwise SUBY is a matrix of size M-by-K, where M is
the number of returned values for X and Y, and K is
the number of negative classes. perfcurve computes
Y values by summing counts over all negative classes.
SUBY gives values of the Y criterion for each negative
class separately. For each negative class, perfcurve
places a new column in SUBY and fills it with Y values for
TN and FP counted just for this class.
SUBYNAMES A cell array of negative class names. If you provide
an input array, negClass, of negative class names,
perfcurve copies it into SUBYNAMES. If you do not provide
negClass, perfcurve extracts SUBYNAMES from input
labels. The order of SUBYNAMES is the same as the order
of columns in SUBY, that is, SUBY(:,1) is for negative
class SUBYNAMES{1} etc.
18-1070
perfcurve
Examples Plot the ROC curve for classification by logistic regression:
load fisheriris
x = meas(51:end,1:2);
% iris data, 2 classes and 2 features
y = (1:100)'>50;
% versicolor=0, virginica=1
b = glmfit(x,y,'binomial');
% logistic regression
p = glmval(b,x,'logit');
% fit probabilities for scores
[X,Y,T,AUC] = perfcurve(species(51:end,:),p,'virginica');
plot(X,Y)
xlabel('False positive rate'); ylabel('True positive rate')
title('ROC for classification by logistic regression')
Obtain errors on TPR by vertical averaging
[X,Y] = perfcurve(species(51:end,:),p,'virginica',...
'nboot',1000,'xvals','all');
% plot errors
18-1071
perfcurve
errorbar(X,Y(:,1),Y(:,2)-Y(:,1),Y(:,3)-Y(:,1));
References [1] T. Fawcett, ROC Graphs: Notes and Practical Considerations for
Researchers, 2004.
[2] M. Zweig and G. Campbell, Receiver-Operating Characteristic
(ROC) Plots: A Fundamental Evaluation Tool in Clinical Medicine,
Clin. Chem. 39/4, 561-577, 1993.
[3] J. Davis and M. Goadrich, The relationship between precision-recall
and ROC curves, in Proceedings of ICML ’06, 233-240, 2006.
[4] C. Moskowitz and M. Pepe, Quantifying and comparing the
predictive accuracy of continuous prognostic factors for binary outcomes,
Biostatistics 5, 113-127, 2004.
[5] Y. Huang, M. Pepe and Z. Feng, Evaluating the Predictiveness of
a Continuous Marker, U. Washington Biostatistics Paper Series, 282,
2006.
[6] W. Briggs and R. Zaretzki, The Skill Plot: A Graphical Technique for
Evaluating Continuous Diagnostic Tests, Biometrics 63, 250-261, 2008.
[7] http://www2.cs.uregina.ca/~hamilton/courses/831/notes/lift_chart/lift_chart.html;
http://www.dmreview.com/news/5329-1.html.
[8] R. Bettinger, Cost-Sensitive Classifier Selection Using the ROC
Convex Hull Method, SAS Institute.
[9] http://www.stata.com/statalist/archive/2003-02/msg00060.html
See Also “Performance Curves” on page 12-53, “Plotting a Performance Curve”
on page 12-49,
See Also bootci | glmfit | mnrfit | classify | NaiveBayes | classregtree
18-1072
perfcurve
How To • “Grouped Data” on page 2-34
• “Performance Curves” on page 12-53
• “Plotting a Performance Curve” on page 12-49
18-1073
perms
Purpose Enumeration of permutations
Syntax P = perms(v)
Description P = perms(v) where v is a row vector of length n, creates a matrix
whose rows consist of all possible permutations of the n elements of v.
The matrix P contains n! rows and n columns.
perms is only practical when n is less than 8 or 9.
Examples perms([2 4 6])
ans =
6 4 2
6 2 4
4 6 2
4 2 6
2 4 6
2 6 4
See Also
combnk
18-1074
categorical.permute
Purpose Permute dimensions of categorical array
Syntax B = permute(A,order)
Description B = permute(A,order) rearranges the dimensions of the categorical
array A so that they are in the order specified by the vector order. The
array produced has the same values as A but the order of the subscripts
needed to access any particular element are rearranged as specified by
order. The elements of order must be a rearrangement of the numbers
from 1 to n.
See Also circshift, ipermute
18-1075
piecewisedistribution class
Purpose Piecewise-defined distributions
Construction piecewisedistribution is an abstract class. To construct a
piecewisedistribution object, use the subclass constructor,
paretotails.
Methods boundary Piecewise distribution boundaries
cdf Cumulative distribution function
for piecewise distribution
disp Display piecewisedistribution
object
display Display piecewisedistribution
object
icdf Inverse cumulative distribution
function for piecewise distribution
nsegments Number of segments
pdf Probability density function for
piecewise distribution
random Random numbers from piecewise
distribution
segment Segments containing values
Properties Objects of the piecewisedistribution class have no properties
accessible by dot indexing, get methods, or set methods. To
obtain information about a piecewisedistribution object, use the
appropriate method.
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
18-1076
piecewisedistribution
Purpose Create piecewise distribution object
Description piecewisedistribution is an abstract class, and you cannot create
instances of it directly. You can create paretotails objects that are
derived from this class.
See Also paretotails
18-1077
plsregress
Purpose Partial least-squares regression
Syntax [XL,YL] = plsregress(X,Y,ncomp)
[XL,YL,XS] = plsregress(X,Y,ncomp)
[XL,YL,XS,YS] = plsregress(X,Y,ncomp)
[XL,YL,XS,YS,BETA] = PLSREGRESS(X,Y,ncomp,...)
[XL,YL,XS,YS,BETA,PCTVAR] = plsregress(X,Y,ncomp)
[XL,YL,XS,YS,BETA,PCTVAR,MSE] = plsregress(X,Y,ncomp)
[XL,YL,XS,YS,BETA,PCTVAR,MSE] = plsregress(...,param1,val1,
param2,val2,...)
[XL,YL,XS,YS,BETA,PCTVAR,MSE,stats] = PLSREGRESS(X,Y,ncomp,
...)
Description [XL,YL] = plsregress(X,Y,ncomp) computes a partial least-squares
(PLS) regression of Y on X, using ncomp PLS components, and returns
the predictor and response loadings in XL and YL, respectively. X is
an n-by-p matrix of predictor variables, with rows corresponding to
observations and columns to variables. Y is an n-by-m response matrix.
XL is a p-by-ncomp matrix of predictor loadings, where each row contains
coefficients that define a linear combination of PLS components that
approximate the original predictor variables. YL is an m-by-ncomp
matrix of response loadings, where each row contains coefficients that
define a linear combination of PLS components that approximate the
original response variables.
[XL,YL,XS] = plsregress(X,Y,ncomp) returns the predictor scores
XS, that is, the PLS components that are linear combinations of the
variables in X. XS is an n-by-ncomp orthonormal matrix with rows
corresponding to observations and columns to components.
[XL,YL,XS,YS] = plsregress(X,Y,ncomp) returns the response
scores YS, that is, the linear combinations of the responses with which
the PLS components XS have maximum covariance. YS is an n-by-ncomp
matrix with rows corresponding to observations and columns to
components. YS is neither orthogonal nor normalized.
plsregress uses the SIMPLS algorithm, first centering X and Y by
subtracting off column means to get centered variables X0 and Y0.
18-1078
plsregress
However, it does not rescale the columns. To perform PLS with
standardized variables, use zscore to normalize X and Y.
If ncomp is omitted, its default value is min(size(X,1)-1,size(X,2)).
The relationships between the scores, loadings, and centered variables
X0 and Y0 are:
XL = (XS\X0)' = X0'*XS,
YL = (XS\Y0)' = Y0'*XS,
XL and YL are the coefficients from regressing X0 and Y0 on XS, and
XS*XL' and XS*YL' are the PLS approximations to X0 and Y0.
plsregress initially computes YS as:
YS = Y0*YL = Y0*Y0'*XS,
By convention, however, plsregress then orthogonalizes each column
of YS with respect to preceding columns of XS, so that XS'*YS is lower
triangular.
[XL,YL,XS,YS,BETA] = PLSREGRESS(X,Y,ncomp,...) returns
the PLS regression coefficients BETA. BETA is a (p+1)-by-m matrix,
containing intercept terms in the first row:
Y = [ones(n,1),X]*BETA + RESIDUALS,
Y0 = X0*BETA(2:end,:) + RESIDUALS.
[XL,YL,XS,YS,BETA,PCTVAR] = plsregress(X,Y,ncomp) returns
a 2-by-ncomp matrix PCTVAR containing the percentage of variance
explained by the model. The first row of PCTVAR contains the percentage
of variance explained in X by each PLS component, and the second row
contains the percentage of variance explained in Y.
[XL,YL,XS,YS,BETA,PCTVAR,MSE] = plsregress(X,Y,ncomp) returns
a 2-by-(ncomp+1) matrix MSE containing estimated mean-squared errors
for PLS models with 0:ncomp components. The first row of MSE contains
mean-squared errors for the predictor variables in X, and the second
row contains mean-squared errors for the response variable(s) in Y.
18-1079
plsregress
[XL,YL,XS,YS,BETA,PCTVAR,MSE] =
plsregress(...,param1,val1,param2,val2,...) specifies optional
parameter name/value pairs from the following table to control the
calculation of MSE.
ParameterValue
'cv' The method used to compute MSE.
• When the value is a positive integer k, plsregress
uses k-fold cross-validation.
• When the value is an object of the cvpartition class,
other forms of cross-validation can be specified.
• When the value is 'resubstitution', plsregress
uses X and Y both to fit the model and to estimate the
mean-squared errors, without cross-validation.
The default is 'resubstitution'.
'mcreps' A positive integer indicating the number of Monte-Carlo
repetitions for cross-validation. The default value
is 1. The value must be 1 if the value of 'cv' is
'resubstitution'.
[XL,YL,XS,YS,BETA,PCTVAR,MSE,stats] =
PLSREGRESS(X,Y,ncomp,...) returns a structure stats with the
following fields:
• W — A p-by-ncomp matrix of PLS weights so that XS = X0*W.
• T2 — The T2 statistic for each point in XS.
• Xresiduals — The predictor residuals, that is, X0-XS*XL'.
• Yresiduals — The response residuals, that is, Y0-XS*YL'.
Examples Load data on near infrared (NIR) spectral intensities of 60 samples of
gasoline at 401 wavelengths, and their octane ratings:
18-1080
plsregress
load spectra
X = NIR;
y = octane;
Perform PLS regression with ten components:
[XL,yl,XS,YS,beta,PCTVAR] = plsregress(X,y,10);
Plot the percent of variance explained in the response variable as a
function of the number of components:
plot(1:10,cumsum(100*PCTVAR(2,:)),'-bo');
xlabel('Number of PLS components');
ylabel('Percent Variance Explained in y');
Compute the fitted response and display the residuals:
18-1081
plsregress
yfit = [ones(size(X,1),1) X]*beta;
residuals = y-yfit;
stem(residuals)
xlabel('Observation');
ylabel('Residual');
References [1] de Jong, S. “SIMPLS: An Alternative Approach to Partial Least
Squares Regression.” Chemometrics and Intelligent Laboratory Systems.
Vol. 18, 1993, pp. 251–263.
[2] Rosipal, R., and N. Kramer. “Overview and Recent Advances
in Partial Least Squares.” Subspace, Latent Structure and Feature
Selection: Statistical and Optimization Perspectives Workshop (SLSFS
18-1082
plsregress
2005), Revised Selected Papers (Lecture Notes in Computer Science
3940). Berlin, Germany: Springer-Verlag, 2006, pp. 34–51.
See Also regress, sequentialfs
18-1083
sobolset.PointOrder property
Purpose Point generation method
Description The PointOrder property contains a string that specifies the order in
which the Sobol sequence points are produced. The property value must
be one of 'standard' or 'graycode'. When set to 'standard' the points
produced match the original Sobol sequence implementation. When set
to 'graycode', the sequence is generated using an implementation that
uses the Gray code of the index instead of the index itself.
18-1084
qrandstream.PointSet property
Purpose Point set from which stream is drawn
Description The PointSet property contains a copy of the point set from which the
stream is providing points. The point set is specified during construction
of a quasi-random stream and cannot subsequently be altered.
Examples Q = qrandstream('sobol', 5, 'Skip', 8);
% Create a new stream based on the same sequence as that in Q
Q2 = qrandstream(Q.PointSet);
u1 = qrand(Q, 10)
u2 = qrand(Q2, 10) % contains exactly the same values as u1
18-1085
poisscdf
Purpose Poisson cumulative distribution function
Syntax P = poisscdf(X,lambda)
Description P = poisscdf(X,lambda) computes the Poisson cdf at each of the
values in X using the corresponding mean parameters in lambda. X and
lambda can be vectors, matrices, or multidimensional arrays that have
the same size. A scalar input is expanded to a constant array with the
same dimensions as the other input. The parameters in lambda must
be positive.
The Poisson cdf is
floor ( x)
i
p = F ( x | ) = e− ∑
i =0
i!
Examples For example, consider a Quality Assurance department that performs
random tests of individual hard disks. Their policy is to shut down the
manufacturing process if an inspector finds more than four bad sectors
on a disk. What is the probability of shutting down the process if the
mean number of bad sectors (λ) is two?
probability = 1-poisscdf(4,2)
probability =
0.0527
About 5% of the time, a normally functioning manufacturing process
produces more than four flaws on a hard disk.
Suppose the average number of flaws (λ) increases to four. What is the
probability of finding fewer than five flaws on a hard drive?
probability = poisscdf(4,4)
probability =
0.6288
18-1086
poisscdf
This means that this faulty manufacturing process continues to operate
after this first inspection almost 63% of the time.
See Also cdf, poisspdf, poissinv, poisstat, poissfit, poissrnd
“Poisson Distribution” on page B-89
18-1087
poissfit
Purpose Poisson parameter estimates
Syntax lambdahat = poissfit(data)
[lambdahat,lambdaci] = poissfit(data)
[lambdahat,lambdaci] = poissfit(data,alpha)
Description lambdahat = poissfit(data) returns the maximum likelihood
estimate (MLE) of the parameter of the Poisson distribution, λ, given
the data data.
[lambdahat,lambdaci] = poissfit(data) also gives 95% confidence
intervals in lamdaci.
[lambdahat,lambdaci] = poissfit(data,alpha) gives
100(1 - alpha)% confidence intervals. For example alpha = 0.001
yields 99.9% confidence intervals.
The sample mean is the MLE of λ.
1 n
̂ = ∑ xi
n i=1
Examples r = poissrnd(5,10,2);
[l,lci] = poissfit(r)
l =
7.4000 6.3000
lci =
5.8000 4.8000
9.1000 7.9000
See Also mle, poisspdf, poisscdf, poissinv, poisstat, poissrnd
“Poisson Distribution” on page B-89
18-1088
poissinv
Purpose Poisson inverse cumulative distribution function
Syntax X = poissinv(P,lambda)
Description X = poissinv(P,lambda) returns the smallest value X such that the
Poisson cdf evaluated at X equals or exceeds P, using mean parameters
in lambda. P and lambda can be vectors, matrices, or multidimensional
arrays that all have the same size. A scalar input is expanded to a
constant array with the same dimensions as the other input.
Examples If the average number of defects (λ) is two, what is the 95th percentile
of the number of defects?
poissinv(0.95,2)
ans =
5
What is the median number of defects?
median_defects = poissinv(0.50,2)
median_defects =
2
See Also icdf, poisscdf, poisspdf, poisstat, poissfit, poissrnd
“Poisson Distribution” on page B-89
18-1089
poisspdf
Purpose Poisson probability density function
Syntax Y = poisspdf(X,lambda)
Description Y = poisspdf(X,lambda) computes the Poisson pdf at each of the
values in X using mean parameters in lambda. X and lambda can be
vectors, matrices, or multidimensional arrays that all have the same
size. A scalar input is expanded to a constant array with the same
dimensions as the other input. The parameters in lambda must all
be positive.
The Poisson pdf is
x −
y = f ( x|) = e I(0,1,...) ( x)
x!
where x can be any nonnegative integer. The density function is zero
unless x is an integer.
Examples A computer hard disk manufacturer has observed that flaws occur
randomly in the manufacturing process at the average rate of two flaws
in a 4 GB hard disk and has found this rate to be acceptable. What is
the probability that a disk will be manufactured with no defects?
In this problem, λ = 2 and x = 0.
p = poisspdf(0,2)
p =
0.1353
See Also pdf, poisscdf, poissinv, poisstat, poissfit, poissrnd
“Poisson Distribution” on page B-89
18-1090
poissrnd
Purpose Poisson random numbers
Syntax R = poissrnd(lambda)
R = poissrnd(lambda,m)
R = poissrnd(lambda,m,n)
Description R = poissrnd(lambda) generates random numbers from the Poisson
distribution with mean parameter lambda. lambda can be a vector, a
matrix, or a multidimensional array. The size of R is the size of lambda.
R = poissrnd(lambda,m) generates random numbers from the Poisson
distribution with mean parameter lambda, where m is a row vector. If m
is a 1-by-2 vector, R is a matrix with m(1) rows and m(2) columns. If m
is 1-by-n, R is an n-dimensional array.
R = poissrnd(lambda,m,n) generates random numbers from the
Poisson distribution with mean parameter lambda, where scalars m and
n are the row and column dimensions of R.
Examples Generate a random sample of 10 pseudo-observations from a Poisson
distribution with λ = 2.
lambda = 2;
random_sample1 = poissrnd(lambda,1,10)
random_sample1 =
1 0 1 2 1 3 4 2 0 0
random_sample2 = poissrnd(lambda,[1 10])
random_sample2 =
1 1 1 5 0 3 2 2 3 4
random_sample3 = poissrnd(lambda(ones(1,10)))
random_sample3 =
3 2 1 1 0 0 4 0 2 0
See Also random, poisspdf, poisscdf, poissinv, poisstat, poissfit
18-1091
poissrnd
“Poisson Distribution” on page B-89
18-1092
poisstat
Purpose Poisson mean and variance
Syntax M = poisstat(lambda)
[M,V] = poisstat(lambda)
Description M = poisstat(lambda) returns the mean of the Poisson distribution
using mean parameters in lambda. The size of M is the size of lambda.
[M,V] = poisstat(lambda) also returns the variance V of the Poisson
distribution.
For the Poisson distribution with parameter λ, both the mean and
variance are equal to λ.
Examples Find the mean and variance for the Poisson distribution with λ = 2.
[m,v] = poisstat([1 2; 3 4])
m =
1 2
3 4
v =
1 2
3 4
See Also poisspdf, poisscdf, poissinv, poissfit, poissrnd
“Poisson Distribution” on page B-89
18-1093
polyconf
Purpose Polynomial confidence intervals
Syntax Y = polyconf(p,X)
[Y,DELTA] = polyconf(p,X,S)
[Y,DELTA] = polyconf(p,X,S,param1,val1,param2,val2,...)
Description Y = polyconf(p,X) evaluates the polynomial p at the values in X. p is
a vector of coefficients in descending powers.
[Y,DELTA] = polyconf(p,X,S) takes outputs p and S from polyfit
and generates 95% prediction intervals Y – DELTA for new observations
at the values in X.
[Y,DELTA] = polyconf(p,X,S,param1,val1,param2,val2,...)
specifies optional parameter name/value pairs chosen from the following
list.
Parameter Value
'alpha' A value between 0 and 1 specifying a confidence level
of 100*(1-alpha)%. The default is 0.05.
'mu' A two-element vector containing centering and
scaling parameters. With this option, polyconf uses
(X-mu(1))/mu(2) in place of X.
'predopt' Either 'observation' (the default) to compute
prediction intervals for new observations at the
values in X, or 'curve' to compute confidence
intervals for the fit evaluated at the values in X. See
below.
'simopt' Either 'off' (the default) for nonsimultaneous
bounds, or 'on' for simultaneous bounds. See below.
The 'predopt' and 'simopt' parameters can be understood in terms
of the following functions:
• p(x) — the unknown mean function estimated by the fit
18-1094
polyconf
• l(x) — the lower confidence bound
• u(x) — the upper confidence bound
Suppose you make a new observation yn+1 at xn+1, so that
yn+1(xn+1) = p(xn+1) + εn+1
By default, the interval [ln+1(xn+1), un+1(xn+1)] is a 95% confidence bound
on yn+1(xn+1).
The following combinations of the 'predopt' and 'simopt' parameters
allow you to specify other bounds.
simopt predopt Bounded Quantity
'off' 'observation' yn+1(xn+1) (default)
'off' 'curve' p(xn+1)
'on' 'observation' yn+1(x), for all x
'on' 'curve' p(x), for all x
In general, 'observation' intervals are wider than 'curve' intervals,
because of the additional uncertainty of predicting a new response
value (the curve plus random errors). Likewise, simultaneous intervals
are wider than nonsimultaneous intervals, because of the additional
uncertainty of bounding values for all predictors x.
18-1095
polyconf
Examples This example uses code from the documentation example function
polydemo, and calls the documentation example function polystr
to convert the coefficient vector p into a string for the polynomial
expression displayed in the figure title. It combines the functions
polyfit, polyval, roots, and polyconf to produce a formatted display
of data with a polynomial fit.
18-1096
polyconf
Note Statistics Toolbox documentation example files are located in the
\help\toolbox\stats\examples subdirectory of your MATLAB root
folder (matlabroot). This subdirectory is not on the MATLAB path
at installation. To use the files in this subdirectory, either add the
subdirectory to the MATLAB path (addpath) or make the subdirectory
your current working folder (cd).
Display simulated data with a quadratic trend, a fitted quadratic
polynomial, and 95% prediction intervals for new observations:
xdata = -5:5;
ydata = x.^2 - 5*x - 3 + 5*randn(size(x));
degree = 2; % Degree of the fit
alpha = 0.05; % Significance level
% Compute the fit and return the structure used by
% POLYCONF.
[p,S] = polyfit(xdata,ydata,degree);
% Compute the real roots and determine the extent of the
% data.
r = roots(p)'; % Roots as a row vector.
real_r = r(imag(r) == 0); % Real roots.
% Assure that the data are row vectors.
xdata = reshape(xdata,1,length(xdata));
ydata = reshape(ydata,1,length(ydata));
% Extent of the data.
mx = min([real_r,xdata]);
Mx = max([real_r,xdata]);
my = min([ydata,0]);
My = max([ydata,0]);
18-1097
polyconf
% Scale factors for plotting.
sx = 0.05*(Mx-mx);
sy = 0.05*(My-my);
% Plot the data, the fit, and the roots.
hdata = plot(xdata,ydata,'md','MarkerSize',5,...
'LineWidth',2);
hold on
xfit = mx-sx:0.01:Mx+sx;
yfit = polyval(p,xfit);
hfit = plot(xfit,yfit,'b-','LineWidth',2);
hroots = plot(real_r,zeros(size(real_r)),...
'bo','MarkerSize',5,...
'LineWidth',2,...
'MarkerFaceColor','b');
grid on
plot(xfit,zeros(size(xfit)),'k-','LineWidth',2)
axis([mx-sx Mx+sx my-sy My+sy])
% Add prediction intervals to the plot.
[Y,DELTA] = polyconf(p,xfit,S,'alpha',alpha);
hconf = plot(xfit,Y+DELTA,'b--');
plot(xfit,Y-DELTA,'b--')
% Display the polynomial fit and the real roots.
approx_p = round(100*p)/100; % Round for display.
htitle = title(['{\bf Fit: }',...
texlabel(polystr(approx_p))]);
set(htitle,'Color','b')
approx_real_r = round(100*real_r)/100; % Round for display.
hxlabel = xlabel(['{\bf Real Roots: }',...
num2str(approx_real_r)]);
set(hxlabel,'Color','b')
% Add a legend.
legend([hdata,hfit,hroots,hconf],...
'Data','Fit','Real Roots of Fit',...
18-1098
polyconf
'95% Prediction Intervals')
See Also polyfit, polyval, polytool
18-1099
polytool
Purpose Interactive polynomial fitting
Syntax polytool
polytool(x,y)
polytool(x,y,n)
polytool(x,y,n,alpha)
polytool(x,y,n,alpha,xname,yname)
h = polytool(...)
Description polytool
polytool(x,y) fits a line to the vectors x and y and displays an
interactive plot of the result in a graphical interface. You can use the
interface to explore the effects of changing the parameters of the fit and
to export fit results to the workspace.
polytool(x,y,n) initially fits a polynomial of degree n. The default
is 1, which produces a linear fit.
polytool(x,y,n,alpha) initially plots 100(1 - alpha)% confidence
intervals on the predicted values. The default is 0.05 which results in
95% confidence intervals.
polytool(x,y,n,alpha,xname,yname) labels the x and y values on the
graphical interface using the strings xname and yname. Specify n and
alpha as [] to use their default values.
h = polytool(...) outputs a vector of handles, h, to the line objects
in the plot. The handles are returned in the degree: data, fit, lower
bounds, upper bounds.
See Also polyfit, polyval, polyconf, invpred
18-1100
gmdistribution.posterior
Purpose Posterior probabilities of components
Syntax P = posterior(obj,X)
[P,nlogl] = posterior(obj,X)
Description P = posterior(obj,X) returns the posterior probabilities of each of
the k components in the Gaussian mixture distribution defined by obj
for each observation in the data matrix X. X is n-by-d, where n is the
number of observations and d is the dimension of the data. obj is an
object created by gmdistribution or fit. P is n-by-k, with P(I,J) the
probability of component J given observation I.
posterior treats NaN values as missing data. Rows of X with NaN values
are excluded from the computation.
[P,nlogl] = posterior(obj,X) also returns nlogl, the negative
log-likelihood of the data.
Examples Generate data from a mixture of two bivariate Gaussian distributions
using the mvnrnd function:
MU1 = [1 2];
SIGMA1 = [2 0; 0 .5];
MU2 = [-3 -5];
SIGMA2 = [1 0; 0 1];
X = [mvnrnd(MU1,SIGMA1,1000);mvnrnd(MU2,SIGMA2,1000)];
scatter(X(:,1),X(:,2),10,'.')
hold on
18-1101
gmdistribution.posterior
Fit a two-component Gaussian mixture model:
obj = gmdistribution.fit(X,2);
h = ezcontour(@(x,y)pdf(obj,[x y]),[-8 6],[-8 6]);
18-1102
gmdistribution.posterior
Compute posterior probabilities of the components:
P = posterior(obj,X);
delete(h)
scatter(X(:,1),X(:,2),10,P(:,1),'.')
hb = colorbar;
ylabel(hb,'Component 1 Probability')
18-1103
gmdistribution.posterior
See Also gmdistribution, fit, cluster, mahal
18-1104
NaiveBayes.posterior
Purpose Compute posterior probability of each class for test data
Syntax post = posterior(nb,test)
[post,cpre] = posterior(nb,test)
[post,cpre,logp] = posterior(nb,test)
[...] = posterior(..., 'HandleMissing',val)
Description post = posterior(nb,test) returns the posterior probability of the
observations in test according to the NaiveBayes object nb. test is
a N-by-nb.ndims matrix, where N is the number of observations in
the test data. Rows of test correspond to points, columns of test
correspond to features. post is a N-by-nb.nclasses matrix containing
the posterior probability of each observation for each class. post(i,j)
is the posterior probability of point I belonging to class j. Classes are
ordered the same as nb.clevels, i.e., column j of post corresponds to
the jth class in nb.clevels. The posterior probabilities corresponding
to any empty classes are NaN.
[post,cpre] = posterior(nb,test) returns cpre, an N-by-1 vector,
containing the class to which each row of test has been assigned. cpre
has the same type as nb.CLevels.
[post,cpre,logp] = posterior(nb,test) returns logp, an N-by-1
vector containing estimates of the log of the probability density function
(PDF). logp(i) is the log of the PDF of point i. The PDF value of point
i is the sum of Prob(point I | class J) * Pr{class J} taken over
all classes.
[...] = posterior(..., 'HandleMissing',val) specifies how
posterior treats NaN (missing values). val can be one of the following:
18-1105
NaiveBayes.posterior
'off' Observations with NaN in any of the columns are not
(default) classified into any class. The corresponding rows in
post and logp are NaN. The corresponding rows in
cpre are NaN (if obj.clevels is numeric or logical),
empty strings (if obj.clevels is char or cell array of
strings) or (if obj.clevels is categorical).
'on' For observations having NaN in some (but not all)
columns, post and cpre are computed using the
columns with non-NaN values. Corresponding logp
values are NaN.
See Also NaiveBayes, fit, predict
18-1106
prctile
Purpose Calculate percentile values
Syntax Y = prctile(X,p)
Y = prctile(X,p,dim)
Description Y = prctile(X,p) returns percentiles of the values in X. p is a scalar
or a vector of percent values. When X is a vector, Y is the same size as
p and Y(i) contains the p(i)th percentile. When X is a matrix, the
ith row of Y contains the p(i)th percentiles of each column of X. For
N-dimensional arrays, prctile operates along the first nonsingleton
dimension of X.
Y = prctile(X,p,dim) calculates percentiles along dimension dim.
The dim’th dimension of Y has length length(p).
Percentiles are specified using percentages, from 0 to 100. For an
n-element vector X, prctile computes percentiles as follows:
1 The sorted values in X are taken to be the 100(0.5/n), 100(1.5/n), ...,
100([n-0.5]/n) percentiles.
2 Linear interpolation is used to compute percentiles for percent values
between 100(0.5/n) and 100([n-0.5]/n).
3 The minimum or maximum values in X are assigned to percentiles
for percent values outside that range.
prctile treats NaNs as missing values and removes them.
Examples x = (1:5)'*(1:5)
x =
1 2 3 4 5
2 4 6 8 10
3 6 9 12 15
4 8 12 16 20
5 10 15 20 25
y = prctile(x,[25 50 75])
18-1107
prctile
y =
1.7500 3.5000 5.2500 7.0000 8.7500
3.0000 6.0000 9.0000 12.0000 15.0000
4.2500 8.5000 12.7500 17.0000 21.2500
18-1108
CompactTreeBagger.predict
Purpose Predict response
Syntax YFIT = predict(B,X)
[YFIT,stdevs] = predict(B,X)
[YFIT,scores] = predict(B,X)
[YFIT,scores,stdevs] = predict(B,X)
Y = predict(B,X,'param1',val1,'param2',val2,...)
Description YFIT = predict(B,X) computes the predicted response of the trained
ensemble B for predictors X. By default, predict takes a democratic
(nonweighted) average vote from all trees in the ensemble. In X, rows
represent observations and columns represent variables. YFIT is a cell
array of strings for classification and a numeric array for regression.
For regression, [YFIT,stdevs] = predict(B,X) also returns standard
deviations of the computed responses over the ensemble of the grown
trees.
For classification, [YFIT,scores] = predict(B,X) returns scores
for all classes. scores is a matrix with one row per observation and
one column per class. For each observation and each class, the score
generated by each tree is the probability of this observation originating
from this class computed as the fraction of observations of this class in a
tree leaf. predict averages these scores over all trees in the ensemble.
[YFIT,scores,stdevs] = predict(B,X)also returns standard
deviations of the computed scores for classification. stdevs is a matrix
with one row per observation and one column per class, with standard
deviations taken over the ensemble of the grown trees.
Y = predict(B,X,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
18-1109
CompactTreeBagger.predict
'trees' Array of tree indices to use for computation of
responses. Default is 'all'.
'treeweights' Array of NTrees weights for weighting votes from
the specified trees.
'useifort' Logical matrix of size Nobs-by-NTrees indicating
which trees to use to make predictions for each
observation. By default all trees are used for all
observations.
See Also classregtree.eval, TreeBagger.predict
18-1110
NaiveBayes.predict
Purpose Predict class label for test data
Syntax cpre = predict(nb,test)
cpre = predict(...,'HandleMissing',val)
Description cpre = predict(nb,test) classifies each row of data in test into one
of the classes according to the NaiveBayes classifier nb, and returns the
predicted class level cpre. test is an N-by-nb.ndims matrix, where N is
the number of observations in the test data. Rows of test correspond to
points, columns of test correspond to features. cpre is an N-by-1 vector
of the same type as nb.CLevels, and it indicates the class to which
each row of test has been assigned.
cpre = predict(...,'HandleMissing',val) specifies how predict
treats NaN (missing values). val can be one of the following:
'off' Observations with NaN in any of the columns are not
(default) classified into any class. The corresponding rows in
post and logp are NaN. The corresponding rows in
cpre are NaN (if obj.clevels is numeric or logical),
empty strings (if obj.clevels is char or cell array of
strings) or (if obj.clevels is categorical).
'on' For observations having NaN in some (but not all)
columns, post and predict computes cpre using the
columns with non-NaN values. Corresponding logp
values are NaN.
See Also NaiveBayes, fit, posterior
18-1111
TreeBagger.predict
Purpose Predict response
Syntax Y = predict(B,X)
[Y,stdevs] = predict(B,X)
[Y,scores] = predict(B,X)
[Y,scores,stdevs] = predict(B,X)
Y = predict(B,X,'param1',val1,'param2',val2,...)
Description Y = predict(B,X) computes predicted response of the trained
ensemble B for data X. The output has one prediction for each row of X.
The returned Y is a cell array of strings for classification and a numeric
array for regression.
For regression, [Y,stdevs] = predict(B,X) also returns standard
deviations of the computed responses over the ensemble of the grown
trees.
For classification, [Y,scores] = predict(B,X) returns scores for
all classes. scores is a matrix with one row per observation and
one column per class. For each observation and each class, the score
generated by each tree is the probability of this observation originating
from this class computed as the fraction of observations of this class in a
tree leaf. predict averages these scores over all trees in the ensemble.
[Y,scores,stdevs] = predict(B,X)also returns standard deviations
of the computed scores for classification. stdevs is a matrix with one
row per observation and one column per class, with standard deviations
taken over the ensemble of the grown trees.
Y = predict(B,X,'param1',val1,'param2',val2,...) specifies
optional parameter name/value pairs:
18-1112
TreeBagger.predict
'trees' Array of tree indices to use for computation of
responses. Default is 'all'.
'treeweights' Array of NTrees weights for weighting votes from
the specified trees.
'useifort' Logical matrix of size Nobs-by-NTrees indicating
which trees to use to make predictions for each
observation. By default all trees are used for all
observations.
See Also CompactTreeBagger.predict
18-1113
princomp
Purpose Principal component analysis on data
Syntax [COEFF,SCORE] = princomp(X)
[COEFF,SCORE,latent] = princomp(X)
[COEFF,SCORE,latent,tsquare] = princomp(X)
[...] = princomp(X,'econ')
Description COEFF = princomp(X) performs principal components analysis on the
n-by-p data matrix X, and returns the principal component coefficients,
also known as loadings. Rows of X correspond to observations, columns
to variables. COEFF is a p-by-p matrix, each column containing
coefficients for one principal component. The columns are in order of
decreasing component variance.
princomp centers X by subtracting off column means, but does not
rescale the columns of X. To perform principal components analysis
with standardized variables, that is, based on correlations, use
princomp(zscore(X)). To perform principal components analysis
directly on a covariance or correlation matrix, use pcacov.
[COEFF,SCORE] = princomp(X) returns SCORE, the principal
component scores; that is, the representation of X in the principal
component space. Rows of SCORE correspond to observations, columns
to components.
[COEFF,SCORE,latent] = princomp(X) returns latent, a vector
containing the eigenvalues of the covariance matrix of X.
[COEFF,SCORE,latent,tsquare] = princomp(X) returns tsquare,
which contains Hotelling’s T2 statistic for each data point.
The scores are the data formed by transforming the original data into
the space of the principal components. The values of the vector latent
are the variance of the columns of SCORE. Hotelling’s T2 is a measure
of the multivariate distance of each observation from the center of the
data set.
When n <= p, SCORE(:,n:p) and latent(n:p) are necessarily zero,
and the columns of COEFF(:,n:p) define directions that are orthogonal
to X.
18-1114
princomp
[...] = princomp(X,'econ') returns only the elements of latent
that are not necessarily zero, and the corresponding columns of COEFF
and SCORE, that is, when n <= p, only the first n-1. This can be
significantly faster when p is much larger than n.
Examples Compute principal components for the ingredients data in the Hald
data set, and the variance accounted for by each component.
load hald;
[pc,score,latent,tsquare] = princomp(ingredients);
pc,latent
pc =
0.0678 -0.6460 0.5673 -0.5062
0.6785 -0.0200 -0.5440 -0.4933
-0.0290 0.7553 0.4036 -0.5156
-0.7309 -0.1085 -0.4684 -0.4844
latent =
517.7969
67.4964
12.4054
0.2372
The following command and plot show that two components account for
98% of the variance:
cumsum(latent)./sum(latent)
ans =
0.86597
0.97886
0.9996
1
biplot(pc(:,1:2),'Scores',score(:,1:2),'VarLabels',...
{'X1' 'X2' 'X3' 'X4'})
18-1115
princomp
For a more detailed example and explanation of this analysis method,
see “Principal Component Analysis” on page 10-31.
References [1] Jackson, J. E., A User’s Guide to Principal Components, John Wiley
and Sons, 1991, p. 592.
[2] Jolliffe, I. T., Principal Component Analysis, 2nd edition, Springer,
2002.
[3] Krzanowski, W. J. Principles of Multivariate Analysis: A User’s
Perspective. New York: Oxford University Press, 1988.
18-1116
princomp
[4] Seber, G. A. F., Multivariate Observations, Wiley, 1984.
See Also “Principal Component Analysis” on page 10-31
barttest, biplot, canoncorr, factoran, pcacov, pcares ,
rotatefactors
18-1117
TreeBagger.Prior property
Purpose Prior class probabilities
Description The Prior property is a vector with prior probabilities for classes. This
property is empty for ensembles of regression trees.
See Also classregtree
18-1118
ProbDist class
Purpose Object representing probability distribution
Description ProbDist is an abstract class representing a probability distribution.
Construction ProbDist is an abstract class. You cannot create instances of this
class directly. You can construct an object in a subclass, such as
ProbDistUnivParam or ProbDistUnivKernel, either by calling the
subclass constructors (ProbDistUnivParam.ProbDistUnivParam or
ProbDistUnivKernel.ProbDistUnivKernel) or by using the fitdist
function.
Methods cdf Return cumulative distribution
function (CDF) for ProbDist object
pdf Return probability density
function (PDF) for ProbDist object
random Generate random number drawn
from ProbDist object
Properties DistName Read-only string containing
probability distribution name of
ProbDist object
InputData Read-only structure containing
information about input data to
ProbDist object
Support Read-only structure containing
information about support of
ProbDist object
Copy Value. To learn how this affects your use of the class, see Copying
Semantics Objects in the MATLAB Programming Fundamentals documentation.
18-1119
ProbDist class
See Also fitdist
ProbDistParametric class
ProbDistKernel class
ProbDistUnivParam class
ProbDistUnivKernel class
ProbDistUnivParam.ProbDistUnivParam constructor
ProbDistUnivKernel.ProbDistUnivKernel constructor
18-1120
ProbDistKernel class
Superclasses ProbDist
Purpose Object representing nonparametric probability distribution defined
by kernel smoothing
Description ProbDistKernel is an abstract class defining the properties and methods
of a nonparametric distribution defined by a kernel smoothing function.
Construction ProbDistKernel is an abstract class. You cannot create instances
of this class directly. You can construct an object in a subclass,
ProbDistUnivKernel, either by calling the subclass constructor,
ProbDistUnivKernel.ProbDistUnivKernel, or by using the fitdist
function.
Methods cdf Return cumulative distribution
function (CDF) for ProbDist object
pdf Return probability density
function (PDF) for ProbDist object
random Generate random number drawn
from ProbDist object
Note The above methods are inherited from the ProbDist class.
Properties BandWidth Read-only value specifying
bandwidth of kernel smoothing
function for ProbDistKernel
object
DistName Read-only string containing
probability distribution name of
ProbDist object
18-1121
ProbDistKernel class
InputData Read-only structure containing
information about input data to
ProbDist object
Kernel Read-only string specifying name
of kernel smoothing function for
ProbDistKernel object
Support Read-only structure containing
information about support of
ProbDist object
Note Some of the above properties are inherited from the ProbDist
class.
Copy Value. To learn how this affects your use of the class, see Copying
Semantics Objects in the MATLAB Programming Fundamentals documentation.
See Also fitdist
ProbDist class
ProbDistUnivKernel class
ProbDistUnivKernel.ProbDistUnivKernel constructor
18-1122
ProbDistParametric class
Superclasses ProbDist
Purpose Object representing parametric probability distribution
Description ProbDistParametric is an abstract class defining the properties and
methods of a parametric probability distribution.
Construction ProbDistParametric is an abstract class. You cannot create instances
of this class directly. You can construct an object in its subclass,
ProbDistUnivParam, either by calling the subclass constructor,
ProbDistUnivParam.ProbDistUnivParam, or by using the fitdist
function.
Methods cdf Return cumulative distribution
function (CDF) for ProbDist object
pdf Return probability density
function (PDF) for ProbDist object
random Generate random number drawn
from ProbDist object
Note The above methods are inherited from the ProbDist class.
Properties DistName Read-only string containing
probability distribution name of
ProbDist object
InputData Read-only structure containing
information about input data to
ProbDist object
18-1123
ProbDistParametric class
NLogL Read-only value specifying
negative log likelihood for input
data to ProbDistParametric
object
NumParams Read-only value specifying
number of parameters of
ProbDistParametric object
ParamCov Read-only covariance matrix
of parameter estimates of
ProbDistParametric object
ParamDescription Read-only cell array specifying
descriptions of parameters of
ProbDistParametric object
ParamIsFixed Read-only logical array
specifying fixed parameters
of ProbDistParametric object
ParamNames Read-only cell array specifying
names of parameters of
ProbDistParametric object
Params Read-only array specifying
values of parameters of
ProbDistParametric object
Support Read-only structure containing
information about support of
ProbDist object
Note Some of the above properties are inherited from the ProbDist
class.
Copy Value. To learn how this affects your use of the class, see Copying
Semantics Objects in the MATLAB Programming Fundamentals documentation.
18-1124
ProbDistParametric class
See Also fitdist
ProbDist class
ProbDistUnivParam class
ProbDistUnivParam.ProbDistUnivParam constructor
18-1125
ProbDistUnivKernel class
Superclasses ProbDistKernel
Purpose Object representing univariate kernel probability distribution
Description A ProbDistUnivKernel object represents a univariate nonparametric
probability distribution defined by kernel smoothing. You create this
object using the fitdist function to fit the distribution to data.
Construction fitdist Fit probability distribution to
data
Methods cdf Return cumulative distribution
function (CDF) for ProbDist object
icdf Return inverse cumulative
distribution function (ICDF) for
ProbDistUnivKernel object
iqr Return interquartile range (IQR)
for ProbDistUnivKernel object
median Return median of
ProbDistUnivKernel object
pdf Return probability density
function (PDF) for ProbDist object
random Generate random number drawn
from ProbDist object
Note Some of the above methods are inherited from the
ProbDistKernel class.
18-1126
ProbDistUnivKernel class
Properties BandWidth Read-only value specifying
bandwidth of kernel smoothing
function for ProbDistKernel
object
DistName Read-only string containing
probability distribution name of
ProbDist object
InputData Read-only structure containing
information about input data to
ProbDist object
Kernel Read-only string specifying name
of kernel smoothing function for
ProbDistKernel object
NLogL Read-only value specifying
negative log likelihood for input
data to ProbDistUnivKernel
object
Support Read-only structure containing
information about support of
ProbDist object
Note Some of the above properties are inherited from the
ProbDistKernel class.
Copy Value. To learn how this affects your use of the class, see Copying
Semantics Objects in the MATLAB Programming Fundamentals documentation.
References [1] Bowman, A. W., and A. Azzalini. Applied Smoothing Techniques for
Data Analysis. New York: Oxford University Press, 1997.
18-1127
ProbDistUnivKernel class
See Also fitdist
ksdensity
ProbDist class
ProbDistKernel class
ProbDistUnivKernel.ProbDistUnivKernel constructor
18-1128
ProbDistUnivKernel
Purpose Construct ProbDistUnivKernel object
Syntax PD = ProbDistUnivKernel(X)
PD = ProbDistUnivKernel(X, param1, val1, param2, val2, ...)
Description
Tip Although you can use this constructor function to create a
ProbDistUnivKernel object, using the fitdist function is an easier way
to create the ProbDistUnivKernel object.
PD = ProbDistUnivKernel(X) creates PD, a ProbDistUnivKernel
object, which represents a nonparametric probability distribution,
based on a normal kernel smoothing function.
PD = ProbDistUnivKernel(X, param1, val1, param2, val2,
...) specifies optional parameter name/value pairs, as described in
the Parameter/Values table. Parameter and value names are case
insensitive.
Input X A column vector of data.
Arguments
Note Any NaN values in X are ignored by the
fitting calculations.
18-1129
ProbDistUnivKernel
Parameter Values
'censoring' A Boolean vector the same size as X, containing 1s when
the corresponding elements in X are right-censored
observations and 0s when the corresponding elements
are exact observations. Default is a vector of 0s.
Note Any NaN values in this censoring vector are
ignored by the fitting calculations.
'frequency' A vector the same size as X, containing nonnegative
integers specifying the frequencies for the corresponding
elements in X. Default is a vector of 1s.
Note Any NaN values in this frequency vector are
ignored by the fitting calculations.
'kernel' A string specifying the type of kernel smoother to use.
Choices are:
• 'normal' (default)
• 'box'
• 'triangle'
• 'epanechnikov'
18-1130
ProbDistUnivKernel
Parameter Values
'support' Any of the following to specify the support:
• 'unbounded' — Default. If the density can extend
over the whole real line.
• 'positive' — To restrict it to positive values.
• A two-element vector giving finite lower and upper
limits for the support of the density.
'width' A value specifying the bandwidth of the kernel
smoothing window. The default is optimal for
estimating normal densities, but you may want to
choose a smaller value to reveal features such as
multiple modes.
Output PD An object in the ProbDistUnivKernel class,
Arguments which is derived from the ProbDist class.
It represents a nonparametric probability
distribution.
References [1] Bowman, A. W., and A. Azzalini. Applied Smoothing Techniques for
Data Analysis. New York: Oxford University Press, 1997.
See Also fitdist
ksdensity
18-1131
ProbDistUnivParam class
Superclasses ProbDistParametric
Purpose Object representing univariate parametric probability distribution
Description A ProbDistUnivParam object represents a univariate parametric
probability distribution. You create this object by using the constructor
(ProbDistUnivParam.ProbDistUnivParam) and supplying parameter
values, or by using the fitdist function to fit the distribution to data.
Construction fitdist Fit probability distribution to
data
ProbDistUnivParam Construct ProbDistUnivParam
object
Methods cdf Return cumulative distribution
function (CDF) for ProbDist object
icdf Return inverse cumulative
distribution function (ICDF) for
ProbDistUnivParam object
iqr Return interquartile range (IQR)
for ProbDistUnivParam object
mean Return mean of
ProbDistUnivParam object
median Return median of
ProbDistUnivParam object
paramci Return parameter confidence
intervals of ProbDistUnivParam
object
pdf Return probability density
function (PDF) for ProbDist object
18-1132
ProbDistUnivParam class
random Generate random number drawn
from ProbDist object
std Return standard deviation of
ProbDistUnivParam object
var Return variance of
ProbDistUnivParam object
Note Some of the above methods are inherited from the
ProbDistParametric class.
Properties DistName Read-only string containing
probability distribution name of
ProbDist object
InputData Read-only structure containing
information about input data to
ProbDist object
NLogL Read-only value specifying
negative log likelihood for input
data to ProbDistParametric
object
NumParams Read-only value specifying
number of parameters of
ProbDistParametric object
ParamCov Read-only covariance matrix
of parameter estimates of
ProbDistParametric object
ParamDescription Read-only cell array specifying
descriptions of parameters of
ProbDistParametric object
18-1133
ProbDistUnivParam class
ParamIsFixed Read-only logical array
specifying fixed parameters
of ProbDistParametric object
ParamNames Read-only cell array specifying
names of parameters of
ProbDistParametric object
Params Read-only array specifying
values of parameters of
ProbDistParametric object
Support Read-only structure containing
information about support of
ProbDist object
Note The above properties are inherited from the ProbDistParametric
class.
Note Parameter values are also properties. For example, if you
create PD, a univariate parametric probability distribution object
that represents a normal distribution, then PD.mu and PD.sigma are
properties that give the values of the mu and sigma parameters.
Copy Value. To learn how this affects your use of the class, see Copying
Semantics Objects in the MATLAB Programming Fundamentals documentation.
References [1] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous
Univariate Distributions. Vol. 1, Hoboken, NJ: Wiley-Interscience,
1993.
18-1134
ProbDistUnivParam class
[2] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous
Univariate Distributions. Vol. 2, Hoboken, NJ: Wiley-Interscience,
1994.
See Also fitdist
ProbDist class
ProbDistParametric class
ProbDistUnivParam.ProbDistUnivParam constructor
Appendix B, “Distribution Reference” — For more information on
parametric distributions
18-1135
ProbDistUnivParam
Purpose Construct ProbDistUnivParam object
Syntax PD = ProbDistUnivParam(DistName, Params)
Description PD = ProbDistUnivParam(DistName, Params) creates PD, a
ProbDistUnivParam object, which represents a probability distribution.
This distribution is defined by the parametric distribution specified by
DistName, with parameters specified by the numeric vector Params.
Input DistName A string specifying a distribution. Choices are:
Arguments
• 'beta'
• 'binomial'
• 'birnbaumsaunders'
• 'exponential'
• 'extreme value' or ev'
• 'gamma'
• 'generalized extreme value' or 'gev'
• 'generalized pareto' or 'gp'
• 'inversegaussian'
• 'logistic'
• 'loglogistic'
• 'lognormal'
• 'nakagami'
• 'negative binomial' or 'nbin'
• 'normal'
• 'poisson'
18-1136
ProbDistUnivParam
• 'rayleigh'
• 'rician'
• 'tlocationscale'
• 'weibull' or 'wbl'
For more information on these parametric
distributions, see Appendix B, “Distribution
Reference”.
Params Numeric vector of distribution parameters.
The number and type of parameters
depends on the distribution you specify with
DistName. For information on parameters
for each distribution type, see Appendix B,
“Distribution Reference”.
Output PD An object in the ProbDistUnivParam class,
Arguments which is derived from the ProbDist class.
It represents a parametric probability
distribution.
Examples 1 Create an object representing a normal distribution with a mean of
100 and a standard deviation of 10.
pd = ProbDistUnivParam('normal',[100 10])
pd =
normal distribution
mu = 100
sigma = 10
18-1137
ProbDistUnivParam
2 Generate a 4-by-5 matrix of random values from this distribution.
random(pd,4,5)
ans =
105.3767 103.1877 135.7840 107.2540 98.7586
118.3389 86.9231 127.6944 99.3695 114.8970
77.4115 95.6641 86.5011 107.1474 114.0903
108.6217 103.4262 130.3492 97.9503 114.1719
References [1] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous
Univariate Distributions. Vol. 1, Hoboken, NJ: Wiley-Interscience,
1993.
[2] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous
Univariate Distributions. Vol. 2, Hoboken, NJ: Wiley-Interscience,
1994.
See Also Appendix B, “Distribution Reference” — For more information on
parametric distributions
fitdist
18-1138
probplot
Purpose Probability plots
Syntax probplot(Y)
probplot(distribution,Y)
probplot(Y,cens,freq)
probplot(ax,Y)
probplot(...,'noref')
probplot(ax,PD)
probplot(ax,fun,params)
h = probplot(...)
Description probplot(Y) produces a normal probability plot comparing the
distribution of the data Y to the normal distribution. Y can be a single
vector, or a matrix with a separate sample in each column. The plot
includes a reference line useful for judging whether the data follow a
normal distribution.
probplot uses midpoint probability plotting positions. The ith sorted
value from a sample of size N is plotted against the midpoint in the
jump of the empirical CDF on the y axis. With uncensored data, that
midpoint is (i–0.5)/N. With censored data (see below), the y value is
more complicated to compute.
probplot(distribution,Y) creates a probability plot for the
distribution specified by distribution. Acceptable strings for
distribution are:
• 'exponential' — Exponential probability plot (nonnegative values)
• 'extreme value' — Extreme value probability plot (all values)
• 'lognormal' — Lognormal probability plot (positive values)
• 'normal' — Normal probability plot (all values)
• 'rayleigh' — Rayleigh probability plot (positive values)
• 'weibull' — Weibull probability plot (positive values)
18-1139
probplot
The y axis scale is based on the selected distribution. The x axis has a
log scale for the Weibull and lognormal distributions, and a linear scale
for the others.
Not all distributions are appropriate for all data sets, and probplot will
error when asked to create a plot with a data set that is inappropriate
for a specified distribution. Appropriate data ranges for each
distribution are given parenthetically in the list above.
probplot(Y,cens,freq) or probplot(distname,Y,cens,freq)
requires a vector Y. cens is a vector of the same size as Y and contains
1 for observations that are right-censored and 0 for observations that
are observed exactly. freq is a vector of the same size as Y, containing
integer frequencies for the corresponding elements in Y.
probplot(ax,Y) takes a handle ax to an existing probability plot, and
adds additional lines for the samples in Y. ax is a handle for a set of axes.
probplot(...,'noref') omits the reference line.
probplot(ax,PD) takes a probability distribution object, PD, and
adds a fitted line to the axes specified by ax to represent the
probability distribution specified by PD. PD is a ProbDist object of the
ProbDistUnivParam class or ProbDistUnivKernel class.
probplot(ax,fun,params) takes a function fun and a set of
parameters, params, and adds fitted lines to the axes of an existing
probability plot specified by ax. fun is a function handle to a cdf
function, specified with @ (for example, @wblcdf). params is the set of
parameters required to evaluate fun, and is specified as a cell array
or vector. The function must accept a vector of X values as its first
argument, then the optional parameters, and must return a vector of
cdf values evaluated at X.
h = probplot(...) returns handles to the plotted lines.
Examples Example 1
The following plot assesses two samples, one from a Weibull distribution
and one from a Rayleigh distribution, to see if they may have come
from a Weibull population.
18-1140
probplot
x1 = wblrnd(3,3,100,1);
x2 = raylrnd(3,100,1);
probplot('weibull',[x1 x2])
legend('Weibull Sample','Rayleigh Sample','Location','NW')
Example 2
Consider the following data, with about 20% outliers:
left_tail = -exprnd(1,10,1);
right_tail = exprnd(5,10,1);
center = randn(80,1);
18-1141
probplot
data = [left_tail;center;right_tail];
Neither a normal distribution nor a t distribution fits the tails very well:
probplot(data);
p = mle(data,'dist','tlo');
t = @(data,mu,sig,df)cdf('tlocationscale',data,mu,sig,df);
h = probplot(gca,t,p);
set(h,'color','r','linestyle','-')
title('{\bf Probability Plot}')
legend('Data','Normal','t','Location','NW')
18-1142
probplot
See Also normplot, ecdf, wblplot
18-1143
procrustes
Purpose Procrustes analysis
Syntax d = procrustes(X,Y)
[d,Z] = procrustes(X,Y)
[d,Z,transform] = procrustes(X,Y)
[...] = procrustes(...,'scaling',flag)
[...] = procrustes(...,'reflection',flag)
Description d = procrustes(X,Y) determines a linear transformation (translation,
reflection, orthogonal rotation, and scaling) of the points in matrix Y to
best conform them to the points in matrix X. The goodness-of-fit criterion
is the sum of squared errors. procrustes returns the minimized value
of this dissimilarity measure in d. d is standardized by a measure of the
scale of X, given by:
sum(sum((X-repmat(mean(X,1),size(X,1),1)).^2,1))
That is, the sum of squared elements of a centered version of X.
However, if X comprises repetitions of the same point, the sum of
squared errors is not standardized.
X and Y must have the same number of points (rows), and procrustes
matches Y(i) to X(i). Points in Y can have smaller dimension (number
of columns) than those in X. In this case, procrustes adds columns
of zeros to Y as necessary.
[d,Z] = procrustes(X,Y) also returns the transformed Y values.
[d,Z,transform] = procrustes(X,Y) also returns the transformation
that maps Y to Z. transform is a structure array with fields:
• c — Translation component
• T — Orthogonal rotation and reflection component
• b — Scale component
That is:
c = transform.c;
18-1144
procrustes
T = transform.T;
b = transform.b;
Z = b*Y*T + c;
[...] = procrustes(...,'scaling',flag), when flag is false,
allows you to compute the transformation without a scale component
(that is, with b equal to 1). The default flag is true.
[...] = procrustes(...,'reflection',flag), when flag is
false, allows you to compute the transformation without a reflection
component (that is, with det(T) equal to 1). The default flag is
'best', which computes the best-fitting transformation, whether
or not it includes a reflection component. A flag of true forces the
transformation to be computed with a reflection component (that is,
with det(T) equal to -1)
Examples This example creates some random points in two dimensions, then
rotates, scales, translates, and adds some noise to those points. It uses
procrustes to conform Y to X, then plots the original X and Y with the
transformed Y.
n = 10;
X = normrnd(0,1,[n 2]);
S = [0.5 -sqrt(3)/2; sqrt(3)/2 0.5];
Y = normrnd(0.5*X*S+2,0.05,n,2);
[d,Z,tr] = procrustes(X,Y);
plot(X(:,1),X(:,2),'rx',...
Y(:,1),Y(:,2),'b.',...
Z(:,1),Z(:,2),'bx');
18-1145
procrustes
References [1] Kendall, David G. “A Survey of the Statistical Theory of Shape.”
Statistical Science. Vol. 4, No. 2, 1989, pp. 87–99.
[2] Bookstein, Fred L. Morphometric Tools for Landmark Data.
Cambridge, UK: Cambridge University Press, 1991.
[3] Seber, G. A. F. Multivariate Observations. Hoboken, NJ: John Wiley
& Sons, Inc., 1984.
See Also cmdscale, factoran
18-1146
CompactTreeBagger.proximity
Purpose Proximity matrix for data
Syntax prox = proximity(B,X)
Description prox = proximity(B,X) computes a numeric matrix of size
Nobs-by-Nobs of proximities for data X, where Nobs is the number of
observations (rows) in X. Proximity between any two observations in the
input data is defined as a fraction of trees in the ensemble B for which
these two observations land on the same leaf. This is a symmetric
matrix with ones on the diagonal and off-diagonal elements ranging
from 0 to 1.
18-1147
TreeBagger.Proximity property
Purpose Proximity matrix for observations
Description The Proximity property is a numeric matrix of size Nobs-by-Nobs,
where Nobs is the number of observations in the training data,
containing measures of the proximity between observations. For any
two observations, their proximity is defined as the fraction of trees for
which these observations land on the same leaf. This is a symmetric
matrix with 1s on the diagonal and off-diagonal elements ranging from
0 to 1.
See Also CompactTreeBagger.proximity, classregtree.varimportance
18-1148
classregtree.prune
Purpose Prune tree
Syntax t2 = prune(t1,'level',level)
t2 = prune(t1,'nodes',nodes)
t2 = prune(t1)
Description t2 = prune(t1,'level',level) takes a decision tree t1 and a pruning
level level, and returns the decision tree t2 pruned to that level. If
level is 0, there is no pruning. Trees are pruned based on an optimal
pruning scheme that first prunes branches giving less improvement
in error cost.
t2 = prune(t1,'nodes',nodes) prunes the nodes listed in the nodes
vector from the tree. Any t1 branch nodes listed in nodes become leaf
nodes in t2, unless their parent nodes are also pruned. Use view to
display the node numbers for any node you select.
t2 = prune(t1) returns the decision tree t2 that is the full, unpruned
t1, but with optimal pruning information added. This is useful only if
t1 is created by pruning another tree, or by using the classregtree
function with the 'prune' parameter set to 'off'. If you plan to prune
a tree multiple times along the optimal pruning sequence, it is more
efficient to create the optimal pruning sequence first.
Pruning is the process of reducing a tree by turning some branch nodes
into leaf nodes and removing the leaf nodes under the original branch.
Examples Display the full tree for Fisher’s iris data:
load fisheriris;
t1 = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'},...
'splitmin',5)
t1 =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
18-1149
classregtree.prune
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t1)
18-1150
classregtree.prune
Display the next largest tree from the optimal pruning sequence:
t2 = prune(t1,'level',1)
t2 =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 class = versicolor
7 class = virginica
view(t2)
18-1151
classregtree.prune
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, test, view
18-1152
TreeBagger.Prune property
Purpose Flag to prune trees
Description The Prune property is true if decision trees are pruned and false if they
are not. Pruning decision trees is not recommended for ensembles. The
default value is false.
See Also classregtree.prune
18-1153
qrandstream.qrand
Purpose Generate quasi-random points from stream
Syntax x = qrand(q)
X = qrand(q,n)
Description x = qrand(q) returns the next value x in the quasi-random number
stream q of the qrandstream class. x is a 1-by-d vector, where d is the
dimension of the stream. The command sets q.State to the index in the
underlying point set of the next value to be returned.
X = qrand(q,n) returns the next n values X in an n-by-d matrix.
Objects q of the qrandstream class encapsulate properties of a specified
quasi-random number stream. Values of the stream are not generated
and stored in memory until q is accessed using qrand.
Examples Use qrandstream to construct a 3-D Halton stream, based on a point
set that skips the first 1000 values and then retains every 101st point:
q = qrandstream('halton',3,'Skip',1e3,'Leap',1e2)
q =
Halton quasi-random stream in 3 dimensions
Point set properties:
Skip : 1000
Leap : 100
ScrambleMethod : none
nextIdx = q.State
nextIdx =
1
Use qrand to generate two samples of size four:
X1 = qrand(q,4)
X1 =
0.0928 0.3475 0.0051
0.6958 0.2035 0.2371
0.3013 0.8496 0.4307
18-1154
qrandstream.qrand
0.9087 0.5629 0.6166
nextIdx = q.State
nextIdx =
5
X2 = qrand(q,4)
X2 =
0.2446 0.0238 0.8102
0.5298 0.7540 0.0438
0.3843 0.5112 0.2758
0.8335 0.2245 0.4694
nextIdx = q.State
nextIdx =
9
Use reset to reset the stream, then generate another sample:
reset(q)
nextIdx = q.State
nextIdx =
1
X = qrand(q,4)
X =
0.0928 0.3475 0.0051
0.6958 0.2035 0.2371
0.3013 0.8496 0.4307
0.9087 0.5629 0.6166
See Also qrandstream, reset
18-1155
qrandset class
Purpose Quasi-random point sets
Description qrandset is a base class that encapsulates a sequence of multi-
dimensional quasi-random numbers. This base class is abstract and
cannot be instantiated directly. Concrete subclasses include sobolset
and haltonset.
Construction qrandset Abstract quasi-random point set
class
Methods disp Display qrandset object
end Last index in indexing expression
for point set
length Length of point set
ndims Number of dimensions in matrix
net Generate quasi-random point set
scramble Scramble quasi-random point set
size Number of dimensions in matrix
subsref Subscripted reference for
qrandset
Properties Dimensions Number of dimensions
Leap Interval between points
ScrambleMethod Settings that control scrambling
18-1156
qrandset class
Skip Number of initial points to omit
from sequence
Type Name of sequence on which point
set P is based
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
See Also haltonset | sobolset
How To • “Quasi-Random Point Sets” on page 6-18
18-1157
qrandset
Purpose Abstract quasi-random point set class
Description qrandset is an abstract class, and you cannot create instances of it
directly. You must use haltonset or sobolset to create a qrandset
object.
See Also haltonset, sobolset
18-1158
qrandstream class
Purpose Quasi-random number streams
Construction qrandstream Construct quasi-random number
stream
Methods addlistener Add listener for event
delete Delete handle object
disp Display qrandstream object
eq Test handle equality
findobj Find objects matching specified
conditions
findprop Find property of MATLAB handle
object
ge Greater than or equal relation for
handles
gt Greater than relation for handles
isvalid Test handle validity
le Less than or equal relation for
handles
lt Less than relation for handles
ne Not equal relation for handles
notify Notify listeners of event
qrand Generate quasi-random points
from stream
rand Generate quasi-random points
from stream
reset Reset state
18-1159
qrandstream class
Properties PointSet Point set from which stream is
drawn
State Current state of the stream
Copy Handle. To learn how this affects your use of the class, see
Semantics Comparing Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
18-1160
qrandstream
Purpose Construct quasi-random number stream
Syntax q = qrandstream(type,d)
q = qrandstream(type,d,prop1,val1,prop2,val2,...)
q = qrandstream(p)
Description q = qrandstream(type,d) constructs a d-dimensional quasi-random
number stream q of the qrandstream class, of type specified by the
string type. type is either 'halton' or 'sobol', and q is based
on a point set from either the haltonset class or sobolset class,
respectively, with default property settings.
q = qrandstream(type,d,prop1,val1,prop2,val2,...) specifies
property name/value pairs for the point set on which the stream is
based. Applicable properties depend on type.
q = qrandstream(p) constructs a stream based on the specified point
set p. p must be a point set from either the haltonset class or sobolset
class.
Examples Construct a 3-D Halton stream, based on a point set that skips the first
1000 values and then retains every 101st point:
q = qrandstream('halton',3,'Skip',1e3,'Leap',1e2)
q =
Halton quasi-random stream in 3 dimensions
Point set properties:
Skip : 1000
Leap : 100
ScrambleMethod : none
nextIdx = q.State
nextIdx =
1
Use qrand to generate two samples of size four:
X1 = qrand(q,4)
18-1161
qrandstream
X1 =
0.0928 0.3475 0.0051
0.6958 0.2035 0.2371
0.3013 0.8496 0.4307
0.9087 0.5629 0.6166
nextIdx = q.State
nextIdx =
5
X2 = qrand(q,4)
X2 =
0.2446 0.0238 0.8102
0.5298 0.7540 0.0438
0.3843 0.5112 0.2758
0.8335 0.2245 0.4694
nextIdx = q.State
nextIdx =
9
Use reset to reset the stream, and then generate another sample:
reset(q)
nextIdx = q.State
nextIdx =
1
X = qrand(q,4)
X =
0.0928 0.3475 0.0051
0.6958 0.2035 0.2371
0.3013 0.8496 0.4307
0.9087 0.5629 0.6166
See Also haltonset | qrand | reset | sobolset
18-1162
qqplot
Purpose Quantile-quantile plot
Syntax qqplot(X)
qqplot(X,Y)
qqplot(X,PD)
qqplot(X,Y,pvec)
h = qqplot(X,Y,pvec)
Description qqplot(X) displays a quantile-quantile plot of the sample quantiles
of X versus theoretical quantiles from a normal distribution. If the
distribution of X is normal, the plot will be close to linear.
qqplot(X,Y) displays a quantile-quantile plot of two samples. If the
samples do come from the same distribution, the plot will be linear.
qqplot(X,PD) makes an empirical quantile-quantile plot of
the quantiles of the data in the vector X versus the quantiles
of the distribution specified by PD, a ProbDist object of the
ProbDistUnivParam class or ProbDistUnivKernel class.
For matrix X and Y, qqplot displays a separate line for each pair of
columns. The plotted quantiles are the quantiles of the smaller data set.
The plot has the sample data displayed with the plot symbol '+'.
Superimposed on the plot is a line joining the first and third quartiles of
each distribution (this is a robust linear fit of the order statistics of the
two samples). This line is extrapolated out to the ends of the sample to
help evaluate the linearity of the data.
Use qqplot(X,Y,pvec) to specify the quantiles in the vector pvec.
h = qqplot(X,Y,pvec) returns handles to the lines in h.
Examples The following example shows a quantile-quantile plot of two samples
from Poisson distributions.
x = poissrnd(10,50,1);
y = poissrnd(5,100,1);
qqplot(x,y);
18-1163
qqplot
See Also normplot
18-1164
quantile
Purpose Quantiles
Syntax Y = quantile(X,p)
Y = quantile(X,p,dim)
Description Y = quantile(X,p) returns quantiles of the values in X. p is a scalar
or a vector of cumulative probability values. When X is a vector, Y
is the same size as p, and Y(i) contains the p(i)th quantile. When
X is a matrix, the ith row of Y contains the p(i)th quantiles of each
column of X. For N-dimensional arrays, quantile operates along the
first nonsingleton dimension of X.
Y = quantile(X,p,dim) calculates quantiles along dimension dim. The
dimth dimension of Y has length length(P).
Quantiles are specified using cumulative probabilities from 0 to 1. For
an n-element vector X, quantile computes quantiles as follows:
1 The sorted values in X are taken as the (0.5/n), (1.5/n), ..., ([n–0.5]/n)
quantiles.
2 Linear interpolation is used to compute quantiles for probabilities
between (0.5/n) and ([n–0.5]/n).
3 The minimum or maximum values in X are assigned to quantiles for
probabilities outside that range.
quantile treats NaNs as missing values and removes them.
Examples y = quantile(x,.50); % the median of x
y = quantile(x,[.025 .25 .50 .75 .975]); % Summary of x
See Also prctile, iqr, median
18-1165
qrandstream.rand
Purpose Generate quasi-random points from stream
Syntax rand
rand(q,n)
rand(q)
rand(q,m,n)
rand(q,[m,n])
rand(q,m,n,p,...)
rand(q,[m,n,p,...])
Description rand returns a matrix of quasi-random values and is intended to allow
objects of the qrandstream class to be used in code that contains calls
to the rand method of the MATLAB pseudo-random randstream class.
Due to the multidimensional nature of quasi-random numbers, only
some syntaxes of rand are supported by the qrandstream class.
rand(q,n) returns an n-by-n matrix only when n is equal to the number
of dimensions. Any other value of n produces an error.
rand(q) returns a scalar only when the stream is in one dimension.
Having more than one dimension in q produces an error.
rand(q,m,n) or rand(q,[m,n]) returns an m-by-n matrix only when n
is equal to the number of dimensions in the stream. Any other value of
n produces an error.
rand(q,m,n,p,...) or rand(q,[m,n,p,...]) produces an error unless
p and all following dimensions sizes are equal to one.
Examples Generate the first 256 points from a 5-D Sobol sequence:
q = qrandstream('sobol',5);
X = rand(q,256,5);
See Also qrandstream, qrand, rand
18-1166
randg
Purpose Gamma random numbers
Syntax Y = randg
Y = randg(A)
Y = randg(A,m)
Y = randg(A,m,n,...)
Y = randg(A,[m,n,p])
Description Y = randg returns a scalar random value chosen from a gamma
distribution with unit scale and shape.
Y = randg(A) returns a matrix of random values chosen from gamma
distributions with unit scale. Y is the same size as A, and randg
generates each element of Y using a shape parameter equal to the
corresponding element of A.
Y = randg(A,m) returns an m-by-m matrix of random values chosen
from gamma distributions with shape parameters A. A is either an
m-by-m matrix or a scalar. If A is a scalar, randg uses that single shape
parameter value to generate all elements of Y.
Y = randg(A,m,n,...) or Y = randg(A,[m,n,p]) returns an
m-by-n-by-p array of random values chosen from gamma distributions
with shape parameters A. A is either an m-by-n-by-p array or a scalar.
randg produces pseudo-random numbers using the MATLAB functions
rand and randn. The sequence of numbers generated is determined by
the state of the default random number stream. Since MATLAB resets
the state at start-up, the sequence of numbers randg generates will be
the same in each session unless those states are changed.
To create reproducible output from randg, reset the state of the default
stream before calling randg. For example:
reset(RandStream.getDefaultStream,0);
r = randg(1,[10,1]);
always produces the same 10 values. See the RandStream
documentation for more information.
18-1167
randg
Calling randg changes the current states of rand, randn, and randi,
and therefore alters the outputs of subsequent calls to those functions.
To generate gamma random numbers and specify both the scale and
shape parameters, you should call gamrnd rather than calling randg
directly.
References [1] Marsaglia, G., and W. W. Tsang. “A Simple Method for Generating
Gamma Variables.” ACM Transactions on Mathematical Software. Vol.
26, 2000, pp. 363–372.
See Also gamrnd
18-1168
random
Purpose Random numbers
Syntax Y = random(name,A)
Y = random(name,A,B)
Y = random(name,A,B,C)
Y = random(...,m,n,...)
Y = random(...,[m,n,...])
Description Y = random(name,A) where name is the name of a distribution
that takes a single parameter, returns random numbers Y from the
one-parameter family of distributions specified by name. Parameter
values for the distribution are given in A.
Y is the same size as A.
Y = random(name,A,B) returns random numbers Y from a
two-parameter family of distributions. Parameter values for the
distribution are given in A and B.
If A and B are arrays, they must be the same size. If either A or B are
scalars, they are expanded to constant matrices of the same size.
Y = random(name,A,B,C) returns random numbers Y from a
three-parameter family of distributions. Parameter values for the
distribution are given in A, B, and C.
If A, B, and C are arrays, they must be the same size. If any of A, B, or C
are scalars, they are expanded to constant matrices of the same size.
Y = random(...,m,n,...) or Y = random(...,[m,n,...]) returns
an m-by-n-by... matrix of random numbers.
If any of A, B, or C are arrays, then the specified dimensions must
match the common dimensions of A, B, and C after any necessary scalar
expansion.
The following table denotes the acceptable strings for name, as well as
the parameters for that distribution:
18-1169
random
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'beta' or “Beta a b —
'Beta' Distribution”
on page B-4
'bino' or “Binomial n: number p: —
'Binomial' Distribution” of trials probability
on page B-7 of success
for each
trial
'chi2' or “Chi-Square ν: degrees — —
'Chisquare' Distribution” of freedom
on page
B-12
'exp' or “Exponential : mean — —
Distribution”
'Exponential'
on page
B-16
'ev' or “Extreme : location σ: scale —
'Extreme Value parameter parameter
Value' Distribution”
on page
B-19
'f' or 'F' “F ν 1: ν 2: —
Distribution” numerator denominator
'gam' or on page
“Gamma degrees
a : shapeof degrees
b : scale of —
'Gamma' B-25
Distribution” freedom
parameter freedom
parameter
on page
B-27
18-1170
random
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'gev' or “Generalized K: shape : location σ: scale
'GeneralizedExtreme parameter parameter parameter
Extreme Value
Value' Distribution”
on page
B-32
'gp' or “Generalized k: tail index σ: scale : threshold
'GeneralizedPareto (shape) parameter (location)
Pareto' Distribution” parameter parameter
on page
B-37
'geo' or “Geometric p: — —
'Geometric' Distribution” probability
'hyge' or on page parameter
“Hypergeometric
M : size of the K: number n: number
B-41
Distribution”
'Hypergeometric' population of items of samples
on page with the drawn
B-43 desired
characteristic
in the
population
'logn' or “Lognormal σ —
'Lognormal' Distribution”
on page
B-51
'nbin' or “Negative r: number p: —
'Negative Binomial of successes probability
Binomial' Distribution” of success
on page in a single
B-72 trial
18-1171
random
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'ncf' or “Noncentral ν1: ν 2: δ:
'Noncentral F numerator denominator noncentrality
F' Distribution” degrees of degrees of parameter
on page freedom freedom
B-78
'nct' or “Noncentral ν: degrees δ: —
'Noncentral t of freedom noncentrality
t' Distribution” parameter
'ncx2' or on page
“Noncentral ν: degrees δ: —
'Noncentral B-80
Chi-Square of freedom noncentrality
Chi-square' Distribution” parameter
on page
B-76
'norm' or “Normal : mean σ: standard —
'Normal' Distribution” deviation
on page
B-83
'poiss' or “Poisson λ: mean — —
'Poisson' Distribution”
on page
B-89
'rayl' or “Rayleigh b: scale — —
'Rayleigh' Distribution” parameter
on page
B-91
't' or 'T' “Student’s t ν: degrees — —
Distribution” of freedom
on page
B-95
18-1172
random
name Distribution Input Input Input
Parameter Parameter Parameter
A B C
'unif' or “Uniform a: lower b: upper —
'Uniform' Distribution endpoint endpoint
(Continuous)” (minimum) (maximum)
on page
B-99
'unid' or “Uniform N: — —
'Discrete Distribution maximum
Uniform' (Discrete)” observable
on page value
B-101
'wbl' or “Weibull a: scale b: shape —
'Weibull' Distribution” parameter parameter
on page
B-103
Examples Generate a 2-by-4 array of random values from the normal distribution
with mean 0 and standard deviation 1:
x1 = random('Normal',0,1,2,4)
x1 =
1.1650 0.0751 -0.6965 0.0591
0.6268 0.3516 1.6961 1.7971
The order of the parameters is the same as for normrnd.
Generate a single random value from Poisson distributions with rate
parameters 1, 2, ..., 6, respectively:
x2 = random('Poisson',1:6,1,6)
x2 =
0 0 1 2 5 7
See Also cdf, pdf, icdf, mle
18-1173
gmdistribution.random
Purpose Random numbers from Gaussian mixture distribution
Syntax y = random(obj)
Y = random(obj,n)
[Y,idx] = random(obj,n)
Description y = random(obj) generates a 1-by-d vector y drawn at random from
the d-dimensional Gaussian mixture distribution defined by obj. obj is
an object created by gmdistribution or fit.
Y = random(obj,n) generates an n-by-d matrix Y of n d-dimensional
random samples.
[Y,idx] = random(obj,n) also returns an n-by-1 vector idx, where
idx(I) is the index of the component used to generate Y(I,:).
Examples Create a gmdistribution object defining a two-component mixture of
bivariate Gaussian distributions:
MU = [1 2;-3 -5];
SIGMA = cat(3,[2 0;0 .5],[1 0;0 1]);
p = ones(1,2)/2;
obj = gmdistribution(MU,SIGMA,p);
ezcontour(@(x,y)pdf(obj,[x y]),[-10 10],[-10 10])
hold on
18-1174
gmdistribution.random
Generate 1000 random values:
Y = random(obj,1000);
scatter(Y(:,1),Y(:,2),10,'.')
18-1175
gmdistribution.random
See Also gmdistribution, fit, mvnrnd
18-1176
piecewisedistribution.random
Purpose Random numbers from piecewise distribution
Syntax r = random(obj)
R = random(obj,n)
R = random(obj,m,n)
R = random(obj,[m,n])
R = random(obj,m,n,p,...)
R = random(obj,[m,n,p,...])
Description r = random(obj) generates a pseudo-random number r drawn from
the piecewise distribution object obj.
R = random(obj,n) generates an n-by-n matrix of pseudo-random
numbers R.
R = random(obj,m,n) or R = random(obj,[m,n]) generates an
m-by-n matrix of pseudo-random numbers R.
R = random(obj,m,n,p,...) or R = random(obj,[m,n,p,...])
generates an m-by-n-by-p-by-... array of pseudo-random numbers R.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
r = random(obj)
r =
0.8285
See Also paretotails, cdf, icdf
18-1177
ProbDist.random
Purpose Generate random number drawn from ProbDist object
Syntax Y = random(PD)
Y = random(PD, N)
Y = random(PD, N, M, ...)
Description Y = random(PD) generates a random number drawn from the
distribution specified by PD, a ProbDist object.
Y = random(PD, N) generates an N-by-N array of random numbers
drawn from the distribution specified by PD, a ProbDist object.
Y = random(PD, N, M, ...) generates an N-by-M-by... array of
random numbers drawn from the distribution specified by PD, a
ProbDist object.
Input PD An object of the class ProbDistUnivParam or
Arguments ProbDistUnivKernel.
N A positive integer.
M A positive integer.
Output Y A random number drawn from the distribution
Arguments specified by PD.
See Also random
18-1178
randsample
Purpose Random sample
Syntax y = randsample(n,k)
y = randsample(population,k)
y = randsample(...,replace)
y = randsample(...,true,w)
y = randsample(s, ...)
Description y = randsample(n,k) returns a k-by-1 vector y of values sampled
uniformly at random, without replacement, from the integers 1 to n.
y = randsample(population,k), where population is a vector of two
or more values, returns a vector of values sampled uniformly at random,
without replacement, from the values in the vector population. The
orientation of y (row or column) is the same as population.
y = randsample(...,replace) returns a sample taken with
replacement if replace is true, or without replacement if replace is
false. The default is false.
y = randsample(...,true,w) returns a weighted sample taken with
replacement, using a vector of positive weights w, whose length is
n. The probability that the integer i is selected for an entry of y is
w(i)/sum(w). Usually, w is a vector of probabilities. randsample does
not support weighted sampling without replacement.
y = randsample(s, ...) uses the stream s for random number
generation. s is a member of the RandStream class. Default is the
MATLAB default random number stream.
Examples The following command generates a random sequence of the characters
A, C, G, and T, with replacement, according to the specified probabilities.
R = randsample('ACGT',48,true,[0.15 0.35 0.35 0.15])
See Also rand, randperm
18-1179
randtool
Purpose Interactive random number generation
Syntax randtool
Description randtool opens the Random Number Generation Tool.
The Random Number Generation Tool is a graphical user interface that
generates random samples from specified probability distributions and
displays the samples as histograms. Use the tool to explore the effects
of changing parameters and sample size on the distributions.
18-1180
randtool
Choose distribution Sample size
Histogram
Parameter
bounds
Parameter
value
Parameter
control Additional Sample again Export to
parameters from the same workspace
distribution
18-1181
randtool
Start by selecting a distribution, then enter the desired sample size.
You can also
• Use the controls at the bottom of the window to set parameter values
for the distribution and to change their upper and lower bounds.
• Draw another sample from the same distribution, with the same
size and parameters.
• Export the current sample to your workspace. A dialog box enables
you to provide a name for the sample.
See Also disttool, dfittool
18-1182
range
Purpose Range of values
Syntax range(X)
y = range(X,dim)
Description range(X) returns the difference between the maximum and the
minimum of a sample. For vectors, range(x) is the range of the
elements. For matrices, range(X) is a row vector containing the range
of each column of X. For N-dimensional arrays, range operates along
the first nonsingleton dimension of X.
y = range(X,dim) operates along the dimension dim of X.
range treats NaNs as missing values and ignores them.
The range is an easily-calculated estimate of the spread of a sample.
Outliers have an undue influence on this statistic, which makes it an
unreliable estimator.
Examples The range of a large sample of standard normal random numbers is
approximately six. This is the motivation for the process capability
indices Cp and Cpk in statistical quality control applications.
rv = normrnd(0,1,1000,5);
near6 = range(rv)
near6 =
6.1451 6.4986 6.2909 5.8894 7.0002
See Also std, iqr, mad
18-1183
ranksum
Purpose Wilcoxon rank sum test
Syntax p = ranksum(x,y)
[p,h] = ranksum(x,y)
[p,h] = ranksum(x,y,'alpha',alpha)
[p,h] = ranksum(...,'method',method)
[p,h,stats] = ranksum(...)
Description p = ranksum(x,y) performs a two-sided rank sum test of the null
hypothesis that data in the vectors x and y are independent samples
from identical continuous distributions with equal medians, against the
alternative that they do not have equal medians. x and y can have
different lengths. The p value of the test is returned in p. The test is
equivalent to a Mann-Whitney U-test.
[p,h] = ranksum(x,y) returns the result of the test in h. h = 1
indicates a rejection of the null hypothesis at the 5% significance
level. h = 0 indicates a failure to reject the null hypothesis at the 5%
significance level.
[p,h] = ranksum(x,y,'alpha',alpha) performs the test at the
(100*alpha)% significance level. The default, when unspecified, is
alpha = 0.05.
[p,h] = ranksum(...,'method',method) computes the p value using
either an exact algorithm, when method is 'exact', or a normal
approximation, when method is 'approximate'. Because the exact
method can be very slow on large samples, the default is exact for small
samples and approximate for large samples.
[p,h,stats] = ranksum(...) returns the structure stats with the
following fields:
• ranksum — Value of the rank sum test statistic
• zval — Value of the z-statistic (computed only for large samples)
18-1184
ranksum
Examples Test the hypothesis of equal medians for two independent unequal-sized
samples. The sampling distributions are identical except for a shift
of 0.25.
x = unifrnd(0,1,10,1);
y = unifrnd(0.25,1.25,15,1);
[p,h] = ranksum(x,y)
p =
0.0375
h =
1
The test rejects the null hypothesis of equal medians at the default
5% significance level.
References [1] Gibbons, J. D. Nonparametric Statistical Inference. New York:
Marcel Dekker, 1985.
[2] Hollander, M., and D. A. Wolfe. Nonparametric Statistical Methods.
Hoboken, NJ: John Wiley & Sons, Inc., 1999.
See Also kruskalwallis, signrank, signtest, ttest2
18-1185
raylcdf
Purpose Rayleigh cumulative distribution function
Syntax P = raylcdf(X,B)
Description P = raylcdf(X,B) computes the Rayleigh cdf at each of the values in
X using the corresponding scale parameter, B. X and B can be vectors,
matrices, or multidimensional arrays that all have the same size. A
scalar input for X or B is expanded to a constant array with the same
dimensions as the other input.
The Rayleigh cdf is
⎛ −t2 ⎞
x t ⎜ 2⎟
∫0 b2
2b ⎠
y = F ( x | b) = e⎝ dt
Examples x = 0:0.1:3;
p = raylcdf(x,1);
plot(x,p)
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3
References [1] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions.
Hoboken, NJ: Wiley-Interscience, 2000. pp. 134–136.
See Also cdf, raylpdf, raylinv, raylstat, raylfit, raylrnd
“Rayleigh Distribution” on page B-91
18-1186
raylfit
Purpose Rayleigh parameter estimates
Syntax raylfit(data,alpha)
[phat,pci] = raylfit(data,alpha)
Description raylfit(data,alpha) returns the maximum likelihood estimates of the
parameter of the Rayleigh distribution given the data in the vector data.
[phat,pci] = raylfit(data,alpha) returns the maximum likelihood
estimate and 100(1 - alpha)% confidence interval given the data. The
default value of the optional parameter alpha is 0.05, corresponding to
95% confidence intervals.
See Also mle, raylpdf, raylcdf, raylinv, raylstat, raylrnd
“Rayleigh Distribution” on page B-91
18-1187
raylinv
Purpose Rayleigh inverse cumulative distribution function
Syntax X = raylinv(P,B)
Description X = raylinv(P,B) returns the inverse of the Rayleigh cumulative
distribution function using the corresponding scale parameter, B at
the corresponding probabilities in P. P and B can be vectors, matrices,
or multidimensional arrays that all have the same size. A scalar input
for P or B is expanded to a constant array with the same dimensions
as the other input.
Examples x = raylinv(0.9,1)
x =
2.1460
See Also icdf, raylcdf, raylpdf, raylrnd, raylstat
“Rayleigh Distribution” on page B-91
18-1188
raylpdf
Purpose Rayleigh probability density function
Syntax Y = raylpdf(X,B)
Description Y = raylpdf(X,B) computes the Rayleigh pdf at each of the values in
X using the corresponding scale parameter, B. X and B can be vectors,
matrices, or multidimensional arrays that all have the same size, which
is also the size of Y. A scalar input for X or B is expanded to a constant
array with the same dimensions as the other input.
The Rayleigh pdf is
⎛ − x2 ⎞
x ⎜ 2⎟
2b ⎠
y = f ( x | b) = e⎝
b2
Examples x = 0:0.1:3;
p = raylpdf(x,1);
plot(x,p)
See Also pdf, raylcdf, raylinv, raylstat, raylfit, raylrnd
“Rayleigh Distribution” on page B-91
18-1189
raylrnd
Purpose Rayleigh random numbers
Syntax R = raylrnd(B)
R = raylrnd(B,v)
R = raylrnd(B,m,n)
Description R = raylrnd(B) returns a matrix of random numbers chosen from
the Rayleigh distribution with scale parameter, B. B can be a vector, a
matrix, or a multidimensional array. The size of R is the size of B.
R = raylrnd(B,v) returns a matrix of random numbers chosen from
the Rayleigh distribution with parameter B, where v is a row vector. If v
is a 1-by-2 vector, R is a matrix with v(1) rows and v(2) columns. If v
is 1-by-n, R is an n-dimensional array.
R = raylrnd(B,m,n) returns a matrix of random numbers chosen from
the Rayleigh distribution with parameter B, where scalars m and n are
the row and column dimensions of R.
Examples r = raylrnd(1:5)
r =
1.7986 0.8795 3.3473 8.9159 3.5182
See Also random, raylpdf, raylcdf, raylinv, raylstat, raylfit
“Rayleigh Distribution” on page B-91
18-1190
raylstat
Purpose Rayleigh mean and variance
Syntax [M,V] = raylstat(B)
Description [M,V] = raylstat(B) returns the mean of and variance for the
Rayleigh distribution with scale parameter B.
The mean of the Rayleigh distribution with parameter b is b / 2 and
the variance is
4 − 2
b
2
Examples [mn,v] = raylstat(1)
mn =
1.2533
v =
0.4292
See Also raylpdf, raylcdf, raylinv, raylfit, raylrnd
“Rayleigh Distribution” on page B-91
18-1191
rcoplot
Purpose Residual case order plot
Syntax rcoplot(r,rint)
Description rcoplot(r,rint) displays an errorbar plot of the confidence intervals
on the residuals from a regression. The residuals appear in the plot in
case order. Inputs r and rint are outputs from the regress function.
Examples The following plots residuals and prediction intervals from a regression
of a linearly additive model to the data in moore.mat:
load moore
X = [ones(size(moore,1),1) moore(:,1:5)];
y = moore(:,6);
alpha = 0.05;
[betahat,Ibeta,res,Ires,stats] = regress(y,X,alpha);
rcoplot(res,Ires)
18-1192
rcoplot
The interval around the first residual, shown in red, does not contain
zero. This indicates that the residual is larger than expected in 95% of
new observations, and suggests the data point is an outlier.
See Also regress
18-1193
refcurve
Purpose Add reference curve to plot
Syntax refcurve(p)
refcurve
hcurve = refcurve(...)
Description refcurve(p) adds a polynomial reference curve with coefficients p to
the current axes. If p is a vector with n+1 elements, the curve is:
y = p(1)*x^n + p(2)*x^(n-1) + ... + p(n)*x + p(n+1)
refcurve with no input arguments adds a line along the x axis.
hcurve = refcurve(...) returns the handle hcurve to the curve.
Examples Example 1
Plot data from a population with a polynomial trend and use refcurve
to add both the population and fitted mean functions:
p = [1 -2 -1 0];
t = 0:0.1:3;
y = polyval(p,t) + 0.5*randn(size(t));
plot(t,y,'ro')
h = refcurve(p);
set(h,'Color','r')
q = polyfit(t,y,3);
refcurve(q)
legend('Data','Population Mean','Fitted Mean',...
'Location','NW')
18-1194
refcurve
Example 2
Plot trajectories of a batted baseball, with and without air resistance.
Relevant physical constants are:
M = 0.145; % Mass (kg)
R = 0.0366; % Radius (m)
A = pi*R^2; % Area (m^2)
rho = 1.2; % Density of air (kg/m^3)
C = 0.5; % Drag coefficient
18-1195
refcurve
D = rho*C*A/2;
% Drag proportional to the square of the speed
g = 9.8; % Acceleration due to gravity (m/s^2)
First, simulate the trajectory with drag proportional to the square of
the speed, assuming constant acceleration in each time interval:
dt = 1e-2; % Simulation time interval (s)
r0 = [0 1]; % Initial position (m)
s0 = 50; % Initial speed (m/s)
alpha0 = 35; % Initial angle (deg)
v0=s0*[cosd(alpha0) sind(alpha0)]; % Initial velocity (m/s)
r = r0;
v = v0;
trajectory = r0;
while r(2) > 0
a = [0 -g]-(D/M)*norm(v)*v;
v = v + a*dt;
r = r + v*dt + (1/2)*a*(dt^2);
trajectory = [trajectory;r];
end
Second, use refcurve to add the drag-free parabolic trajectory (found
analytically) to a plot of trajectory:
plot(trajectory(:,1),trajectory(:,2),'m','LineWidth',2)
xlim([0,250])
h = refcurve([-g/(2*v0(1)^2),...
(g*r0(1)/v0(1)^2)+(v0(2)/v0(1)),...
(-g*r0(1)^2/(2*v0(1)^2))-(v0(2)*r0(1)/v0(1))+r0(2)]);
set(h,'Color','c','LineWidth',2)
axis equal
ylim([0,50])
grid on
xlabel('Distance (m)')
ylabel('Height (m)')
title('{\bf Baseball Trajectories}')
18-1196
refcurve
legend('With Drag','Without Drag')
See Also refline, lsline, gline, polyfit
18-1197
refline
Purpose Add reference line to plot
Syntax refline(m,b)
refline(coeffs)
refline
hline = refline(...)
Description refline(m,b) adds a reference line with slope m and intercept b to
the current axes.
refline(coeffs), where coeffs is a two-element coefficient vector,
adds the line
y = coeffs(1)*x + coeffs(2)
to the figure.
refline with no input arguments is equivalent to lsline.
hline = refline(...) returns the handle hline to the line.
Examples Add a reference line at the mean of a data scatter and its least-squares
line:
x = 1:10;
y = x + randn(1,10);
scatter(x,y,25,'b','*')
lsline
mu = mean(y);
hline = refline([0 mu]);
set(hline,'Color','r')
18-1198
refline
See Also refcurve, lsline, gline
18-1199
regress
Purpose Multiple linear regression
Syntax b = regress(y,X)
[b,bint] = regress(y,X)
[b,bint,r] = regress(y,X)
[b,bint,r,rint] = regress(y,X)
[b,bint,r,rint,stats] = regress(y,X)
[...] = regress(y,X,alpha)
Description b = regress(y,X) returns a p-by-1 vector b of coefficient estimates for
a multilinear regression of the responses in y on the predictors in X. X
is an n-by-p matrix of p predictors at each of n observations. y is an
n-by-1 vector of observed responses.
regress treats NaNs in X or y as missing values, and ignores them.
If the columns of X are linearly dependent, regress obtains a basic
solution by setting the maximum number of elements of b to zero.
[b,bint] = regress(y,X) returns a p-by-2 matrix bint of 95%
confidence intervals for the coefficient estimates. The first column of
bint contains lower confidence bounds for each of the p coefficient
estimates; the second column contains upper confidence bounds.
If the columns of X are linearly dependent, regress returns zeros in
elements of bint corresponding to the zero elements of b.
[b,bint,r] = regress(y,X) returns an n-by-1 vector r of residuals.
[b,bint,r,rint] = regress(y,X) returns an n-by-2 matrix rint of
intervals that can be used to diagnose outliers. If the interval rint(i,:)
for observation i does not contain zero, the corresponding residual is
larger than expected in 95% of new observations, suggesting an outlier.
In a linear model, observed values of y are random variables, and so are
their residuals. Residuals have normal distributions with zero mean
but with different variances at different values of the predictors. To put
residuals on a comparable scale, they are “Studentized,” that is, they are
divided by an estimate of their standard deviation that is independent
of their value. Studentized residuals have t distributions with known
18-1200
regress
degrees of freedom. The intervals returned in rint are shifts of the 95%
confidence intervals of these t distributions, centered at the residuals.
[b,bint,r,rint,stats] = regress(y,X) returns a 1-by-4 vector
stats that contains, in order, the R2 statistic, the F statistic and its p
value, and an estimate of the error variance.
Note When computing statistics, X should include a column of 1s so
that the model contains a constant term. The F statistic and its p value
are computed under this assumption, and they are not correct for models
without a constant. The R2 statistic can be negative for models without
a constant, indicating that the model is not appropriate for the data.
[...] = regress(y,X,alpha) uses a 100*(1-alpha)% confidence
level to compute bint and rint.
Examples Load data on cars; identify weight and horsepower as predictors,
mileage as the response:
load carsmall
x1 = Weight;
x2 = Horsepower; % Contains NaN data
y = MPG;
Compute regression coefficients for a linear model with an interaction
term:
X = [ones(size(x1)) x1 x2 x1.*x2];
b = regress(y,X) % Removes NaN data
b =
60.7104
-0.0102
-0.1882
0.0000
Plot the data and the model:
18-1201
regress
scatter3(x1,x2,y,'filled')
hold on
x1fit = min(x1):100:max(x1);
x2fit = min(x2):10:max(x2);
[X1FIT,X2FIT] = meshgrid(x1fit,x2fit);
YFIT = b(1) + b(2)*X1FIT + b(3)*X2FIT + b(4)*X1FIT.*X2FIT;
mesh(X1FIT,X2FIT,YFIT)
xlabel('Weight')
ylabel('Horsepower')
zlabel('MPG')
view(50,10)
18-1202
regress
References [1] Chatterjee, S., and A. S. Hadi. “Influential Observations, High
Leverage Points, and Outliers in Linear Regression.” Statistical Science.
Vol. 1, 1986, pp. 379–416.
See Also regstats, mvregress, robustfit, stepwisefit, rcoplot
18-1203
regstats
Purpose Regression diagnostics
Syntax regstats(y,X,model)
stats = regstats(...)
stats = regstats(y,X,model,whichstats)
Description regstats(y,X,model) performs a multilinear regression of the
responses in y on the predictors in X. X is an n-by-p matrix of p predictors
at each of n observations. y is an n-by-1 vector of observed responses.
Note By default, regstats adds a first column of 1s to X, corresponding
to a constant term in the model. Do not enter a column of 1s directly
into X.
The optional input model controls the regression model. By default,
regstats uses a linear additive model with a constant term. model can
be any one of the following strings:
• 'linear' — Constant and linear terms (the default)
• 'interaction' — Constant, linear, and interaction terms
• 'quadratic' — Constant, linear, interaction, and squared terms
• 'purequadratic' — Constant, linear, and squared terms
Alternatively, model can be a matrix of model terms accepted by the
x2fx function. See x2fx for a description of this matrix and for a
description of the order in which terms appear. You can use this matrix
to specify other models including ones without a constant term.
With this syntax, the function displays a graphical user interface (GUI)
with a list of diagnostic statistics, as shown in the following figure.
18-1204
regstats
18-1205
regstats
When you select check boxes corresponding to the statistics you want
to compute and click OK, regstats returns the selected statistics to
the MATLAB workspace. The names of the workspace variables are
displayed on the right-hand side of the interface. You can change the
name of the workspace variable to any valid MATLAB variable name.
stats = regstats(...) creates the structure stats, whose fields
contain all of the diagnostic statistics for the regression. This syntax
does not open the GUI. The fields of stats are listed in the following
table.
Field Description
Q Q from the QR decomposition of the design matrix
R R from the QR decomposition of the design matrix
beta Regression coefficients
covb Covariance of regression coefficients
yhat Fitted values of the response data
r Residuals
mse Mean squared error
rsquare R2 statistic
adjrsquare Adjusted R2 statistic
leverage Leverage
hatmat Hat matrix
s2_i Delete-1 variance
beta_i Delete-1 coefficients
standres Standardized residuals
studres Studentized residuals
dfbetas Scaled change in regression coefficients
dffit Change in fitted values
18-1206
regstats
Field Description
dffits Scaled change in fitted values
covratio Change in covariance
cookd Cook’s distance
tstat t statistics and p-values for coefficients
fstat F statistic and p-value
dwstat Durbin-Watson statistic and p-value
Note that the fields names of stats correspond to the names of the
variables returned to the MATLAB workspace when you use the GUI.
For example, stats.beta corresponds to the variable beta that is
returned when you select Coefficients in the GUI and click OK.
stats = regstats(y,X,model,whichstats) returns only the statistics
that you specify in whichstats. whichstats can be a single string
such as 'leverage' or a cell array of strings such as {'leverage'
'standres' 'studres'}. Set whichstats to 'all' to return all of
the statistics.
Note The F statistic is computed under the assumption that the model
contains a constant term. It is not correct for models without a constant.
The R2 statistic can be negative for models without a constant, which
indicates that the model is not appropriate for the data.
Examples Open the regstats GUI using data from hald.mat:
load hald
regstats(heat,ingredients,'linear');
Select Fitted Values and Residuals in the GUI:
18-1207
regstats
Click OK to export the fitted values and residuals to the MATLAB
workspace in variables named yhat and r, respectively.
You can create the same variables using the stats output, without
opening the GUI:
whichstats = {'yhat','r'};
stats = regstats(heat,ingredients,'linear',whichstats);
yhat = stats.yhat;
r = stats.r;
References [1] Belsley, D. A., E. Kuh, and R. E. Welsch. Regression Diagnostics.
Hoboken, NJ: John Wiley & Sons, Inc., 1980.
[2] Chatterjee, S., and A. S. Hadi. “Influential Observations, High
Leverage Points, and Outliers in Linear Regression.” Statistical Science.
Vol. 1, 1986, pp. 379–416.
[3] Cook, R. D., and S. Weisberg. Residuals and Influence in Regression.
New York: Chapman & Hall/CRC Press, 1983.
[4] Goodall, C. R. “Computation Using the QR Decomposition.”
Handbook in Statistics. Vol. 9, Amsterdam: Elsevier/North-Holland,
1993.
See Also x2fx, regress, stepwise, leverage
18-1208
gmdistribution.RegV property
Purpose Value of 'Regularize' parameter
Description The value of the parameter 'Regularize'.
Note This property applies only to gmdistribution objects constructed
with fit.
18-1209
categorical.reorderlevels
Purpose Reorder levels
Syntax B = reorderlevels(A,newlevels)
Description B = reorderlevels(A,newlevels) reorders the levels of the
categorical array A. newlevels is a cell array of strings or a 2-D
character matrix that specifies the new order. newlevels must be a
reordering of getlabels(A).
The order of the levels of an ordinal array has significance for relational
operators, minimum and maximum, and for sorting.
Examples Reorder hockey standings:
standings = ordinal(1:3,{'Leafs','Canadiens','Bruins'});
getlabels(standings)
ans =
'Leafs' 'Canadiens' 'Bruins'
standings = reorderlevels(standings,...
{'Canadiens','Leafs','Bruins'});
getlabels(standings)
ans =
'Canadiens' 'Leafs' 'Bruins'
See Also addlevels,droplevels, getlabels, islevel,mergelevels
18-1210
cvpartition.repartition
Purpose Repartition data for cross-validation
Syntax cnew = repartition(c)
Description cnew = repartition(c) constructs an object cnew of the cvpartition
class defining a random partition of the same type as c, where c is also
an object of the cvpartition class.
Repartitioning is useful for Monte-Carlo repetitions of cross-validation
analyses. repartition is called by crossval when the 'mcreps'
parameter is specified.
Examples Partition and repartition 100 observations for 3-fold cross-validation:
c = cvpartition(100,'kfold',3)
c =
K-fold cross validation partition
N: 100
NumTestSets: 3
TrainSize: 67 66 67
TestSize: 33 34 33
cnew = repartition(c)
cnew =
K-fold cross validation partition
N: 100
NumTestSets: 3
TrainSize: 67 66 67
TestSize: 33 34 33
Check for equality of the test data in the first fold:
isequal(test(c,1),test(cnew,1))
ans =
0
See Also cvpartition
18-1211
dataset.replacedata
Purpose Replace dataset variables
Syntax B = replacedata(A,X)
B = replacedata(A,X,vars)
B = replacedata(A,fun)
B = replacedata(A,fun,vars)
Description B = replacedata(A,X) creates a dataset array B with the same
variables as the dataset array A, but with the data for those variables
replaced by the data in the array X. replacedata creates each variable
in B using one or more columns from X, in order. X must have as many
columns as the total number of columns in all of the variables in A, and
as many rows as A has observations.
B = replacedata(A,X,vars) creates a dataset array B with the same
variables as the dataset array A, but with the data for the variables
specified in vars replaced by the data in the array X. The remaining
variables in B are copies of the corresponding variables in A. vars is a
positive integer, a vector of positive integers, a variable name, a cell
array containing one or more variable names, or a logical vector. Each
variable in B has as many columns as the corresponding variable in A. X
must have as many columns as the total number of columns in all the
variables specified in vars.
B = replacedata(A,fun) or B = replacedata(A,fun,vars) creates
a dataset array B by applying the function fun to the values in A’s
variables. replacedata first horizontally concatenates A’s variables into
a single array, then applies the function fun. The specified variables in
A must have types and sizes compatible with the concatenation. fun is a
function handle that accepts a single input array and returns an array
with the same number of rows and columns as the input.
Examples data = dataset({rand(3,3),'Var1','Var2','Var3'})
% Use ZSCORE to normalize each variable in a dataset array
% separately, by explicitly extracting and transforming the
% data, and then replacing it.
18-1212
dataset.replacedata
X = double(data);
X = zscore(X);
data = replacedata(data,X)
% Equivalently, provide a handle to ZSCORE.
data = replacedata(data,@zscore)
% Use ZSCORE to normalize each observation in a dataset
% array separately by creating an anonymous function.
data = replacedata(data,@(x) zscore(x,[],2))
See Also dataset
18-1213
categorical.repmat
Purpose Replicate and tile categorical array
Syntax B = repmat(A,m,n)
B = repmat(A,[m n p ...])
Description B = repmat(A,m,n) creates a large array B consisting of an m-by-n
tiling of copies of the categorical array A. The size of B is [size(A,1)*M
size(A,2)*N, size(A,3), ...]. repmat(A,n) creates an n-by-n
tiling.
B = repmat(A,[m n p ...]) tiles the categorical array A to produce
a multidimensional array B composed of copies of A. The size of B is
[size(A,1)*M, size(A,2)*N, size(A,3)*P, ...].
See Also ndims, size
18-1214
qrandstream.reset
Purpose Reset state
Syntax reset(q)
Description reset(q) resets the state of the quasi-random number stream q of the
qrandstream class back to its initial state, 1. Subsequent points drawn
from the stream will be the same as those drawn from a new stream.
The command is equivalent to q.State = 1.
Examples Use qrandstream to construct a 3-D Halton stream, based on a point
set that skips the first 1000 values and then retains every 101st point:
q = qrandstream('halton',3,'Skip',1e3,'Leap',1e2)
q =
Halton quasi-random stream in 3 dimensions
Point set properties:
Skip : 1000
Leap : 100
ScrambleMethod : none
nextIdx = q.State
nextIdx =
1
Use qrand to generate two samples of size four:
X1 = qrand(q,4)
X1 =
0.0928 0.3475 0.0051
0.6958 0.2035 0.2371
0.3013 0.8496 0.4307
0.9087 0.5629 0.6166
nextIdx = q.State
nextIdx =
5
X2 = qrand(q,4)
18-1215
qrandstream.reset
X2 =
0.2446 0.0238 0.8102
0.5298 0.7540 0.0438
0.3843 0.5112 0.2758
0.8335 0.2245 0.4694
nextIdx = q.State
nextIdx =
9
Use reset to reset the stream, then generate another sample:
reset(q)
nextIdx = q.State
nextIdx =
1
X = qrand(q,4)
X =
0.0928 0.3475 0.0051
0.6958 0.2035 0.2371
0.3013 0.8496 0.4307
0.9087 0.5629 0.6166
See Also qrandstream, qrand
18-1216
categorical.reshape
Purpose Resize categorical array
Syntax B = reshape(A,M,N)
B = reshape(A,m,n,p,...)
reshape(A,[m n p ...])
B = reshape(A,...,[],...)
Description B = reshape(A,M,N) returns an m-by-n categorical matrix whose
elements are taken columnwise from the categorical array A. An error
results if A does not have m*n elements.
B = reshape(A,m,n,p,...) or reshape(A,[m n p ...]) returns
an array with the same elements as A but reshaped to have the size
m-by-n-by-p-by-... . m*n*p*... must be the same as numel(A).
B = reshape(A,...,[],...) calculates the length of the dimension
represented by[], such that the product of the dimensions equals
numel(A). numel(A) must be evenly divisible by the product of the
known dimensions. You can use only one occurrence of [].
In general, reshape(A,siz) returns an array with the same elements
as A but reshaped to the size siz. prod(siz) must be the same as
numel(A).
See Also shiftdim, squeeze
18-1217
ridge
Purpose Ridge regression
Syntax b = ridge(y,X,k)
b = ridge(y,X,k,scaled)
Description b = ridge(y,X,k) returns a vector b of coefficient estimates for a
multilinear ridge regression of the responses in y on the predictors in X.
X is an n-by-p matrix of p predictors at each of n observations. y is an
n-by-1 vector of observed responses. k is a vector of ridge parameters. If
k has m elements, b is p-by-m. By default, b is computed after centering
and scaling the predictors to have mean 0 and standard deviation 1.
The model does not include a constant term, and X should not contain
a column of 1s.
b = ridge(y,X,k,scaled) uses the {0,1}-valued flag scaled to
determine if the coefficient estimates in b are restored to the scale of the
original data. ridge(y,X,k,0) performs this additional transformation.
In this case, b contains p+1 coefficients for each value of k, with the first
row corresponding to a constant term in the model. ridge(y,X,k,1)
is the same as ridge(y,X,k). In this case, b contains p coefficients,
without a coefficient for a constant term.
The relationship between b0 = ridge(y,X,k,0) and b1 =
ridge(y,X,k,1) is given by
m = mean(X);
s = std(X,0,1)';
b1_scaled = b1./s;
b0 = [mean(y)-m*b1_scaled; b1_scaled]
This can be seen by replacing the xi (i = 1, ..., n) in the multilinear
model y = b00 + b10x1 + ... + bn0xn with the z-scores zi = (xi – μi)/σi , and
replacing y with y – μy.
In general, b1 is more useful for producing plots in which the coefficients
are to be displayed on the same scale, such as a ridge trace (a plot of
the regression coefficients as a function of the ridge parameter). b0 is
more useful for making predictions.
18-1218
ridge
Coefficient estimates for multiple linear regression models rely on
the independence of the model terms. When terms are correlated
and the columns of the design matrix X have an approximate linear
dependence, the matrix (XTX)–1 becomes close to singular. As a result,
the least-squares estimate
ˆ = ( X T X )−1 X T y
becomes highly sensitive to random errors in the observed response y,
producing a large variance. This situation of multicollinearity can arise,
for example, when data are collected without an experimental design.
Ridge regression addresses the problem by estimating regression
coefficients using
ˆ = ( X T X + kI )−1 X T y
where k is the ridge parameter and I is the identity matrix. Small
positive values of k improve the conditioning of the problem and reduce
the variance of the estimates. While biased, the reduced variance of
ridge estimates often result in a smaller mean square error when
compared to least-squares estimates.
Examples Load the data in acetylene.mat, with observations of the predictor
variables x1, x2, x3, and the response variable y:
load acetylene
Plot the predictor variables against each other:
subplot(1,3,1)
plot(x1,x2,'.')
xlabel('x1'); ylabel('x2'); grid on; axis square
subplot(1,3,2)
plot(x1,x3,'.')
xlabel('x1'); ylabel('x3'); grid on; axis square
18-1219
ridge
subplot(1,3,3)
plot(x2,x3,'.')
xlabel('x2'); ylabel('x3'); grid on; axis square
Note the correlation between x1 and the other two predictor variables.
Use ridge and x2fx to compute coefficient estimates for a multilinear
model with interaction terms, for a range of ridge parameters:
X = [x1 x2 x3];
D = x2fx(X,'interaction');
D(:,1) = []; % No constant term
k = 0:1e-5:5e-3;
b = ridge(y,D,k);
Plot the ridge trace:
figure
plot(k,b,'LineWidth',2)
ylim([-100 100])
grid on
xlabel('Ridge Parameter')
ylabel('Standardized Coefficient')
title('{\bf Ridge Trace}')
legend('x1','x2','x3','x1x2','x1x3','x2x3')
18-1220
ridge
The estimates stabilize to the right of the plot. Note that the coefficient
of the x2x3 interaction term changes sign at a value of the ridge
parameter ≈ 5 × 10–4.
18-1221
ridge
References [1] Hoerl, A. E., and R. W. Kennard. “Ridge Regression: Biased
Estimation for Nonorthogonal Problems.” Technometrics. Vol. 12, No.
1, 1970, pp. 55–67.
[2] Hoerl, A. E., and R. W. Kennard. “Ridge Regression: Applications
to Nonorthogonal Problems.” Technometrics. Vol. 12, No. 1, 1970, pp.
69–82.
[3] Marquardt, D.W. “Generalized Inverses, Ridge Regression, Biased
Linear Estimation, and Nonlinear Estimation.” Technometrics. Vol. 12,
No. 3, 1970, pp. 591–612.
[4] Marquardt, D. W., and R.D. Snee. “Ridge Regression in Practice.”
The American Statistician. Vol. 29, No. 1, 1975, pp. 3–20.
See Also regress, stepwise
18-1222
classregtree.risk
Purpose Node risks
Syntax r = risk(t)
r = risk(t,nodes)
Description r = risk(t) returns an n-element vector r of the risk of the nodes
in the tree t, where n is the number of nodes. The risk r(i) for node
i is the node error e(i) (computed by nodeerr) weighted by the node
probability p(i) (computed by nodeprob).
r = risk(t,nodes) takes a vector nodes of node numbers and returns
the risk values for the specified nodes.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-1223
classregtree.risk
e = nodeerr(t);
p = nodeprob(t);
r = risk(t);
r
r =
0.6667
0
0.3333
0.0333
18-1224
classregtree.risk
0.0067
0.0067
0.0133
0
0
e.*p
ans =
0.6667
0
0.3333
0.0333
0.0067
0.0067
0.0133
0
0
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, nodeerr, nodeprob
18-1225
robustdemo
Purpose Interactive robust regression
Syntax robustdemo
robustdemo(x,y)
Description robustdemo shows the difference between ordinary least squares and
robust regression for data with a single predictor. With no input
arguments, robustdemo displays a scatter plot of a sample of roughly
linear data with one outlier. The bottom of the figure displays equations
of lines fitted to the data using ordinary least squares and robust
methods, together with estimates of the root mean squared errors.
Use the right mouse button to click on a point and view its least-squares
leverage and robust weight.
Use the left mouse button to click-and-drag a point. The displays will
update.
robustdemo(x,y) uses x and y data vectors you supply, in place of the
sample data supplied with the function.
Examples The following steps show you how to use robustdemo.
1 Start the demo. To begin using robustdemo with the built-in data,
simply type the function name:
robustdemo
18-1226
robustdemo
The resulting figure shows a scatter plot with two fitted lines. The
red line is the fit using ordinary least-squares regression. The green
line is the fit using robust regression. At the bottom of the figure are
the equations for the fitted lines, together with the estimated root
mean squared errors for each fit.
2 View leverages and robust weights. Right-click on any data point
to see its least-squares leverage and robust weight:
18-1227
robustdemo
In the built-in data, the right-most point has a relatively high
leverage of 0.35. The point exerts a large influence on the
least-squares fit, but its small robust weight shows that it is
effectively excluded from the robust fit.
3 See how changes in the data affect the fits. With the left mouse
button, click and hold on any data point and drag it to a new location.
When you release the mouse button, the displays update:
18-1228
robustdemo
Bringing the right-most data point closer to the least-squares line
makes the two fitted lines nearly identical. The adjusted right-most
data point has significant weight in the robust fit.
See Also robustfit, leverage
18-1229
robustfit
Purpose Robust regression
Syntax b = robustfit(X,y)
b = robustfit(X,y,wfun,tune)
b = robustfit(X,y,wfun,tune,const)
[b,stats] = robustfit(...)
Description b = robustfit(X,y) returns a p-by-1 vector b of coefficient estimates for
a robust multilinear regression of the responses in y on the predictors
in X. X is an n-by-p matrix of p predictors at each of n observations. y is
an n-by-1 vector of observed responses. By default, the algorithm uses
iteratively reweighted least squares with a bisquare weighting function.
Note By default, robustfit adds a first column of 1s to X,
corresponding to a constant term in the model. Do not enter a column of
1s directly into X. You can change the default behavior of robustfit
using the input const, below.
robustfit treats NaNs in X or y as missing values, and removes them.
b = robustfit(X,y,wfun,tune) specifies a weighting function wfun.
tune is a tuning constant that is divided into the residual vector before
computing weights.
The weighting function wfun can be any one of the following strings:
Default
Weight Tuning
Function Equation Constant
'andrews' w = (abs(r)<pi) .* sin(r) ./ r 1.339
'bisquare' w = (abs(r)<1) .* (1 - 4.685
(default) r.^2).^2
'cauchy' w = 1 ./ (1 + r.^2) 2.385
18-1230
robustfit
Default
Weight Tuning
Function Equation Constant
'fair' w = 1 ./ (1 + abs(r)) 1.400
'huber' w = 1 ./ max(1, abs(r)) 1.345
'logistic' w = tanh(r) ./ r 1.205
'ols' Ordinary least squares (no None
weighting function)
'talwar' w = 1 * (abs(r)<1) 2.795
'welsch' w = exp(-(r.^2)) 2.985
If tune is unspecified, the default value in the table is used. Default
tuning constants give coefficient estimates that are approximately
95% as statistically efficient as the ordinary least-squares estimates,
provided the response has a normal distribution with no outliers.
Decreasing the tuning constant increases the downweight assigned
to large residuals; increasing the tuning constant decreases the
downweight assigned to large residuals.
The value r in the weight functions is
r = resid/(tune*s*sqrt(1-h))
where resid is the vector of residuals from the previous iteration, h
is the vector of leverage values from a least-squares fit, and s is an
estimate of the standard deviation of the error term given by
s = MAD/0.6745
Here MAD is the median absolute deviation of the residuals from their
median. The constant 0.6745 makes the estimate unbiased for the
normal distribution. If there are p columns in X, the smallest p absolute
deviations are excluded when computing the median.
You can write your own weight function. The function must take a
vector of scaled residuals as input and produce a vector of weights as
18-1231
robustfit
output. In this case, wfun is specified using a function handle @ (as in
@myfun), and the input tune is required.
b = robustfit(X,y,wfun,tune,const) controls whether or not
the model will include a constant term. const is 'on' to include
the constant term (the default), or 'off' to omit it. When const is
'on', robustfit adds a first column of 1s to X. When const is 'off',
robustfit does not alter X.
[b,stats] = robustfit(...) returns the structure stats, whose
fields contain diagnostic statistics from the regression. The fields of
stats are:
• ols_s — Sigma estimate (RMSE) from ordinary least squares
• robust_s — Robust estimate of sigma
• mad_s — Estimate of sigma computed using the median absolute
deviation of the residuals from their median; used for scaling
residuals during iterative fitting
• s — Final estimate of sigma, the larger of robust_s and a weighted
average of ols_s and robust_s
• resid — Residual
• rstud — Studentized residual (see regress for more information)
• se — Standard error of coefficient estimates
• covb — Estimated covariance matrix for coefficient estimates
• coeffcorr — Estimated correlation of coefficient estimates
• t — Ratio of b to se
• p — p-values for t
• w — Vector of weights for robust fit
• R — R factor in QR decomposition of X
• dfe — Degrees of freedom for error
• h — Vector of leverage values for least-squares fit
18-1232
robustfit
The robustfit function estimates the variance-covariance matrix of
the coefficient estimates using inv(X'*X)*stats.s^2. Standard errors
and correlations are derived from this estimate.
Examples Generate data with the trend y = 10-2*x, then change one value to
simulate an outlier:
x = (1:10)';
y = 10 - 2*x + randn(10,1);
y(10) = 0;
Use both ordinary least squares and robust regression to estimate a
straight-line fit:
bls = regress(y,[ones(10,1) x])
bls =
7.2481
-1.3208
brob = robustfit(x,y)
brob =
9.1063
-1.8231
A scatter plot of the data together with the fits shows that the robust fit
is less influenced by the outlier than the least-squares fit:
scatter(x,y,'filled'); grid on; hold on
plot(x,bls(1)+bls(2)*x,'r','LineWidth',2);
plot(x,brob(1)+brob(2)*x,'g','LineWidth',2)
legend('Data','Ordinary Least Squares','Robust Regression')
18-1233
robustfit
References [1] DuMouchel, W. H., and F. L. O’Brien. “Integrating a Robust Option
into a Multiple Regression Computing Environment.” Computer Science
and Statistics: Proceedings of the 21st Symposium on the Interface.
Alexandria, VA: American Statistical Association, 1989.
[2] Holland, P. W., and R. E. Welsch. “Robust Regression Using
Iteratively Reweighted Least-Squares.” Communications in Statistics:
Theory and Methods, A6, 1977, pp. 813–827.
[3] Huber, P. J. Robust Statistics. Hoboken, NJ: John Wiley & Sons,
Inc., 1981.
18-1234
robustfit
[4] Street, J. O., R. J. Carroll, and D. Ruppert. “A Note on Computing
Robust Regression Estimates via Iteratively Reweighted Least
Squares.” The American Statistician. Vol. 42, 1988, pp. 152–154.
See Also regress, robustdemo
18-1235
categorical.rot90
Purpose Rotate categorical matrix 90 degrees
Syntax B = rot90(A)
B = rot90(A,k)
Description B = rot90(A) returns the 90 degree counterclockwise rotation of the
2-D categorical matrix A.
B = rot90(A,k) returns the k*90 degree rotation of A, k =
+-1,+-2,....
See Also flipdim, fliplr, flipud
18-1236
rotatefactors
Purpose Rotate factor loadings
Syntax B = rotatefactors(A)
B = rotatefactors(A,'Method','orthomax','Coeff',gamma)
B = rotatefactors(A,'Method','procrustes','Target',target)
B = rotatefactors(A,'Method','pattern','Target',target)
B = rotatefactors(A,'Method','promax')
[B,T] = rotatefactors(A,...)
Description B = rotatefactors(A) rotates the d-by-m loadings matrix A to
maximize the varimax criterion, and returns the result in B. Rows of
A and B correspond to variables and columns correspond to factors, for
example, the (i, j)th element of A is the coefficient for the i th variable
on the j th factor. The matrix A usually contains principal component
coefficients created with princomp or pcacov, or factor loadings
estimated with factoran.
B = rotatefactors(A,'Method','orthomax','Coeff',gamma)
rotates A to maximize the orthomax criterion with the coefficient gamma,
i.e., B is the orthogonal rotation of A that maximizes
sum(D*sum(B.^4,1) - GAMMA*sum(B.^2,1).^2)
The default value of 1 for gamma corresponds to varimax rotation.
Other possibilities include gamma = 0, m/2, and d(m - 1)/(d + m - 2),
corresponding to quartimax, equamax, and parsimax. You can also
supply the strings 'varimax', 'quartimax', 'equamax', or 'parsimax'
for the 'method' parameter and omit the 'Coeff' parameter.
If 'Method' is 'orthomax', 'varimax', 'quartimax', 'equamax', or
'parsimax', then additional parameters are
• 'Normalize' — Flag indicating whether the loadings matrix should
be row-normalized for rotation. If 'on' (the default), rows of A
are normalized prior to rotation to have unit Euclidean norm, and
unnormalized after rotation. If 'off', the raw loadings are rotated
and returned.
18-1237
rotatefactors
• 'Reltol' — Relative convergence tolerance in the iterative
algorithm used to find T. The default is sqrt(eps).
• 'Maxit' — Iteration limit in the iterative algorithm used to find T.
The default is 250.
B = rotatefactors(A,'Method','procrustes','Target',target)
performs an oblique procrustes rotation of A to the d-by-m target
loadings matrix target.
B = rotatefactors(A,'Method','pattern','Target',target)
performs an oblique rotation of the loadings matrix A to the d-by-m
target pattern matrix target, and returns the result in B. target
defines the "restricted" elements of B, i.e., elements of B corresponding
to zero elements of target are constrained to have small magnitude,
while elements of B corresponding to nonzero elements of target are
allowed to take on any magnitude.
If 'Method' is 'procrustes' or 'pattern', an additional parameter is
'Type', the type of rotation. If 'Type' is 'orthogonal', the rotation
is orthogonal, and the factors remain uncorrelated. If 'Type' is
'oblique' (the default), the rotation is oblique, and the rotated factors
might be correlated.
When 'Method' is 'pattern', there are restrictions on target. If A has
m columns, then for orthogonal rotation, the jth column of target must
contain at least m - j zeros. For oblique rotation, each column of target
must contain at least m - 1 zeros.
B = rotatefactors(A,'Method','promax') rotates A to maximize
the promax criterion, equivalent to an oblique Procrustes rotation
with a target created by an orthomax rotation. Use the four orthomax
parameters to control the orthomax rotation used internally by promax.
An additional parameter for ’promax’ is 'Power', the exponent for
creating promax target matrix. 'Power' must be 1 or greater. The
default is 4.
[B,T] = rotatefactors(A,...) returns the rotation matrix T used to
create B, that is, B = A*T. inv(T'*T) is the correlation matrix of the
18-1238
rotatefactors
rotated factors. For orthogonal rotation, this is the identity matrix,
while for oblique rotation, it has unit diagonal elements but nonzero
off-diagonal elements.
Examples X = randn(100,10);
% Default (normalized varimax) rotation:
% first three principle components.
LPC = princomp(X);
[L1,T] = rotatefactors(LPC(:,1:3));
% Equamax rotation:
% first three principle components.
[L2,T] = rotatefactors(LPC(:,1:3),...
'method','equamax');
% Promax rotation:
% first three factors.
LFA = factoran(X,3,'Rotate','none');
[L3,T] = rotatefactors(LFA(:,1:3),...
'method','promax',...
'power',2);
% Pattern rotation:
% first three factors.
Tgt = [1 1 1 1 1 0 1 0 1 1; ...
0 0 0 1 1 1 0 0 0 0; ...
1 0 0 1 0 1 1 1 1 0]';
[L4,T] = rotatefactors(LFA(:,1:3),...
'method','pattern',...
'target',Tgt);
inv(T'*T) % Correlation matrix of the rotated factors
References [1] Harman, H. H. Modern Factor Analysis. 3rd ed. Chicago: University
of Chicago Press, 1976.
18-1239
rotatefactors
[2] Lawley, D. N., and A. E. Maxwell. Factor Analysis as a Statistical
Method. 2nd ed. New York: American Elsevier Publishing, 1971.
See Also biplot, factoran, princomp, pcacov, procrustes
18-1240
rowexch
Purpose Row exchange
Syntax dRE = rowexch(nfactors,nruns)
[dRE,X] = rowexch(nfactors,nruns)
[dRE,X] = rowexch(nfactors,nruns,model)
[dRE,X] = rowexch(...,param1,val1,param2,val2,...)
Description dRE = rowexch(nfactors,nruns) uses a row-exchange algorithm to
generate a D-optimal design dRE with nruns runs (the rows of dRE) for a
linear additive model with nfactors factors (the columns of dRE). The
model includes a constant term.
[dRE,X] = rowexch(nfactors,nruns) also returns the associated
design matrix X, whose columns are the model terms evaluated at each
treatment (row) of dRE.
[dRE,X] = rowexch(nfactors,nruns,model) uses the linear
regression model specified in model. model is one of the following
strings:
• 'linear' — Constant and linear terms. This is the default.
• 'interaction' — Constant, linear, and interaction terms
• 'quadratic' — Constant, linear, interaction, and squared terms
• 'purequadratic' — Constant, linear, and squared terms
The order of the columns of X for a full quadratic model with n terms is:
1 The constant term
2 The linear terms in order 1, 2, ..., n
3 The interaction terms in order (1, 2), (1, 3), ..., (1, n), (2, 3), ..., (n–1, n)
4 The squared terms in order 1, 2, ..., n
Other models use a subset of these terms, in the same order.
18-1241
rowexch
Alternatively, model can be a matrix specifying polynomial terms of
arbitrary order. In this case, model should have one column for each
factor and one row for each term in the model. The entries in any row
of model are powers for the factors in the columns. For example, if a
model has factors X1, X2, and X3, then a row [0 1 2] in model specifies
the term (X1.^0).*(X2.^1).*(X3.^2). A row of all zeros in model
specifies a constant term, which can be omitted.
[dRE,X] = rowexch(...,param1,val1,param2,val2,...) specifies
additional parameter/value pairs for the design. Valid parameters and
their values are listed in the following table.
Parameter Value
'bounds' Lower and upper bounds for each factor, specified as
a 2-by-nfactors matrix. Alternatively, this value
can be a cell array containing nfactors elements,
each element specifying the vector of allowable
values for the corresponding factor.
'categorical' Indices of categorical predictors.
'display' Either 'on' or 'off' to control display of the
iteration counter. The default is 'on'.
'excludefun' Handle to a function that excludes undesirable
runs. If the function is f, it must support the syntax
b = f(S), where S is a matrix of treatments with
nfactors columns and b is a vector of Boolean
values with the same number of rows as S. b(i) is
true if the ith row S should be excluded.
'init' Initial design as an nruns-by-nfactors matrix. The
default is a randomly selected set of points.
'levels' Vector of number of levels for each factor.
18-1242
rowexch
Parameter Value
'maxiter' Maximum number of iterations. The default is 10.
'tries' Number of times to try to generate a design from
a new starting point. The algorithm uses random
points for each try, except possibly the first. The
default is 1.
Algorithm Both cordexch and rowexch use iterative search algorithms. They
operate by incrementally changing an initial design matrix X to increase
D = |XTX| at each step. In both algorithms, there is randomness
built into the selection of the initial design and into the choice of the
incremental changes. As a result, both algorithms may return locally,
but not globally, D-optimal designs. Run each algorithm multiple times
and select the best result for your final design. Both functions have a
'tries' parameter that automates this repetition and comparison.
At each step, the row-exchange algorithm exchanges an entire row of
X with a row from a design matrix C evaluated at a candidate set of
feasible treatments. The rowexch function automatically generates a C
appropriate for a specified model, operating in two steps by calling the
candgen and candexch functions in sequence. Provide your own C by
calling candexch directly. In either case, if C is large, its static presence
in memory can affect computation.
Examples Suppose you want a design to estimate the parameters in the following
three-factor, seven-term interaction model:
y = 0 + 1 x 1 + 2 x 2 + 3 x 3 + 12 x 1 x 2 + 13 x 1 x 3 + 23 x 2 x 3 +
Use rowexch to generate a D-optimal design with seven runs:
nfactors = 3;
nruns = 7;
[dRE,X] = rowexch(nfactors,nruns,'interaction','tries',10)
dRE =
-1 -1 1
18-1243
rowexch
1 -1 1
1 -1 -1
1 1 1
-1 -1 -1
-1 1 -1
-1 1 1
X =
1 -1 -1 1 1 -1 -1
1 1 -1 1 -1 1 -1
1 1 -1 -1 -1 -1 1
1 1 1 1 1 1 1
1 -1 -1 -1 1 1 1
1 -1 1 -1 -1 1 -1
1 -1 1 1 -1 -1 1
Columns of the design matrix X are the model terms evaluated at each
row of the design dRE. The terms appear in order from left to right:
constant term, linear terms (1, 2, 3), interaction terms (12, 13, 23). Use
X to fit the model, as described in “Linear Regression” on page 9-3, to
response data measured at the design points in dRE.
See Also candgen, candexch, cordexch
18-1244
rsmdemo
Purpose Interactive response surface demonstration
Syntax rsmdemo
Description rsmdemo opens a group of three graphical user interfaces for
interactively investigating response surface methodology (RSM),
nonlinear fitting, and the design of experiments.
The interfaces allow you to collect and model data from a simulated
chemical reaction. Experimental predictors are concentrations of three
reactants (hydrogen, n-Pentane, and isopentane) and the response is
the reaction rate. The reaction rate is simulated by a Hougen-Watson
model (Bates and Watts, [2], pp. 271–272):
1 x2 − x3 / 5
rate =
1 + 2 x1 + 3 x2 + 4 x3
where rate is the reaction rate, x1, x2, and x3 are the concentrations of
hydrogen, n-pentane, and isopentane, respectively, and β1, β2, ... , β5 are
fixed parameters. Random errors are used to perturb the reaction rate
for each combination of reactants.
Collect data using one of two methods:
1 Manually set reactant concentrations in the Reaction Simulator
interface by editing the text boxes or by adjusting the associated
sliders.
18-1245
rsmdemo
When you click Run, the concentrations and simulated reaction rate
are recorded on the Trial and Error Data interface.
18-1246
rsmdemo
You are allowed up to 13 independent experimental runs for data
collection.
2 Use a designed experiment to set reactant concentrations in the
Experimental Data interface by clicking the Do Experiment
button.
A 13-run D-optimal design for a full quadratic model is generated
by the cordexch function, and the concentrations and simulated
reaction rates are recorded on the same interface.
18-1247
rsmdemo
Once data is collected, scatter plots of reaction rates vs. individual
predictors are generated by selecting one of the following from the Plot
pop-up menu below the recorded data:
• Hydrogen vs. Rate
• n-Pentane vs. Rate
• Isopentane vs. Rate
Fit a response surface model to the data by clicking the Analyze button
below the trial-and-error data or the Response Surface button below
the experimental data. Both buttons load the data into the Response
Surface Tool rstool. By default, trial-and-error data is fit with a linear
additive model and experimental data is fit with a full quadratic model,
but the models can be adjusted in the Response Surface Tool.
For experimental data, you have the additional option of fitting a
Hougen-Watson model. Click the Nonlinear Model button to load the
data and the model in hougen into the Nonlinear Fitting Tool nlintool.
18-1248
rsmdemo
See Also hougen, cordexch, rstool, nlintool
18-1249
rstool
Purpose Interactive response surface modeling
Syntax rstool
rstool(X,Y,model)
rstool(x,y,model,alpha)
rstool(x,y,model,alpha,xname,yname)
Description rstool opens a graphical user interface for interactively investigating
one-dimensional contours of multidimensional response surface models.
By default, the interface opens with the data from hald.mat and a fitted
response surface with constant, linear, and interaction terms.
A sequence of plots is displayed, each showing a contour of the response
surface against a single predictor, with all other predictors held fixed.
18-1250
rstool
95% global confidence intervals for new observations are shown as
dashed red curves above and below the response. Predictor values are
displayed in the text boxes on the horizontal axis and are marked by
vertical dashed blue lines in the plots. Predictor values are changed by
editing the text boxes or by dragging the dashed blue lines. When you
change the value of a predictor, all plots update to show the new point
in predictor space.
The pop-up menu at the lower left of the interface allows you to choose
among the following models:
• Linear — Constant and linear terms (the default)
• Pure Quadratic — Constant, linear, and squared terms
• Interactions — Constant, linear, and interaction terms
• Full Quadratic — Constant, linear, interaction, and squared terms
Click Export to open the following dialog box:
The dialog allows you to save information about the fit to MATLAB
workspace variables with valid names.
rstool(X,Y,model) opens the interface with the predictor data
in X, the response data in Y, and the fitted model model. Distinct
predictor variables should appear in different columns of X. Y can be a
vector, corresponding to a single response, or a matrix, with columns
18-1251
rstool
corresponding to multiple responses. Y must have as many elements (or
rows, if it is a matrix) as X has rows.
The optional input model can be any one of the following strings:
• 'linear' — Constant and linear terms (the default)
• 'purequadratic' — Constant, linear, and squared terms
• 'interaction' — Constant, linear, and interaction terms
• 'quadratic' — Constant, linear, interaction, and squared terms
To specify a polynomial model of arbitrary order, or a model without a
constant term, use a matrix for model as described in x2fx.
rstool(x,y,model,alpha) uses 100(1-alpha)% global confidence
intervals for new observations in the plots.
rstool(x,y,model,alpha,xname,yname) labels the axes using the
strings in xname and yname. To label each subplot differently, xname and
yname can be cell arrays of strings.
Examples The following uses rstool to visualize a quadratic response surface
model of the 3-D chemical reaction data in reaction.mat:
load reaction
alpha = 0.01; % Significance level
rstool(reactants,rate,'quadratic',alpha,xn,yn)
18-1252
rstool
The rstool interface is used by rsmdemo to visualize the results of
simulated experiments with data like that in reaction.mat. As
described in “Response Surface Designs” on page 14-9, rsmdemo uses a
response surface model to generate simulated data at combinations of
predictors specified by either the user or by a designed experiment.
See Also x2fx, rsmdemo, nlintool
18-1253
runstest
Purpose Run test for randomness
Syntax h = runstest(x)
h = runstest(x,v)
h = runstest(x,'ud')
h = runstest(...,param1,val1,param2,val2,...)
[h,p] = runstest(...)
[h,p,stats] = runstest(...)
Description h = runstest(x) performs a runs test on the sequence of observations
in the vector x. This is a test of the null hypothesis that the values in
x come in random order, against the alternative that they do not. The
test is based on the number of runs of consecutive values above or below
the mean of x. Too few runs indicate a tendency for high and low values
to cluster. Too many runs indicate a tendency for high and low values
to alternate. The test returns the logical value h = 1 if it rejects the null
hypothesis at the 5% significance level, and h = 0 if it cannot. The test
treats NaN values in x as missing values, and ignores them.
runstest uses a test statistic which is approximately normally
distributed when the null hypothesis is true. It is the difference between
the number of runs and its mean, divided by its standard deviation.
h = runstest(x,v) performs the test using runs above or below the
value v. Values exactly equal to v are discarded.
h = runstest(x,'ud') performs a test for the number of runs up
or down. This also tests the hypothesis that the values in x come in
random order. Too few runs indicate a trend. Too many runs indicate an
oscillation. Values exactly equal to the preceding value are discarded.
h = runstest(...,param1,val1,param2,val2,...) specifies
additional parameters and their values. Valid parameter/value pairs
are the following:
• 'alpha' — A scalar giving the significance level of the test
• 'method' — Either 'exact' to compute the p value using an exact
algorithm, or 'approximate' to use a normal approximation. The
18-1254
runstest
default is 'exact' for runs above/below, and for runs up/down when
the length of x is 50 or less. The 'exact' method is not available for
runs up/down when the length of x is 51 or greater.
• 'tail' — Performs the test against one of the following alternative
hypotheses:
- 'both' — two-tailed test (sequence is not random)
- 'right' — right-tailed test (like values separate for runs
above/below, direction alternates for runs up/down)
- 'left' — left-tailed test (like values cluster for runs above/below,
values trend for runs up/down)
[h,p] = runstest(...) returns the p value of the test. The output p
is computed from either the test statistic or the exact distribution of the
number of runs, depending on the value of the 'method' parameter.
[h,p,stats] = runstest(...) returns a structure stats with the
following fields:
• nruns — The number of runs
• n1 — The number of values above v
• n0 — The number of values below v
• z — The test statistic
Examples x = randn(40,1);
[h,p] = runstest(x,median(x))
h =
0
p =
0.6286
See Also signrank, signtest
18-1255
TreeBagger.SampleWithReplacement property
Purpose Flag to sample with replacement
Description The SampleWithReplacement property is a logical flag specifying if
data are sampled for each decision tree with replacement. True if
TreeBagger samples data with replacement and false otherwise. True
by default.
18-1256
sampsizepwr
Purpose Sample size and power of test
Syntax n = sampsizepwr(testtype,p0,p1)
n = sampsizepwr(testtype,p0,p1,power)
power = sampsizepwr(testtype,p0,p1,[],n)
p1 = sampsizepwr(testtype,p0,[],power,n)
[...] = sampsizepwr(...,n,param1,val1,param2,val2,...)
Description n = sampsizepwr(testtype,p0,p1) returns the sample size n required
for a two-sided test of the specified type to have a power (probability
of rejecting the null hypothesis when the alternative hypothesis is
true) of 0.90 when the significance level (probability of rejecting the
null hypothesis when the null hypothesis is true) is 0.05. p0 specifies
parameter values under the null hypothesis. p1 specifies the single
parameter value being tested under the alternative hypothesis.
The following values are available for testtype:
• 'z' — z-test for normally distributed data with known standard
deviation. p0 is a two-element vector [mu0 sigma0] of the mean and
standard deviation, respectively, under the null hypothesis. p1 is the
value of the mean under the alternative hypothesis.
• 't' — t-test for normally distributed data with unknown standard
deviation. p0 is a two-element vector [mu0 sigma0] of the mean and
standard deviation, respectively, under the null hypothesis. p1 is the
value of the mean under the alternative hypothesis.
• 'var' — Chi-square test of variance for normally distributed data.
p0 is the variance under the null hypothesis. p1 is the variance under
the alternative hypothesis.
• 'p' — Test of the p parameter (success probability) for a binomial
distribution. p0 is the value of p under the null hypothesis. p1 is the
value of p under the alternative hypothesis.
The 'p' test is a discrete test for which increasing the sample size
does not always increase the power. For n values larger than 200,
18-1257
sampsizepwr
there may be values smaller than the returned n value that also
produce the desired size and power.
n = sampsizepwr(testtype,p0,p1,power) returns the sample size n
such that the power is power for the parameter value p1.
power = sampsizepwr(testtype,p0,p1,[],n) returns the power
achieved for a sample size of n when the true parameter value is p1.
p1 = sampsizepwr(testtype,p0,[],power,n) returns the parameter
value detectable with the specified sample size n and power power.
When computing p1 for the 'p' test, if no alternative can be rejected for
a given null hypothesis and significance level, the function displays a
warning message and returns NaN.
[...] = sampsizepwr(...,n,param1,val1,param2,val2,...)
specifies one or more of the following name/value pairs:
• 'alpha' — Significance level of the test (default 0.05)
• 'tail' — The type of test is one of the following:
- 'both' — Two-sided test for an alternative not equal to p0
- 'right' — One-sided test for an alternative larger than p0
- 'left' — One-sided test for an alternative smaller than p0
Examples Compute the mean closest to 100 that can be determined to be
significantly different from 100 using a t-test with a sample size of 60
and a power of 0.8.
mu1 = sampsizepwr('t',[100 10],[],0.8,60)
mu1 =
103.6770
Compute the sample size n required to distinguish p = 0.26 from p = 0.6
with a binomial test. The result is approximate, so make a plot to see if
any smaller n values also have the required power of 0.5.
18-1258
sampsizepwr
napprox = sampsizepwr('p',0.2,0.26,0.6)
Warning: Values N>200 are approximate. Plotting the power as a function
of N may reveal lower N values that have the required power.
napprox =
244
nn = 1:250;
pwr = sampsizepwr('p',0.2,0.26,[],nn);
nexact = min(nn(pwr>=0.6))
nexact =
213
plot(nn,pwr,'b-',[napprox nexact],pwr([napprox nexact]),'ro');
grid on
18-1259
sampsizepwr
See Also vartest, ttest, ztest, binocdf
18-1260
scatterhist
Purpose Scatter plot with marginal histograms
Syntax scatterhist(x,y)
scatterhist(x,y,nbins)
h = scatterhist(...)
Description scatterhist(x,y) creates a 2-D scatterplot of the data in the vectors x
and y, and puts a univariate histogram on the horizontal and vertical
axes of the plot. x and y must be the same length.
The function is useful for viewing properties of random samples
produced by functions such as copularnd, mvnrnd, lhsdesign.
scatterhist(x,y,nbins) also accepts a two-element vector nbins
specifying the number of bins for the x and y histograms. The default is
to compute the number of bins using a Scott rule based on the sample
standard deviation. Any NaN values in either x or y are treated as
missing, and are removed from both x and y. Therefore the plots reflect
points for which neither x nor y has a missing value.
h = scatterhist(...) returns a vector of three axes handles for the
scatterplot, the histogram along the horizontal axis, and the histogram
along the vertical axis, respectively.
Examples Example 1
Independent normal and lognormal random samples:
x = randn(1000,1);
y = exp(.5*randn(1000,1));
scatterhist(x,y)
18-1261
scatterhist
Example 2
Marginal uniform samples that are not independent:
u = copularnd('Gaussian',.8,1000);
scatterhist(u(:,1),u(:,2))
18-1262
scatterhist
Example 3
Mixed discrete and continuous data:
cars = load('carsmall');
scatterhist(cars.Weight,cars.Cylinders,[10 3])
18-1263
scatterhist
See Also scatter, hist
18-1264
qrandset.scramble
Purpose Scramble quasi-random point set
Syntax ps = scramble(p,type)
ps = scramble(p,'clear')
ps = scramble(p)
Description ps = scramble(p,type) returns a scrambled copy ps of the point set p
of the qrandset class, created using the scramble type specified in the
string type. Point sets from different subclasses of qrandset support
different scramble types, as indicated in the following table.
Subclass Scramble Types
haltonset 'RR2' — A permutation of the radical inverse
coefficients derived by applying a reverse-radix
operation to all of the possible coefficient values.
The scramble is described in [1].
sobolset 'MatousekAffineOwen' — A random linear
scramble combined with a random digital shift.
The scramble is described in [2].
ps = scramble(p,'clear') removes all scramble settings from p and
returns the result in ps.
ps = scramble(p) removes all scramble settings from p and then adds
them back in the order they were originally applied. This typically
results in a different point set because of the randomness of the
scrambling algorithms.
Examples Use haltonset to generate a 3-D Halton point set, skip the first 1000
values, and then retain every 101st point:
p = haltonset(3,'Skip',1e3,'Leap',1e2)
p =
Halton point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
18-1265
qrandset.scramble
Leap : 100
ScrambleMethod : none
Use scramble to apply reverse-radix scrambling:
p = scramble(p,'RR2')
p =
Halton point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : RR2
Use net to generate the first four points:
X0 = net(p,4)
X0 =
0.0928 0.6950 0.0029
0.6958 0.2958 0.8269
0.3013 0.6497 0.4141
0.9087 0.7883 0.2166
Use parenthesis indexing to generate every third point, up to the 11th
point:
X = p(1:3:11,:)
X =
0.0928 0.6950 0.0029
0.9087 0.7883 0.2166
0.3843 0.9840 0.9878
0.6831 0.7357 0.7923
References [1] Kocis, L., and W. J. Whiten. “Computational Investigations of
Low-Discrepancy Sequences.” ACM Transactions on Mathematical
Software. Vol. 23, No. 2, 1997, pp. 266–294.
[2] Matousek, J. “On the L2-Discrepancy for Anchored Boxes.” Journal
of Complexity. Vol. 14, No. 4, 1998, pp. 527–556.
18-1266
qrandset.scramble
See Also haltonset, sobolset
18-1267
qrandset.ScrambleMethod property
Purpose Settings that control scrambling
Description The ScrambleMethod property contains a structure that defines which
scrambles to apply to the sequence. The structure consists of two fields:
• Type: A string containing the name of the scramble.
• Options: A cell array of parameter values for the scramble.
Different point sets support different scramble types as outlined in
the help for each point set class. An error occurs if you set an invalid
scramble type for a given point set.
The ScrambleMethod property also accepts an empty matrix as a value.
This will clear all scrambling and set the property to contain a (0x0)
structure.
The scramble method provides an alternative, easier way to set
scrambles.
Examples Apply a random linear scramble combined with a random digital shift
to a sobolset point set class:
P = sobolset(5);
P = scramble(P, 'MatousekAffineOwen');
P.ScrambleMethod
See Also sobolset | scramble
18-1268
piecewisedistribution.segment
Purpose Segments containing values
Syntax S = segment(obj,X,P)
Description S = segment(obj,X,P) returns an array S of integers indicating which
segment of the piecewise distribution object obj contains each value
of X or, alternatively, P. One of X and P must be empty ([]). If X is
nonempty, S is determined by comparing X with the quantile boundary
values defined for obj. If P is nonempty, S is determined by comparing
P with the probability boundary values.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
pvals = 0:0.2:1;
s = segment(obj,[],pvals)
s =
1 2 2 2 2 3
See Also paretotails, boundary, nsegments
18-1269
sequentialfs
Purpose Sequential feature selection
Syntax inmodel = sequentialfs(fun,X,y)
inmodel = sequentialfs(fun,X,Y,Z,...)
[inmodel,history] = sequentialfs(fun,X,...)
[] = sequentialfs(...,param1,val1,param2,val2,...)
Description inmodel = sequentialfs(fun,X,y) selects a subset of features from
the data matrix X that best predict the data in y by sequentially
selecting features until there is no improvement in prediction. Rows
of X correspond to observations; columns correspond to variables or
features. y is a column vector of response values or class labels for each
observation in X. X and y must have the same number of rows. fun is a
function handle to a function that defines the criterion used to select
features and to determine when to stop. The output inmodel is a logical
vector indicating which features are finally chosen.
Starting from an empty feature set, sequentialfs creates candidate
feature subsets by sequentially adding each of the features not yet
selected. For each candidate feature subset, sequentialfs performs
10-fold cross-validation by repeatedly calling fun with different training
subsets of X and y, XTRAIN and ytrain, and test subsets of X and y,
XTEST and ytest, as follows:
criterion = fun(XTRAIN,ytrain,XTEST,ytest)
XTRAIN and ytrain contain the same subset of rows of X and Y, while
XTEST and ytest contain the complementary subset of rows. XTRAIN
and XTEST contain the data taken from the columns of X that correspond
to the current candidate feature set.
Each time it is called, fun must return a scalar value criterion.
Typically, fun uses XTRAIN and ytrain to train or fit a model, then
predicts values for XTEST using that model, and finally returns some
measure of distance, or loss, of those predicted values from ytest.
In the cross-validation calculation for a given candidate feature set,
sequentialfs sums the values returned by fun and divides that sum
18-1270
sequentialfs
by the total number of test observations. It then uses that mean value
to evaluate each candidate feature subset.
Typical loss measures include sum of squared errors for regression
models (sequentialfs computes the mean-squared error in this case),
and the number of misclassified observations for classification models
(sequentialfs computes the misclassification rate in this case).
Note sequentialfs divides the sum of the values returned by
fun across all test sets by the total number of test observations.
Accordingly, fun should not divide its output value by the number of
test observations.
After computing the mean criterion values for each candidate
feature subset, sequentialfs chooses the candidate feature subset
that minimizes the mean criterion value. This process continues until
adding more features does not decrease the criterion.
inmodel = sequentialfs(fun,X,Y,Z,...) allows any number of
input variables X, Y, Z, ... . sequentialfs chooses features (columns)
only from X, but otherwise imposes no interpretation on X, Y, Z, ... . All
data inputs, whether column vectors or matrices, must have the same
number of rows. sequentialfs calls fun with training and test subsets
of X, Y, Z, ... as follows:
criterion = fun(XTRAIN,YTRAIN,ZTRAIN,...,
XTEST,YTEST,ZTEST,...)
sequentialfs creates XTRAIN, YTRAIN, ZTRAIN, ... , XTEST, YTEST,
ZTEST, ... by selecting subsets of the rows of X, Y, Z, ... . fun must return
a scalar value criterion, but may compute that value in any way.
Elements of the logical vector inmodel correspond to columns of X and
indicate which features are finally chosen.
[inmodel,history] = sequentialfs(fun,X,...) returns
information on which feature is chosen at each step. history is a scalar
structure with the following fields:
18-1271
sequentialfs
• Crit — A vector containing the criterion values computed at each
step.
• In — A logical matrix in which row i indicates the features selected
at step i.
[] = sequentialfs(...,param1,val1,param2,val2,...) specifies
optional parameter name/value pairs from the following table.
Parameter Value
'cv' The validation method used to compute the
criterion for each candidate feature subset.
• When the value is a positive integer k,
sequentialfs uses k-fold cross-validation
without stratification.
• When the value is an object of the cvpartition
class, other forms of cross-validation can be
specified.
• When the value is 'resubstitution', the
original data are passed to fun as both the
training and test data to compute the criterion.
• When the value is 'none', sequentialfs
calls fun as criterion = fun(X,Y,Z,...),
without separating test and training sets.
The default value is 10, that is, 10-fold
cross-validation without stratification.
So-called wrapper methods use a function fun
that implements a learning algorithm. These
methods usually apply cross-validation to select
features. So-called filter methods use a function
fun that measures characteristics of the data
(such as correlation) to select features.
18-1272
sequentialfs
Parameter Value
'mcreps' A positive integer indicating the number of
Monte-Carlo repetitions for cross-validation. The
default value is 1. The value must be 1 if the
value of 'cv' is 'resubstitution' or 'none'.
'direction' The direction of the sequential search. The
default is 'forward'. A value of 'backward'
specifies an initial candidate set including all
features and an algorithm that removes features
sequentially until the criterion increases.
'keepin' A logical vector or a vector of column numbers
specifying features that must be included. The
default is empty.
'keepout' A logical vector or a vector of column numbers
specifying features that must be excluded. The
default is empty.
'nfeatures' The number of features at which sequentialfs
should stop. inmodel includes exactly this
many features. The default value is empty,
indicating that sequentialfs should stop when
a local minimum of the criterion is found. A
nonempty value overrides values of 'MaxIter'
and 'TolFun' in 'options'.
'nullmodel' A logical value, indicating whether or not the null
model (containing no features from X) should be
included in feature selection and in the history
output. The default is false.
18-1273
sequentialfs
Parameter Value
'options' Options structure for the iterative sequential
search algorithm, as created by statset.
sequentialfs uses the following statset
parameters:
• Display — Amount of information displayed
by the algorithm. The default is 'off'.
• MaxIter — Maximum number of iterations
allowed. The default is Inf.
• TolFun — Termination tolerance for the
objective function value. The default is 1e-6 if
'direction' is 'forward'; 0 if 'direction'
is 'backward'.
• TolTypeFun — Use absolute or relative
objective function tolerances. The default is
'rel'.
Examples Perform sequential feature selection for classification of noisy features:
load fisheriris;
X = randn(150,10);
X(:,[1 3 5 7 ])= meas;
y = species;
c = cvpartition(y,'k',10);
opts = statset('display','iter');
fun = @(XT,yT,Xt,yt)...
(sum(~strcmp(yt,classify(Xt,XT,yT,'quadratic'))));
[fs,history] = sequentialfs(fun,X,y,'cv',c,'options',opts)
Start forward sequential feature selection:
Initial columns included: none
18-1274
sequentialfs
Columns that can not be included: none
Step 1, added column 7, criterion value 0.04
Step 2, added column 5, criterion value 0.0266667
Final columns included: 5 7
fs =
0 0 0 0 1 0 1 0 0 0
history =
In: [2x10 logical]
Crit: [0.0400 0.0267]
history.In
ans =
0 0 0 0 0 0 1 0 0 0
0 0 0 0 1 0 1 0 0 0
See Also crossval | cvpartition | stepwisefit | statset
Tutorials • “Example: Sequential Feature Selection” on page 10-24
How To • “Sequential Feature Selection” on page 10-23
18-1275
dataset.set
Purpose Set and display properties
Syntax set(A)
set(A,PropertyName)
A = set(A,PropertyName,PropertyValue,...)
B = set(A,PropertyName,value)
Description set(A) displays all properties of the dataset array A and their possible
values.
set(A,PropertyName) displays possible values for the property
specified by the string PropertyName.
A = set(A,PropertyName,PropertyValue,...) sets property
name/value pairs.
B = set(A,PropertyName,value) returns a dataset array B that is a
copy of A, but with the property 'PropertyName' set to the value value.
Note Using set(A,'PropertyName',value) without assigning
to a variable does not modify A’s properties. Use A =
set(A,'PropertyName',value) to modify A.
Examples Create a dataset array from Fisher’s iris data and add a description:
load fisheriris
NumObs = size(meas,1);
NameObs = strcat({'Obs'},num2str((1:NumObs)','%-d'));
iris = dataset({nominal(species),'species'},...
{meas,'SL','SW','PL','PW'},...
'ObsNames',NameObs);
iris = set(iris,'Description','Fisher''s Iris Data');
get(iris)
Description: 'Fisher's Iris Data'
Units: {}
DimNames: {'Observations' 'Variables'}
18-1276
dataset.set
UserData: []
ObsNames: {150x1 cell}
VarNames: {'species' 'SL' 'SW' 'PL' 'PW'}
See Also get, summary
18-1277
CompactTreeBagger.SetDefaultYfit
Purpose Set default value for predict
Syntax B = SetDefaultYfit(B,Yfit)
Description B = SetDefaultYfit(B,Yfit) sets the default prediction for ensemble
B to Yfit. The default prediction must be a character variable for
classification or a numeric scalar for regression. This setting controls
what predicted value CompactTreeBagger returns when no prediction
is possible, for example when the predict method needs to predict
for an observation which has only false values in the matrix supplied
through 'useifort' argument.
See Also DefaultYfit, predict, TreeBagger.DefaultYfit
18-1278
categorical.setdiff
Purpose Set difference for categorical arrays
Syntax C = setdiff(A,B)
[C,I] = setdiff(A,B)
Description C = setdiff(A,B) when A and B are categorical arrays returns a
categorical vector C containing the values in A that are not in B. The
result C is sorted. The set of categorical levels for C is the sorted union
of the sets of levels of the inputs, as determined by their labels.
[C,I] = setdiff(A,B) also returns index vectors I such that C =
A(I).
See Also intersect, ismember, setxor, union, unique
18-1279
categorical.setlabels
Purpose Label levels
Syntax A = setlabels(A,labels)
A = setlabels(A,labels,levels)
Description A = setlabels(A,labels) labels the levels in the categorical array A
using the cell array of strings or 2-D character matrix labels. Labels
are assigned in the order given in labels.
A = setlabels(A,labels,levels) labels only the levels specified in
the cell array of strings or 2-D character matrix levels.
Examples Example 1
Relabel the species in Fisher’s iris data using new categories:
load fisheriris
species = nominal(species);
species = mergelevels(...
species,{'setosa','virginica'},'parent');
species = setlabels(species,'hybrid','versicolor');
getlabels(species)
ans =
'hybrid' 'parent'
Example 2
1 Load patient data from the CSV file hospital.dat and store the
information in a dataset array with observation names given by the
first column in the data (patient identification):
patients = dataset('file','hospital.dat',...
'delimiter',',',...
'ReadObsNames',true);
2 Make the {0,1}-valued variable smoke nominal, and change the labels
to 'No' and 'Yes':
patients.smoke = nominal(patients.smoke,{'No','Yes'});
18-1280
categorical.setlabels
3 Add new levels to smoke as placeholders for more detailed histories
of smokers:
patients.smoke = addlevels(patients.smoke,...
{'0-5 Years','5-10 Years','LongTerm'});
4 Assuming the nonsmokers have never smoked, relabel the 'No' level:
patients.smoke = setlabels(patients.smoke,'Never','No');
5 Drop the undifferentiated 'Yes' level from smoke:
patients.smoke = droplevels(patients.smoke,'Yes');
Warning: OLDLEVELS contains categorical levels that
were present in A, caused some array elements to have
undefined levels.
Note that smokers now have an undefined level.
6 Set each smoker to one of the new levels, by observation name:
patients.smoke('YPL-320') = '5-10 Years';
See Also getlabels
18-1281
categorical.setxor
Purpose Set exclusive-or for categorical arrays
Syntax C = setxor(A,B)
[C,IA,IB] = setxor(A,B)
Description C = setxor(A,B) when A and B are categorical arrays returns a
categorical vector C containing the values not in the intersection of A and
B. The result C is sorted. The set of categorical levels for C is the sorted
union of the sets of levels of the inputs, as determined by their labels.
[C,IA,IB] = setxor(A,B) also returns index vectors IA and IB such
that C is a sorted combination of the elements A(IA) and B(IB).
See Also intersect, ismember, setdiff, union, unique
18-1282
gmdistribution.SharedCov property
Purpose true if all covariance matrices are restricted to be the same
Description Logical true if all the covariance matrices are restricted to be the same
(pooled estimate); logical false otherwise.
18-1283
categorical.shiftdim
Purpose Shift dimensions of categorical array
Syntax B = shiftdim(A,n)
[B,nshifts] = shiftdim(A)
Description B = shiftdim(A,n) shifts the dimensions of the categorical array A
by N. When n is positive, shiftdim shifts the dimensions to the left
and wraps the n leading dimensions to the end. When n is negative,
shiftdim shifts the dimensions to the right and pads with singletons.
[B,nshifts] = shiftdim(A) returns the array B with the same
number of elements as A but with any leading singleton dimensions
removed. nshifts returns the number of dimensions that are removed.
If A is a scalar, shiftdim has no effect.
See Also circshift, reshape, squeeze
18-1284
gmdistribution.Sigma property
Purpose Input array of covariances
Description Input array of covariances SIGMA.
18-1285
signrank
Purpose Wilcoxon signed rank test
Syntax p = signrank(x)
p = signrank(x,m)
p = signrank(x,y)
[p,h] = signrank(...)
[p,h] = signrank(...,'alpha',alpha)
[p,h] = signrank(...,'method',method)
[p,h,stats] = signrank(...)
Description p = signrank(x) performs a two-sided signed rank test of the
null hypothesis that data in the vector x comes from a continuous,
symmetric distribution with zero median, against the alternative that
the distribution does not have zero median. The p value of the test
is returned in p.
p = signrank(x,m) performs a two-sided signed rank test of the null
hypothesis that data in the vectors x and y are independent samples
from a continuous, symmetric distribution with median m, against the
alternative that the distribution does not have median m. m must be a
scalar.
p = signrank(x,y) performs a paired, two-sided signed rank test of
the null hypothesis that data in the vector x-y come from a continuous,
symmetric distribution with zero median, against the alternative that
the distribution does not have zero median. x and y must have equal
lengths. Note that a hypothesis of zero median for x-y is not equivalent
to a hypothesis of equal median for x and y.
[p,h] = signrank(...) returns the result of the test in h. h = 1
indicates a rejection of the null hypothesis at the 5% significance
level. h = 0 indicates a failure to reject the null hypothesis at the 5%
significance level.
[p,h] = signrank(...,'alpha',alpha) performs the test at the
(100*alpha)% significance level. The default, when unspecified, is
alpha = 0.05.
18-1286
signrank
[p,h] = signrank(...,'method',method) computes the p value
using either an exact algorithm, when method is 'exact', or a normal
approximation, when method is 'approximate'. The default, when
unspecified, is the exact method for small samples and the approximate
method for large samples.
[p,h,stats] = signrank(...) returns the structure stats with the
following fields:
• signedrank — Value of the signed rank test statistic
• zval — Value of the z-statistic (computed only for large samples)
Examples Test the hypothesis of zero median for the difference between two
paired samples.
before = lognrnd(2,.25,10,1);
after = before+trnd(2,10,1);
[p,h] = signrank(before,after)
p =
0.5566
h =
0
The sampling distribution of the difference between before and after
is symmetric with zero median. At the default 5% significance level, the
test fails to reject to the null hypothesis of zero median in the difference.
References [1] Gibbons, J. D. Nonparametric Statistical Inference. New York:
Marcel Dekker, 1985.
[2] Hollander, M., and D. A. Wolfe. Nonparametric Statistical Methods.
Hoboken, NJ: John Wiley & Sons, Inc., 1999.
See Also ranksum, ttest, ztest
18-1287
signtest
Purpose Sign test
Syntax p = signtest(x)
p = signtest(x,m)
p = signtest(x,y)
[p,h] = signtest(...)
[p,h] = signtest(...,'alpha',alpha)
[p,h] = signtest(...,'method',method)
[p,h,stats] = signtest(...)
Description p = signtest(x) performs a two-sided sign test of the null hypothesis
that data in the vector x come from a continuous distribution with zero
median, against the alternative that the distribution does not have zero
median. The p value of the test is returned in p
p = signtest(x,m) performs a two-sided sign test of the null
hypothesis that data in the vector x come from a continuous distribution
with median m, against the alternative that the distribution does not
have median m. m must be a scalar.
p = signtest(x,y) performs a paired, two-sided sign test of the
null hypothesis that data in the vector x-y come from a continuous
distribution with zero median, against the alternative that the
distribution does not have zero median. x and y must be the same
length. Note that a hypothesis of zero median for x-y is not equivalent
to a hypothesis of equal median for x and y.
[p,h] = signtest(...) returns the result of the test in h. h = 1
indicates a rejection of the null hypothesis at the 5% significance
level. h = 0 indicates a failure to reject the null hypothesis at the 5%
significance level.
[p,h] = signtest(...,'alpha',alpha) performs the test at the
(100*alpha)% significance level. The default, when unspecified, is
alpha = 0.05.
[p,h] = signtest(...,'method',method) computes the p value
using either an exact algorithm, when method is 'exact', or a normal
approximation, when method is 'approximate'. The default, when
18-1288
signtest
unspecified, is the exact method for small samples and the approximate
method for large samples.
[p,h,stats] = signtest(...) returns the structure stats with the
following fields:
• sign — Value of the sign test statistic
• zval — Value of the z-statistic (computed only for large samples)
Examples Test the hypothesis of zero median for the difference between two
paired samples.
before = lognrnd(2,.25,10,1);
after = before + (lognrnd(0,.5,10,1) - 1);
[p,h] = signtest(before,after)
p =
0.3438
h =
0
The sampling distribution of the difference between before and after
is symmetric with zero median. At the default 5% significance level, the
test fails to reject to the null hypothesis of zero median in the difference.
References [1] Gibbons, J. D. Nonparametric Statistical Inference. New York:
Marcel Dekker, 1985.
[2] Hollander, M., and D. A. Wolfe. Nonparametric Statistical Methods.
Hoboken, NJ: John Wiley & Sons, Inc., 1999.
See Also ranksum, signrank, ttest, ztest
18-1289
silhouette
Purpose Silhouette plot
Syntax silhouette(X,clust)
s = silhouette(X,clust)
[s,h] = silhouette(X,clust)
[...] = silhouette(X,clust,metric)
[...] = silhouette(X,clust,distfun,p1,p2,...)
Description silhouette(X,clust) plots cluster silhouettes for the n-by-p data
matrix X, with clusters defined by clust. Rows of X correspond to
points, columns correspond to coordinates. clust can be a categorical
variable, numeric vector, character matrix, or cell array of strings
containing a cluster name for each point. (See “Grouped Data” on page
2-34.) silhouette treats NaNs or empty strings in clust as missing
values, and ignores the corresponding rows of X. By default, silhouette
uses the squared Euclidean distance between points in X.
s = silhouette(X,clust) returns the silhouette values in the n-by-1
vector s, but does not plot the cluster silhouettes.
[s,h] = silhouette(X,clust) plots the silhouettes, and returns the
silhouette values in the n-by-1 vector s, and the figure handle in h.
[...] = silhouette(X,clust,metric) plots the silhouettes using
the inter-point distance function specified in metric. Choices for metric
are given in the following table.
Metric Description
'Euclidean' Euclidean distance
'sqEuclidean' Squared Euclidean distance (default)
'cityblock' Sum of absolute differences
'cosine' One minus the cosine of the included angle
between points (treated as vectors)
'correlation' One minus the sample correlation between points
(treated as sequences of values)
18-1290
silhouette
Metric Description
'Hamming' Percentage of coordinates that differ
'Jaccard' Percentage of nonzero coordinates that differ
Vector A numeric distance matrix in upper triangular
vector form, such as is created by pdist. X is
not used in this case, and can safely be set to [].
For more information on each metric, see “Distance Metrics” on page
12-14.
[...] = silhouette(X,clust,distfun,p1,p2,...) accepts a
function handle distfun to a metric of the form
d = distfun(X0,X,p1,p2,...)
where X0 is a 1-by-p point, X is an n-by-p matrix of points, and p1,p2,...
are optional additional arguments. The function distfun returns an
n-by-1 vector d of distances between X0 and each point (row) in X. The
arguments p1, p2,... are passed directly to the function distfun.
Remarks The silhouette value for each point is a measure of how similar that
point is to points in its own cluster compared to points in other clusters,
and ranges from -1 to +1. It is defined as
S(i) = (min(b(i,:),2) - a(i)) ./ max(a(i),min(b(i,:)))
where a(i) is the average distance from the ith point to the other
points in its cluster, and b(i,k) is the average distance from the ith
point to points in another cluster k.
Examples X = [randn(10,2)+ones(10,2);
randn(10,2)-ones(10,2)];
cidx = kmeans(X,2,'distance','sqeuclid');
s = silhouette(X,cidx,'sqeuclid');
18-1291
silhouette
References [1] Kaufman L., and P. J. Rousseeuw. Finding Groups in Data: An
Introduction to Cluster Analysis. Hoboken, NJ: John Wiley & Sons,
Inc., 1990.
See Also “Grouped Data” on page 2-34
dendrogram, kmeans, linkage, pdist
18-1292
categorical.single
Purpose Convert categorical array to single array
Syntax B = single(A)
Description B = single(A) converts the categorical array A to a single array.
Each element of B contains the internal categorical level code for the
corresponding element of A.
See Also double
18-1293
dataset.single
Purpose Convert dataset variables to single array
Syntax B = single(A)
B = single(A,vars)
Description B = single(A) returns the contents of the dataset A, converted to one
single array. The classes of the variables in the dataset must support
the conversion.
B = single(A,vars) returns the contents of the dataset variables
specified by vars. vars is a positive integer, a vector of positive
integers, a variable name, a cell array containing one or more variable
names, or a logical vector.
See Also dataset, double, replacedata
18-1294
categorical.size
Purpose Size of categorical array
Syntax d = size(A)
[m,n] = size(A)
[m1,m2,m3,...,mn] = size(A)
m = size(A,dim)
Description d = size(A) returns the two-element row vector d = [m,n] containing
the number of rows and columns in the matrix for an m-by-n categorical
matrix A. For n-D categorical arrays, size(A) returns a 1-by-n vector of
dimension lengths. Trailing singleton dimensions are ignored.
[m,n] = size(A) for a categorical matrix A, returns the number of
rows and columns in A as separate output variables.
[m1,m2,m3,...,mn] = size(A), for n>1, returns the sizes of the
first n dimensions of the categorical array A. If the number of output
arguments n does not equal ndims(A), then for:
n > ndims(A) size returns ones in the "extra" variables,
i.e., outputs ndims(A)+1 through n.
n < ndims(A) mn contains the product of the sizes of
dimensions n through ndims(A).
m = size(A,dim) returns the length of the dimension specified by the
scalar dim. For example, size(A,1) returns the number of rows. If
dim > ndims(A), m will be 1.
See Also length, ndims, numel
18-1295
dataset.size
Purpose Size of dataset array
Syntax B = single(A)
B = single(A,vars)
Description B = single(A) returns the contents of the dataset A, converted to one
single array. The classes of the variables in the dataset must support
the conversion.
B = single(A,vars) returns the contents of the dataset variables
specified by vars. vars is a positive integer, a vector of positive
integers, a variable name, a cell array containing one or more variable
names, or a logical vector.
See Also length, ndims, numel
18-1296
qrandset.size
Purpose Number of dimensions in matrix
Syntax d = size(p)
[m,n] = size(p)
m = size(p,dim)
Description d = size(p) returns the two-element row vector d = [m,n] containing
the number of points in the point set and the number of dimensions
the points are in, for the point set p. These correspond to the number
of rows and columns in the matrix that would be produced by the
expression p(:,:).
[m,n] = size(p) returns the number of points and dimensions for p as
separate output variables.
m = size(p,dim) returns the length of the dimension specified by the
scalar dim. For example, size(p,1) returns the number of rows (points
in the point set). If dim is greater than 2, m will be 1.
Examples The commands
P = sobolset(12);
d = size(P)
return
d = [9.0072e+015 12]
The command
[m,n] = size(P)
returns
m = 9.0072e+015
n = 12
The command
18-1297
qrandset.size
m2 = size(P, 2)
returns
m2 = 12
See Also length, ndims, qrandset
18-1298
slicesample
Purpose Slice sampler
Syntax rnd = slicesample(initial,nsamples,'pdf',pdf)
rnd = slicesample(...,'width',w)
rnd = slicesample(...,'burnin',k)
rnd = slicesample(...,'thin',m)
[rnd,neval] = slicesample(...)
Description rnd = slicesample(initial,nsamples,'pdf',pdf) generates
nsamples random samples from a target distribution whose density
function is defined by pdf using the slice sampling method. initial
is a row vector or scalar containing the initial value of the random
sample sequences. initial must be within the domain of the target
distribution. nsamples is the number of samples to be generated. pdf is
a function handle created using @. pdf accepts only one argument that
must be the same type and size as initial. It defines a function that is
proportional to the target density function. If the log density function
is preferred, 'pdf' can be replaced with 'logpdf'. The log density
function is not necessarily normalized.
rnd = slicesample(...,'width',w) performs slice sampling for the
target distribution with a typical width w. w is a scalar or vector. If it is
a scalar, all dimensions are assumed to have the same typical widths.
If it is a vector, each element of the vector is the typical width of the
marginal target distribution in that dimension. The default value of
w is 10.
rnd = slicesample(...,'burnin',k) generates random samples
with values between the starting point and the kth point omitted in
the generated sequence. Values beyond the kth point are kept. k is a
nonnegative integer with default value of 0.
rnd = slicesample(...,'thin',m) generates random samples with
m-1 out of m values omitted in the generated sequence. m is a positive
integer with default value of 1.
[rnd,neval] = slicesample(...) also returns neval, the averaged
number of function evaluations that occurred in the slice sampling.
neval is a scalar.
18-1299
slicesample
Examples Generate random samples from a distribution with a user-defined pdf.
First, define the function that is proportional to the pdf for a
multi-modal distribution.
f = @(x) exp( -x.^2/2).*(1+(sin(3*x)).^2).* ...
(1+(cos(5*x).^2));
Next, use the slicesample function to generate the random samples for
the function defined above.
x = slicesample(1,2000,'pdf',f,'thin',5,'burnin',1000);
Now, plot a histogram of the random samples generated.
hist(x,50)
set(get(gca,'child'),'facecolor',[0.8 .8 1]);
hold on
xd = get(gca,'XLim'); % Gets the xdata of the bins
binwidth = (xd(2)-xd(1)); % Finds the width of each bin
% Use linspace to normalize the histogram
y = 5.6398*binwidth*f(linspace(xd(1),xd(2),1000));
plot(linspace(xd(1),xd(2),1000),y,'r','LineWidth',2)
18-1300
slicesample
See Also rand, mhsample, randsample
18-1301
skewness
Purpose Skewness
Syntax y = skewness(X)
y = skewness(X,flag)
Description y = skewness(X) returns the sample skewness of X. For vectors,
skewness(x) is the skewness of the elements of x. For matrices,
skewness(X) is a row vector containing the sample skewness of each
column. For N-dimensional arrays, skewness operates along the first
nonsingleton dimension of X.
y = skewness(X,flag) specifies whether to correct for bias (flag = 0)
or not (flag = 1, the default). When X represents a sample from a
population, the skewness of X is biased; that is, it will tend to differ
from the population skewness by a systematic amount that depends
on the size of the sample. You can set flag = 0 to correct for this
systematic bias.
skewness(X,flag,dim) takes the skewness along dimension dim of X.
skewness treats NaNs as missing values and removes them.
Algorithm Skewness is a measure of the asymmetry of the data around the sample
mean. If skewness is negative, the data are spread out more to the
left of the mean than to the right. If skewness is positive, the data are
spread out more to the right. The skewness of the normal distribution
(or any perfectly symmetric distribution) is zero.
The skewness of a distribution is defined as
E(x − )
3
s=
3
where µ is the mean of x, σ is the standard deviation of x, and E(t)
represents the expected value of the quantity t. skewness computes a
sample version of this population value.
When you set flag to 1, the following equation applies:
18-1302
skewness
1 n
∑(x − x)
3
n i=1 i
s1 =
3
⎛ 1 n ⎞
⎜ n∑
⎜ ( xi − x ) 2⎟
⎟
⎝ i=1 ⎠
When you set flag to 0, the following equation applies:
n ( n − 1)
s0 = s1
n−2
This bias-corrected formula requires that X contain at least three
elements.
Examples X = randn([5 4])
X =
1.1650 1.6961 -1.4462 -0.3600
0.6268 0.0591 -0.7012 -0.1356
0.0751 1.7971 1.2460 -1.3493
0.3516 0.2641 -0.6390 -1.2704
-0.6965 0.8717 0.5774 0.9846
y = skewness(X)
y =
-0.2933 0.0482 0.2735 0.4641
See Also kurtosis, mean, moment, std, var
18-1303
qrandset.Skip property
Purpose Number of initial points to omit from sequence
Description The Skip property of a point set contains a positive integer which
specifies the number of initial points in the sequence to omit from the
point set. The default Skip value is 0.
Initial points of a sequence sometimes exhibit undesirable properties,
for example the first point is often (0,0,0,...) and this may
"unbalance" the sequence since its counterpart, (1,1,1,...), never
appears. Another common reason is that initial points often exhibit
correlations among different dimensions which disappear later in the
sequence.
Examples Examine the difference between skipping and not skipping points:
% No skipping produces the standard Sobol sequence.
P = sobolset(5);
P(1:3,:)
% Skip the first point of the sequence. The point set now
% starts at the second point of the basic Sobol sequence.
P.Skip = 1;
P(1:3,:)
See Also Leap | net | qrandset | subsref
18-1304
sobolset class
Superclasses qrandset
Purpose Sobol quasi-random point sets
Description sobolset is a quasi-random point set class that produces points from
the Sobol sequence. The Sobol sequence is a base-2 digital sequence
that fills space in a highly uniform manner.
Construction sobolset Construct Sobol quasi-random
point set
Methods Inherited Methods
Methods in the following table are inherited from qrandset.
disp Display qrandset object
end Last index in indexing expression
for point set
length Length of point set
ndims Number of dimensions in matrix
net Generate quasi-random point set
scramble Scramble quasi-random point set
size Number of dimensions in matrix
subsref Subscripted reference for
qrandset
Properties PointOrder Point generation method
Inherited Properties
Properties in the following table are inherited from qrandset.
18-1305
sobolset class
Dimensions Number of dimensions
Leap Interval between points
ScrambleMethod Settings that control scrambling
Skip Number of initial points to omit
from sequence
Type Name of sequence on which point
set P is based
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
References [1] Bratley, P., and B. L. Fox, "ALGORITHM 659 Implementing
Sobol’s Quasirandom Sequence Generator," ACM Transactions on
Mathematical Software, Vol. 14, No. 1, pp. 88-100, 1988.
[2] Joe, S., and F. Y. Kuo, "Remark on Algorithm 659: Implementing
Sobol’s Quasirandom Sequence Generator," ACM Transactions on
Mathematical Software, Vol. 29, No. 1, pp. 49-57, 2003.
[3] Hong, H. S., and F. J. Hickernell, "ALGORITHM 823: Implementing
Scrambled Digital Sequences," ACM Transactions on Mathematical
Software, Vol. 29, No. 2, pp. 95-109, 2003.
[4] Matousek, J., "On the L2-discrepancy for anchored boxes," Journal
of Complexity, Vol. 14, pp. 527-556, 1998.
See Also “Quasi-Random Point Sets” on page 6-18
haltonset
18-1306
sobolset
Purpose Construct Sobol quasi-random point set
Syntax p = sobolset(d)
p = sobolset(d,prop1,val1,prop2,val2,...)
Description p = sobolset(d) constructs a d-dimensional point set p of the
sobolset class, with default property settings.
p = sobolset(d,prop1,val1,prop2,val2,...) specifies property
name/value pairs used to construct p.
The object p returned by sobolset encapsulates properties of a
specified quasi-random sequence. The point set is finite, with a length
determined by the Skip and Leap properties and by limits on the size
of point set indices (maximum value of 253). Values of the point set are
not generated and stored in memory until you access p using net or
parenthesis indexing.
Examples Generate a 3-D Sobol point set, skip the first 1000 values, and then
retain every 101st point:
p = sobolset(3,'Skip',1e3,'Leap',1e2)
p =
Sobol point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
Leap : 100
ScrambleMethod : none
PointOrder : standard
Use scramble to apply a random linear scramble combined with a
random digital shift:
p = scramble(p,'MatousekAffineOwen')
p =
Sobol point set in 3 dimensions (8.918019e+013 points)
Properties:
Skip : 1000
18-1307
sobolset
Leap : 100
ScrambleMethod : MatousekAffineOwen
PointOrder : standard
Use net to generate the first four points:
X0 = net(p,4)
X0 =
0.7601 0.5919 0.9529
0.1795 0.0856 0.0491
0.5488 0.0785 0.8483
0.3882 0.8771 0.8755
Use parenthesis indexing to generate every third point, up to the 11th
point:
X = p(1:3:11,:)
X =
0.7601 0.5919 0.9529
0.3882 0.8771 0.8755
0.6905 0.4951 0.8464
0.1955 0.5679 0.3192
References [1] Bratley, P., and B. L. Fox. “Algorithm 659 Implementing
Sobol’s Quasirandom Sequence Generator.” ACM Transactions on
Mathematical Software. Vol. 14, No. 1, 1988, pp. 88–100.
[2] Joe, S., and F. Y. Kuo. “Remark on Algorithm 659: Implementing
Sobol’s Quasirandom Sequence Generator.” ACM Transactions on
Mathematical Software. Vol. 29, No. 1, 2003, pp. 49–57.
[3] Hong, H. S., and F. J. Hickernell. “Algorithm 823: Implementing
Scrambled Digital Sequences.” ACM Transactions on Mathematical
Software. Vol. 29, No. 2, 2003, pp. 95–109.
[4] Matousek, J. “On the L2-Discrepancy for Anchored Boxes.” Journal
of Complexity. Vol. 14, No. 4, 1998, pp. 527–556.
18-1308
sobolset
See Also haltonset, net, scramble
18-1309
ordinal.sort
Purpose Sort elements of ordinal array
Syntax B = sort(A)
B = sort(A,dim)
B = sort(A,dim,mode)
[B,I] = sort(A,...)
Description B = sort(A), when A is an ordinal vector, sorts the elements of A in
ascending order. For ordinal matrices, sort(A) sorts each column of
A in ascending order. For N-D ordinal arrays, sort(A) sorts the along
the first nonsingleton dimension of A. B is an ordinal array with the
same levels as A.
B = sort(A,dim) sorts A along dimension dim.
B = sort(A,dim,mode) sorts A in the order specified by mode. mode is
'ascend' for ascending order, or 'descend' for descending order.
[B,I] = sort(A,...) also returns an index matrix I. If A is a vector,
then B = A(I). If A is an m-by-n matrix and dim is 1, then B(:,j) =
A(I(:,j),j) for j = 1:n.
Elements with undefined levels are sorted to the end.
Examples Sort the columns of an ordinal array in ascending order:
A = ordinal([6 2 5; 2 4 1; 3 2 4],...
{'lo','med','hi'},[],[0 2 4 6])
A =
hi med hi
med hi lo
med med hi
B = sort(A)
B =
med med lo
med med hi
hi hi hi
18-1310
ordinal.sort
See Also sortrows
18-1311
dataset.sortrows
Purpose Sort rows of dataset array
Syntax B = sortrows(A)
B = sortrows(A,vars)
B = sortrows(A,'obsnames')
B = sortrows(A,vars,mode)
[B,idx] = sortrows(A)
Description B = sortrows(A) returns a copy of the dataset array A, with the
observations sorted in ascending order by all of the variables in A. The
observations in B are sorted first by the first variable, next by the second
variable, and so on. The variables in A must be scalar valued (i.e.,
column vectors) and be from a class for which a sort method exists.
B = sortrows(A,vars) sorts the observations in A by the variables
specified by vars. vars is a positive integer, a vector of positive
integers, variable names, a cell array containing one or more variable
names, or a logical vector.
B = sortrows(A,'obsnames') sorts the observations in A by
observation name.
B = sortrows(A,vars,mode) sorts in the direction specified by mode.
mode is 'ascend' (the default) or 'descend'. Use [] for vars to sort
using all variables.
[B,idx] = sortrows(A) also returns an index vector idx such that
B = A(idx,:).
Examples Sort the data in hospital.mat by age and then by last name:
load hospital
hospital(1:5,1:3)
ans =
LastName Sex Age
YPL-320 'SMITH' Male 38
GLI-532 'JOHNSON' Male 43
PNI-258 'WILLIAMS' Female 38
MIJ-579 'JONES' Female 40
18-1312
dataset.sortrows
XLK-030 'BROWN' Female 49
hospital = sortrows(hospital,{'Age','LastName'});
hospital(1:5,1:3)
ans =
LastName Sex Age
REV-997 'ALEXANDER' Male 25
FZR-250 'HALL' Male 25
LIM-480 'HILL' Female 25
XUE-826 'JACKSON' Male 25
SCQ-914 'JAMES' Male 25
See Also sortrows
18-1313
ordinal.sortrows
Purpose Sort rows
Syntax B = sortrows(A)
B = sortrows(A,col)
[B,I] = sortrows(A)
[B,I] = sortrows(A,col)
Description B = sortrows(A) sorts the rows of the 2-D ordinal matrix A in
ascending order, as a group. B is an ordinal array with the same levels
as A.
B = sortrows(A,col) sorts A based on the columns specified in the
vector col. If an element of col is positive, the corresponding column
in A is sorted in ascending order; if an element of col is negative, the
corresponding column in A is sorted in descending order.
[B,I] = sortrows(A) and [B,I] = sortrows(A,col) also returns an
index matrix I such that B = A(I,:).
Elements with undefined levels are sorted to the end.
Examples Sort the rows of an ordinal array in ascending order for the first column,
and then in descending order for the second column:
A = ordinal([6 2 5; 2 4 1; 3 2 4],...
{'lo','med','hi'},[],[0 2 4 6])
A =
hi med hi
med hi lo
med med hi
B = sortrows(A,[1 -2])
B =
med hi lo
med med hi
hi med hi
See Also sort, sortrows
18-1314
squareform
Purpose Format distance matrix
Syntax Z = squareform(y)
y = squareform(Z)
Z = squareform(y,'tovector')
Y = squareform(Z,'tomatrix')
Description Z = squareform(y), where y is a vector as created by the pdist
function, converts y into a square, symmetric format Z, in which Z(i,j)
denotes the distance between the ith and jth objects in the original
data.
y = squareform(Z), where Z is a square, symmetric matrix with zeros
along the diagonal, creates a vector y containing the Z elements below
the diagonal. y has the same format as the output from the pdist
function.
Z = squareform(y,'tovector') forces squareform to treat y as a
vector.
Y = squareform(Z,'tomatrix') forces squareform to treat Z as a
matrix.
The last two formats are useful if the input has a single element, so that
it is ambiguous whether the input is a vector or square matrix.
Examples y = 1:6
y =
1 2 3 4 5 6
X = [0 1 2 3; 1 0 4 5; 2 4 0 6; 3 5 6 0]
X =
0 1 2 3
1 0 4 5
2 4 0 6
3 5 6 0
Then squareform(y) = X and squareform(X) = y.
18-1315
squareform
See Also pdist
18-1316
categorical.squeeze
Purpose Squeeze singleton dimensions from categorical array
Syntax B = squeeze(A)
Description B = squeeze(A) returns an array B with the same elements as the
categorical array A but with all the singleton dimensions removed. A
singleton is a dimension such that size(A,dim)==1. 2-D arrays are
unaffected by squeeze so that row vectors remain rows.
See Also shiftdim
18-1317
dataset.stack
Purpose Stack data from multiple variables into single variable
Syntax tall = stack(wide,datavars)
[tall,iwide] = stack(wide,datavars)
tall = stack(wide,datavars,Parameter,value)
Description tall = stack(wide,datavars) converts a wide-format dataset array
into a tall-format array, by stacking multiple variables in wide into a
single variable in tall. In general, tall contains fewer variables but
more observations than wide.
datavars specifies a group of m data variables in wide. stack creates a
single data variable in tall by interleaving their values, and if wide
has n observations, then tall has m-by-n observations. In other words,
stack takes the m data values from each observation in wide and stacks
them up to create m observations in tall. datavars is a positive integer,
a vector of positive integers, a variable name, a cell array containing
one or more variable names, or a logical vector. stack also creates a
grouping variable in tall to indicate which of the m data variables in
wide each observation in tall corresponds to.
stack assigns values for the "per-variable properties (e.g., Units
and VarDescription) for the new data variable in tall from the
corresponding property values for the first variable listed in datavars.
stack copies the remaining variables from wide to tall without
stacking, by replicating each of their values m times. These variables are
typically grouping variables. Because their values are constant across
each group of m observations in tall, they identify which observation in
wide an observation in tall came from.
[tall,iwide] = stack(wide,datavars) returns an index vector iwide
indicating the correspondence between observations in tall and those
in wide. stack creates tall(j,:) using wide(iwide(j),datavarss).
For more information on grouping variables, see “Grouping Variables”
on page 2-34.
18-1318
dataset.stack
Input tall = stack(wide,datavars,Parameter,value) uses the following
Arguments parameter name/value pairs to control how stack converts variables in
wide to variables in tall:
'ConstVars' Variables in wide to copy to tall
without stacking. ConstVars
is a positive integer, a vector
of positive integers, a variable
name, a cell array containing
one or more variable names, or a
logical vector. The default is all
variables in wide not specified in
datavars.
'NewdatavarsName' A name for the data variable to be
created in tall. The default is a
concatenation of the names of the
m variables that are stacked up.
'DataIndVarName' A name for the grouping
variable to create in tall to
indicate the source of each
value in the new data variable.
The default is based on the
'NewdatavarsNames' parameter.
You can also specify multiple groups of data variables in wide, each of
which becomes a variable in tall. All groups must contain the same
number of variables. Use a cell array to contain multiple parameter
values for datavars, and a cell array of strings to contain multiple
'NewdatavarsNames'.
Examples Convert a wide format data set to tall format, and then back to a
different wide format:
load flu
flu2 = stack(flu, 2:11, 'NewdatavarsName','FluRate',...
'IndVarName','Region')
dateNames = cellstr(datestr(flu.Date,'mmm_DD_YYYY'));
18-1319
dataset.stack
flu3 = unstack(flu2, 'FluRate', 'Date',...
'NewdatavarsNames',dateNames)
See Also dataset.unstack | dataset.join | dataset.grpstats
How To • “Grouping Variables” on page 2-34
18-1320
qrandstream.State property
Purpose Current state of the stream
Description The State property of a quasi-random stream contains the index into
the associated point set of the next point to draw in the stream. Getting
and resetting the State property allows you to return a stream to a
previous state. The initial value of State is 1.
Examples Q = qrandstream('sobol', 5);
s = Q.State;
u1 = qrand(Q, 10)
Q.State = s;
u2 = qrand(Q, 10) % contains exactly the same values as u1
See Also qrand
18-1321
statget
Purpose Access values in statistics options structure
Syntax val = statget(options,param)
val = statget(options,param,default)
Description val = statget(options,param) returns the value of the parameter
specified by the string param in the statistics options structure options.
If the parameter is undefined in options, statget returns []. You need
to type only enough leading characters to define the parameter name
uniquely. statget ignores case for parameter names. For available
options, see Inputs.
val = statget(options,param,default) returns default if the
specified parameter is undefined in the optimization options structure
options.
Input DerivStep
Arguments Relative difference used in finite difference derivative calculations.
A positive scalar, or a vector of positive scalars the same size
as the vector of parameters estimated by the Statistics Toolbox
function using the options structure.
Display
Amount of information displayed by the algorithm.
• 'off' — Displays no information.
• 'final' — Displays the final output.
• 'iter' — Displays iterative output to the command window
for some functions; otherwise displays the final output.
FunValCheck
Check for invalid values, such as NaN or Inf, from the objective
function.
• 'off'
18-1322
statget
• 'on'
GradObj
Flags whether the objective function returns a gradient vector
as a second output.
• 'off'
• 'on'
Jacobian
Flags whether the objective function returns a Jacobian as a
second output.
• 'off'
• 'on'
MaxFunEvals
Maximum number of objective function evaluations allowed.
Positive integer.
MaxIter
Maximum number of iterations allowed. Positive integer.
OutputFcn
The solver calls all output functions after each iteration.
• Function handle specified using @
• a cell array with function handles
• an empty array (default)
Robust
Invoke robust fitting option.
18-1323
statget
• 'off'
• 'on'
Streams
A single instance of the RandStream class, or a cell array of
RandStream instances. The Streams option is accepted by some
functions to govern what stream(s) to use in generating random
numbers within the function. If 'UseSubstreams' is 'always',
the Streams value must be a scalar, or must be empty. If
'UseParallel' is 'always' and 'UseSubstreams' is 'never',
then the Streams argument must either be empty, or its length
must match the number of processors used in the computation:
equal to the matlabpool size if a matlabpool is open, a scalar
otherwise.
TolBnd
Parameter bound tolerance. Positive scalar.
TolFun
Termination tolerance for the objective function value. Positive
scalar.
TolTypeFun
Use TolFun for absolute or relative objective function tolerances.
• 'abs'
• 'rel'
TolTypeX
Use TolX for absolute or relative parameter tolerances.
• 'abs'
• 'rel'
18-1324
statget
TolX
Termination tolerance for the parameters. Positive scalar.
Tune
The tuning constant used in robust fitting to normalize the
residuals before applying the weight function. The default value
depends upon the weight function. This parameter is necessary
if you specify the weight function as a function handle. Positive
scalar.
UseParallel
Flag indicating whether eligible functions should use capabilities
of the Parallel Computing Toolbox (PCT), if the capabilities are
available. That is, if the PCT is installed, and a PCT matlabpool
is in effect. Eligible functions are bootci, bootstrp, crossval,
jackknife, and the TreeBagger constructor. Valid values are
'never' (the default), for serial computation, and 'always', for
parallel computation.
UseSubstreams
Flag indicating whether the random number generator in eligible
functions should use Substream property of the RandStream
class. 'never' (default) or 'always'. 'always', high level
iterations within the function will set the Substream property
to the value of the iteration. This behavior helps to generate
reproducible random number streams in parallel and/or serial
mode computation. Eligible functions are bootci, bootstrp,
crossval, and the TreeBagger constructor.
WgtFun
A weight function for robust fitting. Valid only when Robust is
'on'. Can also be a function handle that accepts a normalized
residual as input and returns the robust weights as output.
• 'bisquare'
• 'andrews'
18-1325
statget
• 'cauchy'
• 'fair'
• 'huber'
• 'logistic'
• 'talwar'
• 'welsch'
Examples This statement returns the value of the Display statistics options
parameter from the structure called my_options.
val = statget(my_options,'Display')
Return the value of the Display statistics options parameter from
the structure called my_options (as in the previous example). If the
Display parameter is undefined, statget returns the value 'final'.
optnew = statget(my_options,'Display','final');
See Also statset
18-1326
statset
Purpose Create statistics options structure
Syntax statset
statset(statfun)
options = statset(...)
options = statset(fieldname1,val1,fieldname2,val2,...)
options = statset(oldopts,fieldname1,val1,fieldname2,val2,
...)
options = statset(oldopts,newopts)
Description statset with no input arguments and no output arguments displays all
fields of a statistics options structure and their possible values.
statset(statfun) displays fields and default values used by the
Statistics Toolbox function statfun. Specify statfun using a string
name or a function handle.
options = statset(...) creates a statistics options structure options.
With no input arguments, all fields of the options structure are an
empty array ([]). With a specified statfun, function-specific fields are
default values and the remaining fields are []. Function-specific fields
set to [] indicate that the function is to use its default value for that
parameter. For available options, see Inputs.
options = statset(fieldname1,val1,fieldname2,val2,...)
creates an options structure in which the named fields have the
specified values. Any unspecified values are []. Use strings for field
names. For fields that are string-valued, you must input the complete
string for the value. If you provide an invalid string for a value, statset
uses the default.
options =
statset(oldopts,fieldname1,val1,fieldname2,val2,...)
creates a copy of oldopts with the named parameters changed to
the specified values.
options = statset(oldopts,newopts) combines an existing options
structure, oldopts, with a new options structure, newopts. Any
18-1327
statset
parameters in newopts with nonempty values overwrite corresponding
parameters in oldopts.
Input DerivStep
Arguments Relative difference used in finite difference derivative calculations.
A positive scalar, or a vector of positive scalars the same size
as the vector of parameters estimated by the Statistics Toolbox
function using the options structure.
Display
Amount of information displayed by the algorithm.
• 'off' — Displays no information.
• 'final' — Displays the final output.
• 'iter' — Displays iterative output to the command window
for some functions; otherwise displays the final output.
FunValCheck
Check for invalid values, such as NaN or Inf, from the objective
function.
• 'off'
• 'on'
GradObj
Flags whether the objective function returns a gradient vector
as a second output.
• 'off'
• 'on'
Jacobian
18-1328
statset
Flags whether the objective function returns a Jacobian as a
second output.
• 'off'
• 'on'
MaxFunEvals
Maximum number of objective function evaluations allowed.
Positive integer.
MaxIter
Maximum number of iterations allowed. Positive integer.
OutputFcn
The solver calls all output functions after each iteration.
• Function handle specified using @
• a cell array with function handles
• an empty array (default)
Robust
Invoke robust fitting option.
• 'off'
• 'on'
Streams
A single instance of the RandStream class, or a cell array of
RandStream instances. The Streams option is accepted by some
functions to govern what stream(s) to use in generating random
numbers within the function. If 'UseSubstreams' is 'always',
the Streams value must be a scalar, or must be empty. If
'UseParallel' is 'always' and 'UseSubstreams' is 'never',
18-1329
statset
then the Streams argument must either be empty, or its length
must match the number of processors used in the computation:
equal to the matlabpool size if a matlabpool is open, a scalar
otherwise.
TolBnd
Parameter bound tolerance. Positive scalar.
TolFun
Termination tolerance for the objective function value. Positive
scalar.
TolTypeFun
Use TolFun for absolute or relative objective function tolerances.
• 'abs'
• 'rel'
TolTypeX
Use TolX for absolute or relative parameter tolerances.
• 'abs'
• 'rel'
TolX
Termination tolerance for the parameters. Positive scalar.
Tune
The tuning constant used in robust fitting to normalize the
residuals before applying the weight function. The default value
depends upon the weight function. This parameter is necessary
if you specify the weight function as a function handle. Positive
scalar.
UseParallel
18-1330
statset
Flag indicating whether eligible functions should use capabilities
of the Parallel Computing Toolbox (PCT), if the capabilities are
available. That is, if the PCT is installed, and a PCT matlabpool
is in effect. Eligible functions are bootci, bootstrp, crossval,
jackknife, and the TreeBagger constructor. Valid values are
'never' (the default), for serial computation, and 'always', for
parallel computation.
UseSubstreams
Flag indicating whether the random number generator in eligible
functions should use Substream property of the RandStream
class. 'never' (default) or 'always'. 'always', high level
iterations within the function will set the Substream property
to the value of the iteration. This behavior helps to generate
reproducible random number streams in parallel and/or serial
mode computation. Eligible functions are bootci, bootstrp,
crossval, and the TreeBagger constructor.
WgtFun
A weight function for robust fitting. Valid only when Robust is
'on'. Can also be a function handle that accepts a normalized
residual as input and returns the robust weights as output.
• 'bisquare'
• 'andrews'
• 'cauchy'
• 'fair'
• 'huber'
• 'logistic'
• 'talwar'
• 'welsch'
18-1331
statset
Examples Suppose you want to change the default parameter values for the
function evfit, which fits an extreme value distribution to data. The
defaults parameter values are:
statset('evfit')
ans =
Display: 'off'
MaxFunEvals: []
MaxIter: []
TolBnd: []
TolFun: []
TolX: 1.0000e-006
GradObj: []
DerivStep: []
FunValCheck: []
Robust: []
WgtFun: []
Tune: []
The only parameters that evfit uses are Display and TolX. To create
an options structure with the value of TolX set to 1e-8, enter:
options = statset('TolX',1e-8)
% Pass options to evfit:
mu = 1;
sigma = 1;
data = evrnd(mu,sigma,1,100);
paramhat = evfit(data,[],[],[],options)
See Also statget
18-1332
ProbDistUnivParam.std
Purpose Return standard deviation of ProbDistUnivParam object
Syntax S = std(PD)
Description S = std(PD) returns S, the standard deviation of the
ProbDistUnivParam object PD.
Input PD An object of the class ProbDistUnivParam.
Arguments
Output S The standard deviation of the
Arguments ProbDistUnivParam object PD.
See Also std
18-1333
stepwise
Purpose Interactive stepwise regression
Syntax stepwise
stepwise(X,y)
stepwise(X,y,inmodel,penter,premove)
Description stepwise uses the sample data in hald.mat to display a graphical user
interface for performing stepwise regression of the response values in
heat on the predictive terms in ingredients.
18-1334
stepwise
The upper left of the interface displays estimates of the coefficients for
all potential terms, with horizontal bars indicating 90% (colored) and
95% (grey) confidence intervals. The red color indicates that, initially,
the terms are not in the model. Values displayed in the table are those
that would result if the terms were added to the model.
The middle portion of the interface displays summary statistics for the
entire model. These statistics are updated with each step.
18-1335
stepwise
The lower portion of the interface, Model History, displays the RMSE
for the model. The plot tracks the RMSE from step to step, so you can
compare the optimality of different models. Hover over the blue dots
in the history to see which terms were in the model at a particular
step. Click on a blue dot in the history to open a copy of the interface
initialized with the terms in the model at that step.
Initial models, as well as entrance/exit tolerances for the p-values
of F-statistics, are specified using additional input arguments to
stepwise. Defaults are an initial model with no terms, an entrance
tolerance of 0.05, and an exit tolerance of 0.10.
To center and scale the input data (compute z-scores) to improve
conditioning of the underlying least-squares problem, select Scale
Inputs from the Stepwise menu.
You proceed through a stepwise regression in one of two ways:
1 Click Next Step to select the recommended next step. The
recommended next step either adds the most significant term or
removes the least significant term. When the regression reaches a
local minimum of RMSE, the recommended next step is “Move no
terms.” You can perform all of the recommended steps at once by
clicking All Steps.
2 Click a line in the plot or in the table to toggle the state of the
corresponding term. Clicking a red line, corresponding to a term not
currently in the model, adds the term to the model and changes the
line to blue. Clicking a blue line, corresponding to a term currently
in the model, removes the term from the model and changes the line
to red.
To call addedvarplot and produce an added variable plot from the
stepwise interface, select Added Variable Plot from the Stepwise
menu. A list of terms is displayed. Select the term you want to add,
and then click OK.
Click Export to display a dialog box that allows you to select
information from the interface to save to the MATLAB workspace.
18-1336
stepwise
Check the information you want to export and, optionally, change the
names of the workspace variables to be created. Click OK to export
the information.
stepwise(X,y) displays the interface using the p predictive terms in
the n-by-p matrix X and the response values in the n-by-1 vector y.
Distinct predictive terms should appear in different columns of X.
Note stepwise automatically includes a constant term in all models.
Do not enter a column of 1s directly into X.
stepwise treats NaN values in either X or y as missing values, and
ignores them.
stepwise(X,y,inmodel,penter,premove) additionally specifies the
initial model (inmodel) and the entrance (penter) and exit (premove)
tolerances for the p-values of F-statistics. inmodel is either a logical
vector with length equal to the number of columns of X, or a vector of
indices, with values ranging from 1 to the number of columns in X. The
value of penter must be less than or equal to the value of premove.
Algorithm Stepwise regression is a systematic method for adding and removing
terms from a multilinear model based on their statistical significance
in a regression. The method begins with an initial model and then
compares the explanatory power of incrementally larger and smaller
models. At each step, the p value of an F-statistic is computed to test
models with and without a potential term. If a term is not currently
in the model, the null hypothesis is that the term would have a zero
coefficient if added to the model. If there is sufficient evidence to reject
the null hypothesis, the term is added to the model. Conversely, if a
term is currently in the model, the null hypothesis is that the term
has a zero coefficient. If there is insufficient evidence to reject the null
hypothesis, the term is removed from the model. The method proceeds
as follows:
1 Fit the initial model.
18-1337
stepwise
2 If any terms not in the model have p-values less than an entrance
tolerance (that is, if it is unlikely that they would have zero coefficient
if added to the model), add the one with the smallest p value and
repeat this step; otherwise, go to step 3.
3 If any terms in the model have p-values greater than an exit tolerance
(that is, if it is unlikely that the hypothesis of a zero coefficient can
be rejected), remove the one with the largest p value and go to step
2; otherwise, end.
Depending on the terms included in the initial model and the order in
which terms are moved in and out, the method may build different
models from the same set of potential terms. The method terminates
when no single step improves the model. There is no guarantee,
however, that a different initial model or a different sequence of steps
will not lead to a better fit. In this sense, stepwise models are locally
optimal, but may not be globally optimal.
See Also addedvarplot | regress | stepwisefit
How To • “Stepwise Regression” on page 9-19
18-1338
stepwisefit
Purpose Stepwise regression
Syntax b = stepwisefit(X,y)
[b,se,pval,inmodel,stats,nextstep,history] = stepwisefit(...)
[...] = stepwisefit(X,y,param1,val1,param2,val2,...)
Description b = stepwisefit(X,y) uses a stepwise method to perform a
multilinear regression of the response values in the n-by-1 vector y
on the p predictive terms in the n-by-p matrix X. Distinct predictive
terms should appear in different columns of X. b is a p-by-1 vector of
estimated coefficients for all of the terms in X. The value in b for a term
not included in the final model is the coefficient estimate that would
result from adding the term to the model.
Note stepwisefit automatically includes a constant term in all
models. Do not enter a column of 1s directly into X.
stepwisefit treats NaN values in either X or y as missing values, and
ignores them.
[b,se,pval,inmodel,stats,nextstep,history] =
stepwisefit(...) returns the following additional information:
• se — A vector of standard errors for b
• pval — A vector of p-values for testing whether elements of b are 0
• inmodel — A logical vector, with length equal to the number of
columns in X, specifying which terms are in the final model
• stats — A structure of additional statistics with the following fields.
All statistics pertain to the final model except where noted.
- source — The string 'stepwisefit'
- dfe — Degrees of freedom for error
- df0 — Degrees of freedom for the regression
18-1339
stepwisefit
- SStotal — Total sum of squares of the response
- SSresid — Sum of squares of the residuals
- fstat — F-statistic for testing the final model vs. no model (mean
only)
- pval — p value of the F-statistic
- rmse — Root mean square error
- xr — Residuals for predictors not in the final model, after
removing the part of them explained by predictors in the model
- yr — Residuals for the response using predictors in the final model
- B — Coefficients for terms in final model, with values for a term
not in the model set to the value that would be obtained by adding
that term to the model
- SE — Standard errors for coefficient estimates
- TSTAT — t statistics for coefficient estimates
- PVAL — p-values for coefficient estimates
- intercept — Estimated intercept
- wasnan — Indicates which rows in the data contained NaN values
• nextstep — The recommended next step—either the index of the
next term to move in or out of the model, or 0 if no further steps are
recommended
• history — A structure containing information on steps taken, with
the following fields:
- rmse — Root mean square errors for the model at each step
- df0 — Degrees of freedom for the regression at each step
- in — Logical array indicating which predictors are in the model
at each step
18-1340
stepwisefit
[...] = stepwisefit(X,y,param1,val1,param2,val2,...)
specifies one or more of the name/value pairs described in the following
table.
Parameter Value
'inmodel' A logical vector specifying terms to include in the
initial fit. The default is to specify no terms.
'penter' The maximum p value for a term to be added. The
default is 0.05.
'premove' The minimum p value for a term to be removed. The
default is the maximum of the value of 'penter' and
0.10.
'display' 'on' displays information about each step in the
command window. This is the default.
'off' omits the display.
'maxiter' The maximum number of steps in the regression. The
default is Inf.
'keep' A logical vector specifying terms to keep in their initial
state. The default is to specify no terms.
'scale' 'on' centers and scales each column of X (computes
z-scores) before fitting.
'off' does not scale the terms. This is the default.
Algorithm Stepwise regression is a systematic method for adding and removing
terms from a multilinear model based on their statistical significance
in a regression. The method begins with an initial model and then
compares the explanatory power of incrementally larger and smaller
models. At each step, the p value of an F-statistic is computed to test
models with and without a potential term. If a term is not currently
in the model, the null hypothesis is that the term would have a zero
coefficient if added to the model. If there is sufficient evidence to reject
18-1341
stepwisefit
the null hypothesis, the term is added to the model. Conversely, if a
term is currently in the model, the null hypothesis is that the term
has a zero coefficient. If there is insufficient evidence to reject the null
hypothesis, the term is removed from the model. The method proceeds
as follows:
1 Fit the initial model.
2 If any terms not in the model have p-values less than an entrance
tolerance (that is, if it is unlikely that they would have zero coefficient
if added to the model), add the one with the smallest p value and
repeat this step; otherwise, go to step 3.
3 If any terms in the model have p-values greater than an exit tolerance
(that is, if it is unlikely that the hypothesis of a zero coefficient can
be rejected), remove the one with the largest p value and go to step
2; otherwise, end.
Depending on the terms included in the initial model and the order in
which terms are moved in and out, the method may build different
models from the same set of potential terms. The method terminates
when no single step improves the model. There is no guarantee,
however, that a different initial model or a different sequence of steps
will not lead to a better fit. In this sense, stepwise models are locally
optimal, but may not be globally optimal.
Examples Load the data in hald.mat, which contains observations of the heat of
reaction of various cement mixtures:
load hald
whos
Name Size Bytes Class Attributes
Description 22x58 2552 char
hald 13x5 520 double
heat 13x1 104 double
ingredients 13x4 416 double
18-1342
stepwisefit
The response (heat) depends on the quantities of the four predictors
(the columns of ingredients).
Use stepwisefit to carry out the stepwise regression algorithm,
beginning with no terms in the model and using entrance/exit tolerances
of 0.05/0.10 on the p-values:
stepwisefit(ingredients,heat,...
'penter',0.05,'premove',0.10);
Initial columns included: none
Step 1, added column 4, p=0.000576232
Step 2, added column 1, p=1.10528e-006
Final columns included: 1 4
'Coeff' 'Std.Err.' 'Status' 'P'
[ 1.4400] [ 0.1384] 'In' [1.1053e-006]
[ 0.4161] [ 0.1856] 'Out' [ 0.0517]
[-0.4100] [ 0.1992] 'Out' [ 0.0697]
[-0.6140] [ 0.0486] 'In' [1.8149e-007]
stepwisefit automatically includes an intercept term in the model, so
you do not add it explicitly to ingredients as you would for regress.
For terms not in the model, coefficient estimates and their standard
errors are those that result if the term is added.
The inmodel parameter is used to specify terms in an initial model:
initialModel = ...
[false true false false]; % Force in 2nd term
stepwisefit(ingredients,heat,...
'inmodel',initialModel,...
'penter',.05,'premove',0.10);
Initial columns included: 2
Step 1, added column 1, p=2.69221e-007
Final columns included: 1 2
'Coeff' 'Std.Err.' 'Status' 'P'
[ 1.4683] [ 0.1213] 'In' [2.6922e-007]
[ 0.6623] [ 0.0459] 'In' [5.0290e-008]
[ 0.2500] [ 0.1847] 'Out' [ 0.2089]
18-1343
stepwisefit
[-0.2365] [ 0.1733] 'Out' [ 0.2054]
The preceding two models, built from different initial models, use
different subsets of the predictive terms. Terms 2 and 4, swapped in the
two models, are highly correlated:
term2 = ingredients(:,2);
term4 = ingredients(:,4);
R = corrcoef(term2,term4)
R =
1.0000 -0.9730
-0.9730 1.0000
To compare the models, use the stats output of stepwisefit:
[betahat1,se1,pval1,inmodel1,stats1] = ...
stepwisefit(ingredients,heat,...
'penter',.05,'premove',0.10,...
'display','off');
[betahat2,se2,pval2,inmodel2,stats2] = ...
stepwisefit(ingredients,heat,...
'inmodel',initialModel,...
'penter',.05,'premove',0.10,...
'display','off');
RMSE1 = stats1.rmse
RMSE1 =
2.7343
RMSE2 = stats2.rmse
RMSE2 =
2.4063
The second model has a lower Root Mean Square Error (RMSE).
References [1] Draper, N. R., and H. Smith. Applied Regression Analysis. Hoboken,
NJ: Wiley-Interscience, 1998. pp. 307–312.
See Also stepwise, addedvarplot, regress
18-1344
categorical.subsasgn
Purpose Subscripted assignment for categorical array
Syntax A = subsasgn(A,S,B)
Description A = subsasgn(A,S,B) is called for the syntax A(i)=B. S is a structure
array with the fields:
type String containing '()' specifying
the subscript type. Only
parenthesis subscripting is
allowed.
subs Cell array or string containing
the actual subscripts.
See Also categorical, subsref
18-1345
classregtree.subsasgn
Purpose Subscripted reference for classregtree object
Syntax
Description Subscript assignment is not allowed for a classregtree object.
See Also classregtree
18-1346
dataset.subsasgn
Purpose Subscripted assignment to dataset array
Description A = subsasgn(A,S,B) is called for the syntax A(i,j)=B, A{i,j}=B, or
A. var=B when A is a dataset array. S is a structure array with the fields:
type String containing '()', '{}', or
'.' specifying the subscript type.
subs Cell array or string containing
the actual subscripts.
A(i,j) = B assigns the contents of the dataset array B to a subset of
the observations and variables in the dataset array A. i and j are one of
the following types:
• positive integers
• vectors of positive integers
• observation/variable names
• cell arrays containing one or more observation/variable names
• logical vectors
The assignment does not use observation names, variable names, or any
other properties of B to modify properties of A; however properties of A
are extended with default values if the assignment expands the number
of observations or variables in A. Elements of B are assigned into A by
position, not by matching names.
A{i,j} = B assigns the value B into an element of the dataset array A.
i and J are positive integers, or logical vectors. Cell indexing cannot
assign into multiple dataset elements, that is, the subscripts i and
j must each refer to only a single observation or variable. B is cast
to the type of the target variable if necessary. If the dataset element
already exists, A{i,j} may also be followed by further subscripting as
supported by the variable.
18-1347
dataset.subsasgn
For dataset variables that are cell arrays, assignments such as
A{1,'CellVar'} = B assign into the contents of the target dataset
element in the same way that {}-indexing of an ordinary cell array does.
For dataset variables that are n-D arrays, i.e., each observation is a
matrix or array, an assignment such as A{1,'ArrayVar'} = B assigns
into the second and following dimensions of the target dataset element,
i.e., the assignment adds a leading singleton dimension to B to account
for the observation dimension of the dataset variable.
A.var = B or A.(varname) = B assigns B to a dataset variable. var is
a variable name literal, or varname is a character variable containing a
variable name. If the dataset variable already exists, the assignment
completely replaces that variable. To assign into an element of the
variable, A.var or A.(varname) may be followed by further subscripting
as supported by the variable. In particular, A.var(obsnames,...) =
B and A.var{obsnames,...} = B (when supported by var) provide
assignment into a dataset variable using observation names.
A.properties.propertyname = P assigns to a dataset property.
propertyname is one of the following:
• 'ObsNames'
• 'VarNames'
• 'Description'
• 'Units'
• 'DimNames'
• 'UserData'
• 'VarDescription'
To assign into an element of the property, A.properties.propertyname
may also be followed by further subscripting as supported by the
property.
You cannot assign multiple values into dataset variables or
properties using assignments such as [A.CellVar{1:2}] = B,
18-1348
dataset.subsasgn
[A.StructVar(1:2).field] = B, or [A.Properties.ObsNames{1:2}]
= B. Use multiple assignments of the form A.CellVar{1} = B instead.
Similarly, if a dataset variable is a cell array with multiple columns
or is an n-D cell array, then the contents of that variable for a single
observation consists of multiple cells, and you cannot assign to all of
them using the syntax A{1,'CellVar'} = B. Use multiple assignments
of the form [A.CellVar{1,1}] = B instead.
See Also dataset, set, subsref
18-1349
gmdistribution.subsasgn
Purpose Subscripted reference for Gaussian mixture distribution object
Description Subscript assignment is not allowed for gmdistribution objects.
See Also gmdistribution
18-1350
NaiveBayes.subsasgn
Purpose Subscripted reference for NaiveBayes object
Description Subscript assignment is not allowed for a NaiveBayes object.
18-1351
categorical.subsindex
Purpose Subscript index for categorical array
Syntax I = subsindex(A)
Description I = subsindex(A) is called for the syntax 'X(A)' when A is a
categorical array and X is one of the built-in types (most commonly
'double'). subsindex returns the internal categorical level codes
of A converted to zero-based integer indices. subsindex is invoked
separately on all the subscripts in an expression such as X(A,B).
Examples load fisheriris
a = ordinal(species,[],unique(species));
colmeans = grpstats(meas,a,@mean);
residuals = meas - colmeans(a,:);
See Also categorical, double
18-1352
classregtree.subsref
Purpose Subscripted reference for classregtree object
Syntax B = subsref(T,S)
Description B = subsref(T,S) is called for the syntax T(X) when T is a
classregtree object. S is a structure array with the fields:
type String containing '()', '{}', or '.' specifying
the subscript type.
subs Cell array or string containing the actual
subscripts.
[...]=T(...) invokes the eval method for the tree T.
See Also classregtree, eval
18-1353
categorical.subsref
Purpose Subscripted reference for categorical array
Syntax A = subsref(A,S,B)
Description A = subsref(A,S,B) is called for the syntax A(I)=B. S is a structure
array with the fields:
type String containing '()' specifying
the subscript type. Only
parenthesis subscripting is
allowed.
subs Cell array or string containing
the actual subscripts.
See Also categorical, subsasgn
18-1354
dataset.subsref
Purpose Subscripted reference for dataset array
Syntax B = subsref(A,S)
Description B = subsref(A,S) is called for the syntax A(i,j), A{i,j}, or A.var
when A is a dataset array. S is a structure array with the fields:
type String containing '()', '{}', or '.' specifying the
subscript type.
subs Cell array or string containing the actual
subscripts.
B = A(i,j) returns a dataset array that contains a subset of the
observations and variables in the dataset array A. i and j are one of
the following types:
• positive integers
• vectors of positive integers
• observation/variable names
• cell arrays containing one or more observation/variable names
• logical vectors
B contains the same property values as A, subsetted for observations
or variables where appropriate.
B = A{i,j} returns an element of a dataset variable. i and j are
positive integers, or logical vectors. Cell indexing cannot return
multiple dataset elements, that is, the subscripts i and j must each
refer to only a single observation or variable. A{i,j} may also be
followed by further subscripting as supported by the variable.
For dataset variables that are cell arrays, expressions such as
A{1,'CellVar'} return the contents of the referenced dataset element
in the same way that {}-indexing on an ordinary cell array does. If the
dataset variable is a single column of cells, the contents of a single cell
18-1355
dataset.subsref
is returned. If the dataset variable has multiple columns or is n-D,
multiple outputs containing the contents of multiple cells are returned.
For dataset variables that are n-D arrays, i.e., each observation is
a matrix or an array, expressions such as A{1,'ArrayVar'} return
A.ArrayVar(1,:,...) with the leading singleton dimension squeezed
out.
B = A.var or A.(varname) returns a dataset variable. var is a
variable name literal, or varname is a character variable containing
a variable name. A.var or A.(varname) may also be followed by
further subscripting as supported by the variable. In particular,
A.var(obsnames,...) and A.var{obsnames,...} (when supported by
var) provide subscripting into a dataset variable using observation
names.
P = A.Properties.propertyname returns a dataset property.
propertyname is one of the following:
• 'ObsNames'
• 'VarNames'
• 'Description'
• 'Units'
• 'DimNames'
• 'UserData'
• 'VarDescription'
A.properties.propertyname may also be followed by further
subscripting as supported by the property.
Limitations
Subscripting expressions such as A.CellVar{1:2},
A.StructVar(1:2).field, or A.Properties.ObsNames{1:2}
are valid, but result in subsref returning multiple outputs in the
form of a comma-separated list. If you explicitly assign to output
18-1356
dataset.subsref
arguments on the left-hand side of an assignment, for example,
[cellval1,cellval2] = A.CellVar{1:2}, those variables will receive
the corresponding values. However, if there are no output arguments,
only the first output in the comma-separated list is returned.
Similarly, if a dataset variable is a cell array with multiple columns
or is an n-D cell array, then subscripting expressions such as
A{1,'CellVar'} result in subsref returning the contents of multiple
cells. You should explicitly assign to output arguments on the
left-hand side of an assignment, for example, [cellval1,cellval2]
= A{1,'CellVar'}.
See Also dataset, set, subsasgn
18-1357
gmdistribution.subsref
Purpose Subscripted reference for Gaussian mixture distribution object
Syntax B = subsref(T,S)
Description B = subsref(T,S) is called for the syntax T(X) when T is a
gmdistribution object. S is a structure array with the following fields:
type String containing '()', '{}', or '.' specifying the
subscript type.
subs Cell array or string containing the actual subscripts.
See Also gmdistribution
18-1358
NaiveBayes.subsref
Purpose Subscripted reference for NaiveBayes object
Syntax b = subsref(nb,s)
Description b = subsref(nb,s) is called for the syntax nb(s) when nb is a
NaiveBayes object. S is a structure array with the fields:
type string containing ’()’, ’{}’, or ’.’
specifying the subscript type.
subs Cell array or string containing
the actual subscripts.
18-1359
qrandset.subsref
Purpose Subscripted reference for qrandset
Syntax x = p(i,j)
x = subsref(p,s)
Description x = p(i,j) returns a matrix that contains a subset of the points from
the point set p. The indices in i select points from the set and the
indices in j select columns from those points. i and j are vector of
positive integers or logical vectors. A colon used as a subscript, as in
p(i,:), indicates the entire row (or column).
x = subsref(p,s) is called for the syntax p(i), p{i}, or p.i. s is a
structure array with the fields:
type string containing ’()’, ’{}’, or ’.’ specifying the subscript
type.
subs Cell array or string containing the actual subscripts.
Examples Command Returns
p = sobolset(5); The fifth point
x = p(1:10,:) All columns of the first 10 points
x = p(end,1) The first column of the last point
x = p([1,4,5], :) Points 1, 4, and 5
See Also qrandset
18-1360
categorical.summary
Purpose Summary statistics for categorical array
Syntax summary(A)
C = summary(A)
[C,labels] = summary(A)
Description summary(A) displays the number of elements in the categorical array
A equal to each of the possible levels in A. If A contains any undefined
elements, the output also includes the number of undefined elements.
C = summary(A) returns counts of the number of elements in the
categorical array A equal to each of the possible levels in A. If A is
a matrix or N-dimensional array, C is a matrix or array with rows
corresponding to the levels of A. If A contains any undefined elements, C
contains one more row than the number of levels of A, with the number
of undefined elements in c(end) or c(end,:).
[C,labels] = summary(A) also returns the list of categorical level
labels corresponding to the counts in C.
Examples Count the number of patients in each age group in the data in
hospital.mat:
load hospital
edges = 0:10:100;
labels = strcat(num2str((0:10:90)','%d'),{'s'});
AgeGroup = ordinal(hospital.Age,labels,[],edges);
[c,labels] = summary(AgeGroup);
Table = dataset({labels,'AgeGroup'},{c,'Count'});
Table(3:6,:)
ans =
AgeGroup Count
'20s' 15
'30s' 41
'40s' 42
'50s' 2
18-1361
categorical.summary
See Also islevel, ismember, levelcounts
18-1362
dataset.summary
Purpose Print summary of dataset array
Syntax summary(A)
s = summary(A)
Description summary(A) prints a summary of a dataset array and the variables
that it contains.
s = summary(A) returns a scalar structure s that contains a summary
of the dataset A and the variables that A contains. For more information
on the fields in s, see Outputs.
Summary information depends on the type of the variables in the data
set:
• For numerical variables, summary computes a five-number summary
of the data, giving the minimum, the first quartile, the median, the
third quartile, and the maximum.
• For logical variables, summary counts the number of trues and
falses in the data.
• For categorical variables, summary counts the number of data at
each level.
Output The following list describes the fields in the structure s:
Arguments
• Description — A character array containing the dataset description.
• Variables — A structure array with one element for each dataset
variable in A. Each element has the following fields:
- Name — A character string containing the name of the variable.
- Description — A character string containing the variable’s
description.
- Units — A character string containing the variable’s units.
- Size — A numeric vector containing the size of the variable.
18-1363
dataset.summary
- Class — A character string containing the class of the variable.
- Data — A scalar structure containing the following fields.
For numeric variables:
• Probabilities — A numeric vector containing the probabilities
[0.0 .25 .50 .75 1.0] and NaN (if any are present in the
corresponding dataset variable).
• Quantiles — A numeric vector containing the values that
correspond to ’Probabilities’ for the corresponding dataset
variable, and a count of NaNs (if any are present).
For logical variables:
• Values — The logical vector [true false].
• Counts — A numeric vector of counts for each logical value.
For categorical variables:
• Levels — A cell array containing the labels for each level of the
corresponding dataset variable.
• Counts — A numeric vector of counts for each level.
'Data' is empty if variable is not numeric, categorical, or
logical. If a dataset variable has more than one column, then the
corresponding 'Quantiles' or 'Counts' field is a matrix or an
array.
Examples Summarize Fisher’s iris data:
load fisheriris
species = nominal(species);
data = dataset(species,meas);
summary(data)
species: [150x1 nominal]
setosa versicolor virginica
50 50 50
18-1364
dataset.summary
meas: [150x4 double]
min 4.3000 2 1 0.1000
1st Q 5.1000 2.8000 1.6000 0.3000
median 5.8000 3 4.3500 1.3000
3rd Q 6.4000 3.3000 5.1000 1.8000
max 7.9000 4.4000 6.9000 2.5000
Summarize the data in hospital.mat:
load hospital
summary(hospital)
Dataset array created from the data file hospital.dat.
The first column of the file ("id") is used for observation
names. Other columns ("sex" and "smoke") have been
converted from their original coded values into categorical
and logical variables. Two sets of columns ("sys" and
"dia", "trial1" through "trial4") have been combined into
single variables with multivariate observations. Column
headers have been replaced with more descriptive variable
names. Units have been added where appropriate.
LastName: [100x1 cell string]
Sex: [100x1 nominal]
Female Male
53 47
Age: [100x1 double, Units = Yrs]
min 1st Q median 3rd Q max
25 32 39 44 50
Weight: [100x1 double, Units = Lbs]
min 1st Q median 3rd Q max
111 130.5000 142.5000 180.5000 202
18-1365
dataset.summary
Smoker: [100x1 logical]
true false
34 66
BloodPressure: [100x2 double, Units = mm Hg]
Systolic/Diastolic
min 109 68
1st Q 117.5000 77.5000
median 122 81.5000
3rd Q 127.5000 89
max 138 99
Trials: [100x1 cell, Units = Counts]
From zero to four measurement trials performed
See Also get, set, grpstats
18-1366
ProbDist.Support property
Purpose Read-only structure containing information about support of ProbDist
object
Description Support is a read-only property of the ProbDist class. Support is a
structure containing information about the support of a ProbDist object.
It includes the following fields:
• range
• closedbound
• iscontinuous
Values The values for the three fields in the structure are:
• range — A two-element vector [L, U], such that all of the probability
is contained from L to U.
• closedbound — A two-element logical vector indicating whether the
corresponding range endpoint is included. Possible values for each
endpoint are 1 (true) or 0 (false).
• iscontinuous — A logical value indicates if the distribution takes
values on the entire interval from L to U (true), or if it takes only
integer values within this range (false). Possible values are 1 (true)
or 0 (false).
Use this information to view and compare information about the
support of distributions.
18-1367
surfht
Purpose Interactive contour plot
Syntax surfht(Z)
surfht(x,y,Z)
Description surfht(Z) is an interactive contour plot of the matrix Z treating the
values in Z as height above the plane. The x-values are the column
indices of Z while the y-values are the row indices of Z.
surfht(x,y,Z) where x and y are vectors specify the x and y-axes on
the contour plot. The length of x must match the number of columns in
Z, and the length of y must match the number of rows in Z.
There are vertical and horizontal reference lines on the plot whose
intersection defines the current x value and y value. You can drag
these dotted white reference lines and watch the interpolated z value
(at the top of the plot) update simultaneously. Alternatively, you can
get a specific interpolated z value by typing the x value and y value into
editable text fields on the x-axis and y-axis respectively.
18-1368
tabulate
Purpose Frequency table
Syntax TABLE = tabulate(x)
tabulate(x)
Description TABLE = tabulate(x) creates a frequency table of data in vector x.
Information in TABLE is arranged as follows:
• 1st column — The unique values of x
• 2nd column — The number of instances of each value
• 3rd column — The percentage of each value
If x is a numeric array, TABLE is a numeric matrix. If the elements of x
are nonnegative integers, TABLE includes 0 counts for integers between
1 and max(x) that do not appear in x.
If x is a categorical variable, character array, or cell array of strings,
TABLE is a cell array.
tabulate(x) with no output arguments displays the table in the
command window.
Examples tabulate([1 2 4 4 3 4])
Value Count Percent
1 1 16.67%
2 1 16.67%
3 1 16.67%
4 3 50.00%
See Also “Grouped Data” on page 2-34
pareto
18-1369
tblread
Purpose Read tabular data from file
Syntax [data,varnames,casenames] = tblread
[data,varnames,casenames] = tblread(filename)
[data,varnames,casenames] = tblread(filename,delimiter)
Description [data,varnames,casenames] = tblread displays the File Open dialog
box for interactive selection of a tabular data file. The file format has
variable names in the first row, case names in the first column and data
starting in the (2, 2) position. Outputs are:
• data — Numeric matrix with a value for each variable-case pair
• varnames — String matrix containing the variable names in the first
row of the file
• casenames — String matrix containing the names of each case in
the first column of the file
[data,varnames,casenames] = tblread(filename) allows command
line specification of the name of a file in the current folder, or the
complete path name of any file, using the string filename.
[data,varnames,casenames] = tblread(filename,delimiter)
reads from the file using delimiter as the delimiting character.
Accepted values for delimiter are:
• ' ' or 'space'
• '\t' or 'tab'
• ',' or 'comma'
• ';' or 'semi'
• '|' or 'bar'
The default value of delimiter is 'space'.
18-1370
tblread
Examples [data,varnames,casenames] = tblread('sat.dat')
data =
470 530
520 480
varnames =
Male
Female
casenames =
Verbal
Quantitative
See Also tblwrite, tdfread, caseread
18-1371
tblwrite
Purpose Write tabular data to file
Syntax tblwrite(data,varnames,casenames)
tblwrite(data,varnames,casenames,filename)
tblwrite(data,varnames,casenames,filename,delimiter)
Description tblwrite(data,varnames,casenames) displays the File Open dialog
box for interactive specification of the tabular data output file. The
file format has variable names in the first row, case names in the first
column and data starting in the (2,2) position.
varnames is a string matrix containing the variable names. casenames
is a string matrix containing the names of each case in the first column.
data is a numeric matrix with a value for each variable-case pair.
tblwrite(data,varnames,casenames,filename) specifies a file in
the current folder, or the complete path name of any file in the string
filename.
tblwrite(data,varnames,casenames,filename,delimiter) writes
to the file using delimiter as the delimiting character. The following
table lists the accepted character values for delimiter and their
equivalent string values.
Character String
' ' 'space'
'\t' 'tab'
',' 'comma'
';' 'semi'
'|' 'bar'
The default value of delimiter is 'space'.
Examples Continuing the example from tblread:
tblwrite(data,varnames,casenames,'sattest.dat')
18-1372
tblwrite
type sattest.dat
Male Female
Verbal 470 530
Quantitative 520 480
See Also casewrite, tblread
18-1373
tcdf
Purpose Student’s t cumulative distribution function
Syntax P = tcdf(X,V)
Description P = tcdf(X,V) computes Student’s t cdf at each of the values in X
using the corresponding degrees of freedom in V. X and V can be vectors,
matrices, or multidimensional arrays that all have the same size. A
scalar input is expanded to a constant array with the same dimensions
as the other inputs.
The t cdf is
⎛ + 1 ⎞
Γ⎜ ⎟
x
p = F ( x | ) = ∫ ⎝ 2 ⎠ 1 1
dt
−∞ ⎛ ⎞ +1
Γ⎜ ⎟ ⎛ t2 ⎞ 2
⎝2⎠ ⎜1 + ⎟
⎜ ⎟⎠
⎝
The result, p, is the probability that a single observation from the
t distribution with ν degrees of freedom will fall in the interval [–∞, x).
Examples mu = 1; % Population mean
sigma = 2; % Population standard deviation
n = 100; % Sample size
x = normrnd(mu,sigma,n,1); % Random sample from population
xbar = mean(x); % Sample mean
s = std(x); % Sample standard deviation
t = (xbar-mu)/(s/sqrt(n)) % t-statistic
t =
0.2489
p = 1-tcdf(t,n-1) % Probability of larger t-statistic
p =
0.4020
This probability is the same as the p value returned by a t-test of the
null hypothesis that the sample comes from a normal population with
mean μ:
18-1374
tcdf
[h,ptest] = ttest(x,mu,0.05,'right')
h =
0
ptest =
0.4020
See Also cdf, tpdf, tinv, tstat, trnd
“Student’s t Distribution” on page B-95
18-1375
tdfread
Purpose Read tab-delimited file
Syntax tdfread
tdfread(filename)
tdfread(filename,delimiter)
s = tdfread(filename,...)
Description tdfread displays the File Open dialog box for interactive selection
of a data file, then reads data from the file. The file should have
variable names separated by tabs in the first row, and data values
separated by tabs in the remaining rows. tdfread creates variables in
the workspace, one for each column of the file. The variable names
are taken from the first row of the file. If a column of the file contains
only numeric data in the second and following rows, tdfread creates a
double variable. Otherwise, tdfread creates a char variable. After all
values are imported, tdfread displays information about the imported
values using the format of the tdfread command.
tdfread(filename) allows command line specification of the name of a
file in the current folder, or the complete path name of any file, using
the string filename.
tdfread(filename,delimiter) indicates that the character specified
by delimiter separates columns in the file. Accepted values for
delimiter are:
• ' ' or 'space'
• '\t' or 'tab'
• ',' or 'comma'
• ';' or 'semi'
• '|' or 'bar'
The default delimiter is 'tab'.
s = tdfread(filename,...) returns a scalar structure s whose fields
each contain a variable.
18-1376
tdfread
Examples The following displays the contents of the file sat2.dat:
type sat2.dat
Test,Gender,Score
Verbal,Male,470
Verbal,Female,530
Quantitative,Male,520
Quantitative,Female,480
The following creates the variables Gender, Score, and Test from the
file sat2.dat and displays the contents of the MATLAB workspace:
tdfread('sat2.dat',',')
Name Size Bytes Class Attributes
Gender 4x6 48 char
Score 4x1 32 double
Test 4x12 96 char
See Also tblread, caseread
18-1377
classregtree.test
Purpose Error rate
Syntax cost = test(t,'resubstitution')
cost = test(t,'test',X,y)
cost = test(t,'crossvalidate',X,y)
[cost,secost,ntnodes,bestlevel] = test(...)
[...] = test(...,param1,val1,param2,val2,...)
Description cost = test(t,'resubstitution') computes the cost of the tree
t using a resubstitution method. t is a decision tree as created by
classregtree. The cost of the tree is the sum over all terminal nodes
of the estimated probability of a node times the cost of a node. If t is a
classification tree, the cost of a node is the sum of the misclassification
costs of the observations in that node. If t is a regression tree, the cost
of a node is the average squared error over the observations in that
node. cost is a vector of cost values for each subtree in the optimal
pruning sequence for t. The resubstitution cost is based on the same
sample that was used to create the original tree, so it under estimates
the likely cost of applying the tree to new data.
cost = test(t,'test',X,y) uses the matrix of predictors X and the
response vector y as a test sample, applies the decision tree t to that
sample, and returns a vector cost of cost values computed for the test
sample. X and y should not be the same as the learning sample, that
is, the sample that was used to fit the tree t.
cost = test(t,'crossvalidate',X,y) uses 10-fold cross-validation
to compute the cost vector. X and y should be the learning sample, that
is, the sample that was used to fit the tree t. The function partitions the
sample into 10 subsamples, chosen randomly but with roughly equal
size. For classification trees, the subsamples also have roughly the same
class proportions. For each subsample, test fits a tree to the remaining
data and uses it to predict the subsample. It pools the information from
all subsamples to compute the cost for the whole sample.
[cost,secost,ntnodes,bestlevel] = test(...) also returns the
vector secost containing the standard error of each cost value, the
vector ntnodes containing the number of terminal nodes for each
18-1378
classregtree.test
subtree, and the scalar bestlevel containing the estimated best level
of pruning. A bestlevel of 0 means no pruning. The best level is the
one that produces the smallest tree that is within one standard error of
the minimum-cost subtree.
[...] = test(...,param1,val1,param2,val2,...) specifies
optional parameter name/value pairs for methods other than
'resubstitution', chosen from the following:
• 'weights' — Observation weights.
• 'nsamples' — The number of cross-validation samples (default is
10).
• 'treesize' — Either 'se' (default) to choose the smallest tree
whose cost is within one standard error of the minimum cost, or
'min' to choose the minimal cost tree.
Examples Find the best tree for Fisher’s iris data using cross-validation. Start
with a large tree:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'},...
'splitmin',5)
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 if PW<1.55 then node 10 elseif PW>=1.55 then node 11 else virginica
8 class = versicolor
9 class = virginica
10 class = virginica
18-1379
classregtree.test
11 class = versicolor
view(t)
Find the minimum-cost tree:
[c,s,n,best] = test(t,'cross',meas,species);
tmin = prune(t,'level',best)
tmin =
Decision tree for classification
18-1380
classregtree.test
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 class = versicolor
5 class = virginica
view(tmin)
Plot the smallest tree within one standard error of the minimum cost
tree:
18-1381
classregtree.test
[mincost,minloc] = min(c);
plot(n,c,'b-o',...
n(best+1),c(best+1),'bs',...
n,(mincost+s(minloc))*ones(size(n)),'k--')
xlabel('Tree size (number of terminal nodes)')
ylabel('Cost')
The solid line shows the estimated cost for each tree size, the dashed
line marks one standard error above the minimum, and the square
marks the smallest tree under the dashed line.
18-1382
classregtree.test
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree, eval, view, prune
18-1383
cvpartition.test
Purpose Test indices for cross-validation
Syntax idx = test(c)
idx = test(c,i)
Description idx = test(c) returns the logical vector idx of test indices for an object
c of the cvpartition class of type 'holdout' or 'resubstitution'.
If c.Type is 'holdout', idx specifies the observations in the test set.
If c.Type is 'resubstitution', idx specifies all observations.
idx = test(c,i) returns the logical vector idx of test indices for
repetition i of an object c of the cvpartition class of type 'kfold'
or 'leaveout'.
If c.Type is 'kfold', idx specifies the observations in the test set in
fold i.
If c.Type is 'leaveout', idx specifies the observation left out at
repetition i.
Examples Identify the test indices in the first fold of a partition of 10 observations
for 3-fold cross-validation:
c = cvpartition(10,'kfold',3)
c =
K-fold cross validation partition
N: 10
NumTestSets: 3
TrainSize: 7 6 7
TestSize: 3 4 3
test(c,1)
ans =
1
1
0
0
18-1384
cvpartition.test
0
0
0
0
1
0
See Also cvpartition, training
18-1385
cvpartition.TestSize property
Purpose Size of each test set
Description Value is a vector in partitions of type 'kfold' and 'leaveout'.
Value is a scalar in partitions of type 'holdout' and
'resubstitution'.
18-1386
tiedrank
Purpose Rank adjusted for ties
Syntax [R,TIEADJ] = tiedrank(X)
[R,TIEADJ] = tiedrank(X,1)
[R,TIEADJ] = tiedrank(X,0,1)
Description [R,TIEADJ] = tiedrank(X) computes the ranks of the values in the
vector X. If any X values are tied, tiedrank computes their average
rank. The return value TIEADJ is an adjustment for ties required by the
nonparametric tests signrank and ranksum, and for the computation
of Spearman’s rank correlation.
[R,TIEADJ] = tiedrank(X,1) computes the ranks of the values in
the vector X. TIEADJ is a vector of three adjustments for ties required
in the computation of Kendall’s tau. tiedrank(X,0) is the same as
tiedrank(X).
[R,TIEADJ] = tiedrank(X,0,1) computes the ranks from each end, so
that the smallest and largest values get rank 1, the next smallest and
largest get rank 2, etc. These ranks are used in the Ansari-Bradley test.
Examples Counting from smallest to largest, the two 20 values are 2nd and 3rd,
so they both get rank 2.5 (average of 2 and 3):
tiedrank([10 20 30 40 20])
ans =
1.0000 2.5000 4.0000 5.0000 2.5000
See Also ansaribradley, corr, partialcorr, ranksum, signrank
18-1387
categorical.times
Purpose Product of categorical arrays
Syntax C = times(A,B)
Description C = times(A,B) returns a categorical array each of whose elements
has the level formed from the concatenation of the levels of the
corresponding elements of A and B. The set of levels of C is the cartesian
product of the sets of levels of A and of B. The syntax A .* B calls C
= times(A,B).
See Also categorical
18-1388
tinv
Purpose Student’s t inverse cumulative distribution function
Syntax X = tinv(P,V)
Description X = tinv(P,V) computes the inverse of Student’s t cdf using the degrees
of freedom in V for the corresponding probabilities in P. P and V can be
vectors, matrices, or multidimensional arrays that are the same size. A
scalar input is expanded to a constant array with the same dimensions
as the other inputs. The values in P must lie on the interval [0 1].
The t inverse function in terms of the t cdf is
x = F −1 ( p| ) = { x : F ( x | ) = p}
where
⎛ + 1 ⎞
Γ⎜ ⎟
p = F ( x | ) = ∫
x ⎝ 2 ⎠ 1 1
dt
−∞ ⎛ ⎞ +1
Γ⎜ ⎟ ⎛ t ⎞ 2
2
⎝2⎠ ⎜1 + ⎟
⎜ ⎟⎠
⎝
The result, x, is the solution of the cdf integral with parameter ν, where
you supply the desired probability p.
Examples What is the 99th percentile of the t distribution for one to six degrees
of freedom?
percentile = tinv(0.99,1:6)
percentile =
31.8205 6.9646 4.5407 3.7469 3.3649 3.1427
See Also icdf, tcdf, tpdf, trnd, tstat
“Student’s t Distribution” on page B-95
18-1389
tpdf
Purpose Student’s t probability density function
Syntax Y = tpdf(X,V)
Description Y = tpdf(X,V) computes Student’s t pdf at each of the values in X
using the corresponding degrees of freedom in V. X and V can be vectors,
matrices, or multidimensional arrays that have the same size. A scalar
input is expanded to a constant array with the same dimensions as
the other inputs.
Student’s t pdf is
⎛ + 1 ⎞
Γ⎜
2 ⎟⎠ 1
y = f ( x | ) = ⎝
1
⎛ ⎞ +1
Γ⎜ ⎟ ⎛ x 2⎞ 2
⎝2⎠ ⎜1 + ⎟
⎜ ⎟⎠
⎝
Examples The mode of the t distribution is at x = 0. This example shows that
the value of the function at the mode is an increasing function of the
degrees of freedom.
tpdf(0,1:6)
ans =
0.3183 0.3536 0.3676 0.3750 0.3796 0.3827
The t distribution converges to the standard normal distribution as the
degrees of freedom approaches infinity. How good is the approximation
for v = 30?
difference = tpdf(-2.5:2.5,30)-normpdf(-2.5:2.5)
difference =
0.0035 -0.0006 -0.0042 -0.0042 -0.0006 0.0035
See Also pdf, tcdf, tinv, tstat, trnd
“Student’s t Distribution” on page B-95
18-1390
cvpartition.training
Purpose Training indices for cross-validation
Syntax idx = training(c)
idx = training(c,i)
Description idx = training(c) returns the logical vector idx of training indices
for an object c of the cvpartition class of type 'holdout' or
'resubstitution'.
If c.Type is 'holdout', idx specifies the observations in the training
set.
If c.Type is 'resubstitution', idx specifies all observations.
idx = training(c,i) returns the logical vector idx of training indices
for repetition i of an object c of the cvpartition class of type 'kfold'
or 'leaveout'.
If c.Type is 'kfold', idx specifies the observations in the training
set in fold i.
If c.Type is 'leaveout', idx specifies the observations left in at
repetition i.
Examples Identify the training indices in the first fold of a partition of 10
observations for 3-fold cross-validation:
c = cvpartition(10,'kfold',3)
c =
K-fold cross validation partition
N: 10
NumTestSets: 3
TrainSize: 7 6 7
TestSize: 3 4 3
training(c,1)
ans =
0
0
18-1391
cvpartition.training
1
1
1
1
1
1
0
1
See Also cvpartition, test
18-1392
cvpartition.TrainSize property
Purpose Size of each training set
Description Value is a vector in partitions of type 'kfold' and 'leaveout'.
Value is a scalar in partitions of type 'holdout' and
'resubstitution'.
See Also type
18-1393
categorical.transpose
Purpose Transpose categorical matrix
Syntax B = transpose(A)
Description B = transpose(A) returns the transpose of the 2-D categorical matrix
A. ctranspose is identical to transpose for categorical arrays. The
syntax A.' calls transpose.
See Also ctranspose, permute
18-1394
TreeBagger.TreeArgs property
Purpose Cell array of arguments for classregtree
Description The TreeArgs property is a cell array of arguments for the
classregtree constructor. TreeBagger uses these arguments in
growing new trees for the ensemble.
18-1395
TreeBagger class
Purpose Bootstrap aggregation for ensemble of decision trees
Description TreeBagger bags an ensemble of decision trees for either classification
or regression. Bagging stands for bootstrap aggregation. Every tree in
the ensemble is grown on an independently drawn bootstrap replica of
input data. Observations not included in this replica are "out of bag"
for this tree. To compute prediction of an ensemble of trees for unseen
data, TreeBagger takes an average of predictions from individual
trees. To estimate the prediction error of the bagged ensemble, you
can compute predictions for each tree on its out-of-bag observations,
average these predictions over the entire ensemble for each observation
and then compare the predicted out-of-bag response with the true value
at this observation.
TreeBagger relies on the classregtree functionality for growing
individual trees. In particular, classregtree accepts the number of
features selected at random for each decision split as an optional input
argument.
The Compact property contains another class, CompactTreeBagger,
with sufficient information to make predictions using new data. This
information includes the tree ensemble, variable names, and class
names (for classification). CompactTreeBagger requires less memory
than TreeBagger, but only TreeBagger has methods for growing more
trees for the ensemble. Once you grow an ensemble of trees using
TreeBagger and no longer need access to the training data, you can opt
to work with the compact version of the trained ensemble from then on.
Construction TreeBagger Create ensemble of bagged
decision trees
Methods append Append new trees to ensemble
compact Compact ensemble of decision
trees
18-1396
TreeBagger class
error Error (misclassification
probability or MSE)
fillProximities Proximity matrix for training
data
growTrees Train additional trees and add to
ensemble
margin Classification margin
mdsProx Multidimensional scaling of
proximity matrix
meanMargin Mean classification margin
oobError Out-of-bag error
oobMargin Out-of-bag margins
oobMeanMargin Out-of-bag mean margins
oobPredict Ensemble predictions for
out-of-bag observations
predict Predict response
Properties ClassNames Names of classes
ComputeOOBPrediction Flag to compute out-of-bag
predictions
ComputeOOBVarImp Flag to compute out-of-bag
variable importance
Cost Misclassification costs
DefaultYfit Default value returned by
predict and oobPredict
DeltaCritDecisionSplit Split criterion contributions for
each predictor
18-1397
TreeBagger class
FBoot Fraction of in-bag observations
MergeLeaves Flag to merge leaves that do not
improve risk
Method Method used by trees
(classification or regression)
MinLeaf Minimum number of observations
per tree leaf
NTrees Number of decision trees in
ensemble
NVarToSample Number of variables for random
feature selection
OOBIndices Indicator matrix for out-of-bag
observations
OOBInstanceWeight Count of out-of-bag trees for each
observation
OOBPermutedVarCountRaiseMargin
Variable importance for raising
margin
OOBPermutedVarDeltaError Variable importance for
classification error
OOBPermutedVarDeltaMeanMargin
Variable importance for
classification margin
OutlierMeasure Measure for determining outliers
Prior Prior class probabilities
Proximity Proximity matrix for observations
Prune Flag to prune trees
SampleWithReplacement Flag to sample with replacement
TreeArgs Cell array of arguments for
classregtree
Trees Decision trees in ensemble
18-1398
TreeBagger class
VarNames Variable names
X X data used to create ensemble
Y Y data used to create ensemble
Copy Value. To learn how this affects your use of the class, see Comparing
Semantics Handle and Value Classes in the MATLAB Object-Oriented
Programming documentation.
See Also “Regression and Classification by Bagging Decision Trees” on page 12-30
Classification Trees
Regression Tress
Grouped Data
18-1399
TreeBagger
Purpose Create ensemble of bagged decision trees
Syntax B = TreeBagger(ntrees,X,Y)
B = TreeBagger(ntrees,X,Y,’param1’,val1,'param2',val2,...)
Description B = TreeBagger(ntrees,X,Y) creates an ensemble B of ntrees
decision trees for predicting response Y as a function of predictors
X. By default TreeBagger builds an ensemble of classification trees.
The function can build an ensemble of regression trees by setting the
optional input argument 'method' to 'regression'.
X is a numeric matrix of training data. Each row represents an
observation and each column represents a predictor or feature. Y is an
array of true class labels for classification or numeric function values for
regression. True class labels can be a numeric vector, character matrix,
vector cell array of strings or categorical vector. TreeBagger converts
labels to a cell array of strings for classification.
B = TreeBagger(ntrees,X,Y,’param1’,val1,'param2',val2,...)
specifies optional parameter name/value pairs:
'FBoot' Fraction of input data to sample with replacement
from the input data for growing each new tree.
Default value is 1.
'OOBPred' 'on' to store info on what observations are
out of bag for each tree. This info can be used
by oobPredict to compute the predicted class
probabilities for each tree in the ensemble. Default
is 'off'.
'OOBVarImp' 'on' to store out-of-bag estimates of feature
importance in the ensemble. Default is 'off'.
Specifying 'on' also sets the 'OOBPred' value to
'on'.
'Method' Either 'classification' or 'regression'.
Regression requires a numeric Y.
18-1400
TreeBagger
'NVarToSample' Number of variables to select at random for each
decision split. Default is the square root of the
number of variables for classification and one third
of the number of variables for regression. Valid
values are 'all' or a positive integer. Setting this
argument to any valid value but 'all' invokes
Breiman’s ’random forest’ algorithm.
'NPrint' Number of training cycles (grown trees) after
which TreeBagger displays a diagnostic message
showing training progress. Default is no diagnostic
messages.
'MinLeaf' Minimum number of observations per tree leaf.
Default is 1 for classification and 5 for regression.
'Options' A struct that specifies options that govern the
computation when growing the ensemble of
decision trees. One option requests that the
computation of decision trees on multiple bootstrap
replicates uses multiple processors, if the Parallel
Computing Toolbox is available. Two options
specify the random number streams to use in
selecting bootstrap replicates. You can create
this argument with a call to statset. You can
retrieve values of the individual fields with a call
to statget. Applicable statset parameters are:
• 'UseParallel' — If 'always' and if a
matlabpool of the Parallel Computing Toolbox
is open, compute decision trees drawn on
separate boostrap replicates in parallel. If the
Parallel Computing Toolbox is not installed, or
a matlabpool is not open, computation occurs
in serial mode. Default is 'never', or serial
computation.
• 'UseSubstreams' — If 'always' select each
bootstrap replicate using a separate Substream
18-1401
TreeBagger
of the random number generator (aka Stream).
This option is available only with RandStream
types that support Substreams. Default is
'never', do not use a different Substream to
compute each bootstrap replicate.
• 'Streams' — An object of the RandStream class,
or a cell array of RandStream objects. Default is
an empty cell array. If you do not supply a value
for this parameter, TreeBagger uses the default
RandStream on each MATLAB executable
in selecting bootstrap replicates. Otherwise,
TreeBagger selects bootstrap replicates using
the supplied RandStream object(s). If you select
'UseSubstreams', the Streams parameter, if
present, must be a scalar RandStream object. If
you do not select 'UseSubstreams', then the
Streams parameter, if present, must match the
number of processors used for the computation.
For serial computation, the Streams parameter
must be a scalar. If computation is distributed
('UseParallel' is 'always' and a matlabpool
is open), then the Streams parameter must be a
cell array of the same length as the matlabpool
size. In this case, each element of the cell
array supplies the random number generator
for bootstrap sampling on one of the parallel
workers.
In addition to the optional arguments above, this method accepts all
optional classregtree arguments with the exception of 'minparent'.
Refer to the documentation for classregtree for more detail.
Examples load fisheriris
b = TreeBagger(50,meas,species,'OOBPred','on')
plot(oobError(b))
xlabel('number of grown trees')
ylabel('out-of-bag classification error')
18-1402
TreeBagger
returns
b =
Ensemble with 50 bagged decision trees:
Training X: [150x4]
Training Y: [150x1]
Method: classification
Nvars: 4
NVarToSample: 2
MinLeaf: 1
FBoot: 1
SampleWithReplacement: 1
ComputeOOBPrediction: 1
ComputeOOBVarImp: 0
Proximity: []
Prune: 0
MergeLeaves: 0
TreeArgs:
ClassNames:'setosa' 'versicolor' 'virginica'
18-1403
TreeBagger
See Also “Regression and Classification by Bagging Decision Trees” on page
12-30, “Grouped Data” on page 2-34
classregtree, CompactTreeBagger
18-1404
treedisp
Purpose Plot tree
Note treedisp will be removed in a future release. Use
classregtree.view instead.
Syntax treedisp(t)
treedisp(t,param1,val1,param2,val2,...)
Description
Note This function is superseded by the view method of the
classregtree class and is maintained only for backwards compatibility.
It accepts objects t created with the classregtree constructor.
treedisp(t) takes as input a decision tree t as computed by the
treefit function, and displays it in a figure window. Each branch in
the tree is labeled with its decision rule, and each terminal node is
labeled with the predicted value for that node.
For each branch node, the left child node corresponds to the points that
satisfy the condition, and the right child node corresponds to the points
that do not satisfy the condition.
The Click to display pop-up menu at the top of the figure enables
you to display more information about each node, as described in the
following table.
Menu Choice Displays
Identity The node number, whether the node is a branch
or a leaf, and the rule that governs the node
Variable ranges The range of each of the predictor variables for
that node
Node statistics Descriptive statistics for the observations falling
into this node
18-1405
treedisp
After you select the type of information you want, click any node to
display the information for that node.
The Pruning level button displays the number of levels that have
been cut from the tree and the number of levels in the unpruned tree.
For example, 1 of 6 indicates that the unpruned tree has six levels,
and that one level has been cut from the tree. Use the spin button to
change the pruning level.
treedisp(t,param1,val1,param2,val2,...) specifies optional
parameter name-value pairs, listed in the following table.
Parameter Value
'names' A cell array of names for the predictor variables,
in the order in which they appear in the X matrix
from which the tree was created (see treefit)
'prunelevel' Initial pruning level to display
Examples Create and graph classification tree for Fisher’s iris data. The names in
this example are abbreviations for the column contents (sepal length,
sepal width, petal length, and petal width).
load fisheriris;
t = treefit(meas,species);
treedisp(t,'names',{'SL' 'SW' 'PL' 'PW'});
18-1406
treedisp
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also treefit, treeprune, treetest
18-1407
treefit
Purpose Fit tree
Note treefit will be removed in a future release. Use classregtree
instead.
Syntax t = treefit(X,y)
t = treefit(X,y,param1,val1,param2,val2,...)
Description
Note This function is superseded by the classregtree constructor
of the classregtree class and is maintained only for backwards
compatibility. It returns objects t in the classregtree class.
t = treefit(X,y) creates a decision tree t for predicting response y as
a function of predictors X. X is an n-by-m matrix of predictor values. y is
either a vector of n response values (for regression), or a character array
or cell array of strings containing n class names (for classification).
Either way, t is a binary tree where each non-terminal node is split
based on the values of a column of X.
t = treefit(X,y,param1,val1,param2,val2,...) specifies optional
parameter name-value pairs. Valid parameter strings are:
The following table lists parameters available for all trees.
Parameter Value
'catidx' Vector of indices of the columns of X. treefit
treats these columns as unordered categorical
values.
'method' Either 'classification' (default if y is text) or
'regression' (default if y is numeric).
18-1408
treefit
Parameter Value
'splitmin' A number n such that impure nodes must have n
or more observations to be split (default 10).
'prune' 'on' (default) to compute the full tree and a
sequence of pruned subtrees, or 'off' for the full
tree without pruning.
The following table lists parameters available for classification trees
only.
Parameter Value
'cost' p-by-p matrix C, where p is the number of distinct
response values or class names in the input y.
C(i,j) is the cost of classifying a point into class
j if its true class is i. (The default has C(i,j)=1
if i~=j, and C(i,j)=0 if i=j.) C can also be a
structure S with two fields: S.group containing
the group names (see “Grouped Data” on page
2-34), and S.cost containing a matrix of cost
values.
'splitcriterion' Criterion for choosing a split: either 'gdi'
(default) for Gini’s diversity index, 'twoing' for
the twoing rule, or 'deviance' for maximum
deviance reduction.
'priorprob' Prior probabilities for each class, specified as a
vector (one value for each distinct group name)
or as a structure S with two fields: S.group
containing the group names, and S.prob
containing a vector of corresponding probabilities.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = treefit(meas,species);
18-1409
treefit
treedisp(t,'names',{'SL' 'SW' 'PL' 'PW'});
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also “Grouped Data” on page 2-34
treedisp, treetest
18-1410
treeprune
Purpose Prune tree
Note treeprune will be removed in a future release. Use
classregtree.prune instead.
Syntax t2 = treeprune(t1,'level',level)
t2 = treeprune(t1,'nodes',nodes)
t2 = treeprune(t1)
Description
Note This function is superseded by the prune method of the
classregtree class and is maintained only for backwards compatibility.
It accepts objects t1 created with the classregtree constructor and
returns objects t2 in the classregtree class.
t2 = treeprune(t1,'level',level) takes a decision tree t1 as
created by the treefit function, and a pruning level, and returns the
decision tree t2 pruned to that level. Setting level to 0 means no
pruning. Trees are pruned based on an optimal pruning scheme that
first prunes branches giving less improvement in error cost.
t2 = treeprune(t1,'nodes',nodes) prunes the nodes listed in the
nodes vector from the tree. Any t1 branch nodes listed in nodes
become leaf nodes in t2, unless their parent nodes are also pruned. The
treedisp function can display the node numbers for any node you select.
t2 = treeprune(t1) returns the decision tree t2 that is the same as
t1, but with the optimal pruning information added. This is useful
only if you created t1 by pruning another tree, or by using the treefit
function with pruning set 'off'. If you plan to prune a tree multiple
times, it is more efficient to create the optimal pruning sequence first.
Pruning is the process of reducing a tree by turning some branch nodes
into leaf nodes, and removing the leaf nodes under the original branch.
18-1411
treeprune
Examples Display the full tree for Fisher’s iris data, as well as the next largest
tree from the optimal pruning sequence:
load fisheriris;
t1 = treefit(meas,species,'splitmin',5);
treedisp(t1,'names',{'SL' 'SW' 'PL' 'PW'});
t2 = treeprune(t1,'level',1);
treedisp(t2,'names',{'SL' 'SW' 'PL' 'PW'});
18-1412
treeprune
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also treefit, treetest, treedisp
18-1413
CompactTreeBagger.Trees property
Purpose Decision trees in ensemble
Description The Trees property is a cell array of size NTrees-by-1 containing the
trees in the ensemble.
See Also NTrees
18-1414
TreeBagger.Trees property
Purpose Decision trees in ensemble
Description The Trees property is a cell array of size NTrees-by-1 containing the
trees in the ensemble.
See Also NTrees
18-1415
treetest
Purpose Error rate
Syntax cost = treetest(t,'resubstitution')
cost = treetest(t,'test',X,y)
cost = treetest(t,'crossvalidate',X,y)
[cost,secost,ntnodes,bestlevel] = treetest(...)
[...] = treetest(...,param1,val1,param2,val2,...)
Note treetest will be removed in a future release. Use
classregtree.test instead.
Description
Note This function is superseded by the test method of the
classregtree class class and is maintained only for backwards
compatibility. It accepts objects t created with the classregtree
constructor.
cost = treetest(t,'resubstitution') computes the cost of the tree
t using a resubstitution method. t is a decision tree as created by the
treefit function. The cost of the tree is the sum over all terminal nodes
of the estimated probability of that node times the node’s cost. If t is a
classification tree, the cost of a node is the sum of the misclassification
costs of the observations in that node. If t is a regression tree, the cost
of a node is the average squared error over the observations in that
node. cost is a vector of cost values for each subtree in the optimal
pruning sequence for t. The resubstitution cost is based on the same
sample that was used to create the original tree, so it underestimates
the likely cost of applying the tree to new data.
cost = treetest(t,'test',X,y) uses the predictor matrix X and
response y as a test sample, applies the decision tree t to that sample,
and returns a vector cost of cost values computed for the test sample.
X and y should not be the same as the learning sample, which is the
sample that was used to fit the tree t.
18-1416
treetest
cost = treetest(t,'crossvalidate',X,y) uses 10-fold
cross-validation to compute the cost vector. X and y should be the
learning sample, which is the sample that was used to fit the tree t. The
function partitions the sample into 10 subsamples, chosen randomly
but with roughly equal size. For classification trees, the subsamples
also have roughly the same class proportions. For each subsample,
treetest fits a tree to the remaining data and uses it to predict the
subsample. It pools the information from all subsamples to compute the
cost for the whole sample.
[cost,secost,ntnodes,bestlevel] = treetest(...) also returns
the vector secost containing the standard error of each cost value,
the vector ntnodes containing number of terminal nodes for each
subtree, and the scalar bestlevel containing the estimated best level
of pruning. bestlevel = 0 means no pruning, i.e., the full unpruned
tree. The best level is the one that produces the smallest tree that is
within one standard error of the minimum-cost subtree.
[...] = treetest(...,param1,val1,param2,val2,...) specifies
optional parameter name-value pairs chosen from the following table.
Parameter Value
'nsamples' The number of cross-validations samples (default
is 10).
'treesize' Either 'se' (default) to choose the smallest tree
whose cost is within one standard error of the
minimum cost, or 'min' to choose the minimal cost
tree.
Examples Find the best tree for Fisher’s iris data using cross-validation. The
solid line shows the estimated cost for each tree size, the dashed line
marks one standard error above the minimum, and the square marks
the smallest tree under the dashed line.
% Start with a large tree.
load fisheriris;
t = treefit(meas,species','splitmin',5);
18-1417
treetest
% Find the minimum-cost tree.
[c,s,n,best] = treetest(t,'cross',meas,species);
tmin = treeprune(t,'level',best);
% Plot smallest tree within 1 std of minimum cost tree.
[mincost,minloc] = min(c);
plot(n,c,'b-o',...
n(best+1),c(best+1),'bs',...
n,(mincost+s(minloc))*ones(size(n)),'k--');
xlabel('Tree size (number of terminal nodes)')
ylabel('Cost')
18-1418
treetest
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also treefit, treedisp
18-1419
treeval
Purpose Predicted responses
Note treeval will be removed in a future release. Use
classregtree.eval instead.
Syntax yfit = treeval(t,X)
yfit = treeval(t,X,subtrees)
[yfit,node] = treeval(...)
[yfit,node,cname] = treeval(...)
Description
Note This function is superseded by the eval method of the
classregtree class and is maintained only for backwards compatibility.
It accepts objects t created with the classregtree constructor.
yfit = treeval(t,X) takes a classification or regression tree t as
produced by the treefit function and a matrix X of predictor values,
and produces a vector yfit of predicted response values. For a
regression tree, yfit(i) is the fitted response value for a point having
the predictor values X(i,:). For a classification tree, yfit(i) is the
class number into which the tree would assign the point with data
X(i,:). To convert the number into a class name, use the third output
argument, cname (described below).
yfit = treeval(t,X,subtrees) takes an additional vector subtrees
of pruning levels, with 0 representing the full, unpruned tree. T must
include a pruning sequence as created by the treefit or prunetree
function. If subtree has k elements and X has n rows, the output yfit
is an n-by-k matrix, with the jth column containing the fitted values
produced by the subtrees(j) subtree. subtrees must be sorted in
ascending order.
[yfit,node] = treeval(...) also returns an array node of the same
size as yfit containing the node number assigned to each row of X. The
treedisp function can display the node numbers for any node you select.
18-1420
treeval
[yfit,node,cname] = treeval(...) is valid only for classification
trees. It returns a cell array cname containing the predicted class names.
Examples Find the predicted classifications for Fisher’s iris data:
load fisheriris;
t = treefit(meas,species); % Create decision tree
sfit = treeval(t,meas); % Find assigned class numbers
sfit = t.classname(sfit); % Get class names
mean(strcmp(sfit,species)) % Proportion in correct class
ans =
0.9800
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also treefit, treeprune, treetest
18-1421
trimmean
Purpose Mean excluding outliers
Syntax m = trimmean(X,percent)
trimmean(X,percent,dim)
m = trimmean(X,percent,flag)
m = trimmean(x,percent,flag,dim)
Description m = trimmean(X,percent) calculates the trimmed mean of the values
in X. For a vector input, m is the mean of X, excluding the highest and
lowest k data values, where k=n*(percent/100)/2 and where n is the
number of values in X. For a matrix input, m is a row vector containing
the trimmed mean of each column of X. For n-D arrays, trimmean
operates along the first non-singleton dimension. percent is a scalar
between 0 and 100.
trimmean(X,percent,dim) takes the trimmed mean along dimension
dim of X.
m = trimmean(X,percent,flag) controls how to trim when k is not an
integer. flag can be chosen from the following:
'round' Round k to the nearest integer (round to a smaller
integer if k is a half integer). This is the default.
'floor' Round k down to the next smaller integer.
'weight' If k=i+f where i is the integer part and f is the
fraction, compute a weighted mean with weight
(1-f) for the (i+1)th and (n-i)th values, and
full weight for the values between them.
m = trimmean(x,percent,flag,dim) takes the trimmed mean along
dimension dim of x.
Remarks The trimmed mean is a robust estimate of the location of a sample.
If there are outliers in the data, the trimmed mean is a more
representative estimate of the center of the body of the data than the
mean. However, if the data is all from the same probability distribution,
18-1422
trimmean
then the trimmed mean is less efficient than the sample mean as an
estimator of the location of the data.
Examples Example 1
This example shows a Monte Carlo simulation of the efficiency of the
10% trimmed mean relative to the sample mean for normal data.
x = normrnd(0,1,100,100);
m = mean(x);
trim = trimmean(x,10);
sm = std(m);
strim = std(trim);
efficiency = (sm/strim).^2
efficiency =
0.9702
Example 2
Generate random data from the t distribution, which tends to have
outliers:
reset(RandStream.getDefaultStream)
x = trnd(1,40,1);
probplot(x)
18-1423
trimmean
Though the distribution is symmetric around zero, there are several
outliers which will affect the mean. The trimmed mean is much closer
to zero, which is much more representative of the data:
mean(x)
ans =
2.7991
trimmean(x,25)
ans =
18-1424
trimmean
0.8797
See Also mean, median, geomean, harmmean
18-1425
trnd
Purpose Student’s t random numbers
Syntax R = trnd(V)
R = trnd(v,m)
R = trnd(V,m,n)
Description R = trnd(V) generates random numbers from Student’s t distribution
with V degrees of freedom. V can be a vector, a matrix, or a
multidimensional array. The size of R is the size of V.
R = trnd(v,m) generates random numbers from Student’s t
distribution with v degrees of freedom, where v is a row vector. If v is a
1-by-2 vector, R is a matrix with v(1) rows and v(2) columns. If v is
1-by-n, R is an n-dimensional array.
R = trnd(V,m,n) generates random numbers from Student’s t
distribution with V degrees of freedom, where scalars m and n are the
row and column dimensions of R.
Examples noisy = trnd(ones(1,6))
noisy =
19.7250 0.3488 0.2843 0.4034 0.4816 -2.4190
numbers = trnd(1:6,[1 6])
numbers =
-1.9500 -0.9611 -0.9038 0.0754 0.9820 1.0115
numbers = trnd(3,2,6)
numbers =
-0.3177 -0.0812 -0.6627 0.1905 -1.5585 -0.0433
0.2536 0.5502 0.8646 0.8060 -0.5216 0.0891
See Also random, tpdf, tcdf, tinv, tstat
“Student’s t Distribution” on page B-95
18-1426
tstat
Purpose Student’s t mean and variance
Syntax [M,V] = tstat(NU)
Description [M,V] = tstat(NU) returns the mean of and variance for Student’s t
distribution using the degrees of freedom in NU. M and V are the same
size as NU.
The mean of the Student’s t distribution with parameter ν is zero for
values of ν greater than 1. If ν is one, the mean does not exist. The
variance for values of ν greater than 2 is ν/(ν-2).
Examples Find the mean of and variance for 1 to 30 degrees of freedom.
[m,v] = tstat(reshape(1:30,6,5))
m =
NaN 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
v =
NaN 1.4000 1.1818 1.1176 1.0870
NaN 1.3333 1.1667 1.1111 1.0833
3.0000 1.2857 1.1538 1.1053 1.0800
2.0000 1.2500 1.1429 1.1000 1.0769
1.6667 1.2222 1.1333 1.0952 1.0741
1.5000 1.2000 1.1250 1.0909 1.0714
Note that the variance does not exist for one and two degrees of freedom.
See Also tpdf, tcdf, tinv, trnd
18-1427
ttest
Purpose One-sample and paired-sample t-test
Syntax h = ttest(x)
h = ttest(x,m)
h = ttest(x,y)
h = ttest(...,alpha)
h = ttest(...,alpha,tail)
h = ttest(...,alpha,tail,dim)
[h,p] = ttest(...)
[h,p,ci] = ttest(...)
[h,p,ci,stats] = ttest(...)
Description h = ttest(x) performs a t-test of the null hypothesis that data in the
vector x are a random sample from a normal distribution with mean 0
and unknown variance, against the alternative that the mean is not 0.
The result of the test is returned in h. h = 1 indicates a rejection of the
null hypothesis at the 5% significance level. h = 0 indicates a failure to
reject the null hypothesis at the 5% significance level.
x can also be a matrix or an N-dimensional array. For matrices, ttest
performs separate t-tests along each column of x and returns a vector
of results. For N-dimensional arrays, ttest works along the first
non-singleton dimension of x.
The test treats NaN values as missing data, and ignores them.
h = ttest(x,m) performs a t-test of the null hypothesis that data in
the vector x are a random sample from a normal distribution with mean
m and unknown variance, against the alternative that the mean is not m.
h = ttest(x,y) performs a paired t-test of the null hypothesis
that data in the difference x-y are a random sample from a normal
distribution with mean 0 and unknown variance, against the alternative
that the mean is not 0. x and y must be vectors of the same length,
or arrays of the same size.
h = ttest(...,alpha) performs the test at the (100*alpha)%
significance level. The default, when unspecified, is alpha = 0.05.
18-1428
ttest
h = ttest(...,alpha,tail) performs the test against the alternative
specified by the string tail. There are three options for tail:
• 'both' — Mean is not 0 (or m) (two-tailed test). This is the default,
when tail is unspecified.
• 'right' — Mean is greater than 0 (or m) (right-tail test)
• 'left' — Mean is less than 0 (or m) (left-tail test)
tail must be a single string, even when x is a matrix or an
N-dimensional array.
h = ttest(...,alpha,tail,dim) works along dimension dim of x, or
of x-y for a paired test. Use [] to pass in default values for m, alpha,
or tail.
[h,p] = ttest(...) returns the p value of the test. The p value is the
probability, under the null hypothesis, of observing a value as extreme
or more extreme of the test statistic
x−
t=
s/ n
where x is the sample mean, μ = 0 (or m) is the hypothesized population
mean, s is the sample standard deviation, and n is the sample size.
Under the null hypothesis, the test statistic will have Student’s t
distribution with n – 1 degrees of freedom.
[h,p,ci] = ttest(...) returns a 100*(1 – alpha)% confidence
interval on the population mean, or on the difference of population
means for a paired test.
[h,p,ci,stats] = ttest(...) returns the structure stats with the
following fields:
• tstat — Value of the test statistic
• df — Degrees of freedom of the test
• sd — Sample standard deviation
18-1429
ttest
Examples Simulate a random sample of size 100 from a normal distribution with
mean 0.1:
x = normrnd(0.1,1,1,100);
Test the null hypothesis that the sample comes from a normal
distribution with mean 0:
[h,p,ci] = ttest(x,0)
h =
0
p =
0.8323
ci =
-0.1650 0.2045
The test fails to reject the null hypothesis at the default α = 0.05
significance level. Under the null hypothesis, the probability of
observing a value as extreme or more extreme of the test statistic, as
indicated by the p value, is much greater than α. The 95% confidence
interval on the mean contains 0.
Simulate a larger random sample of size 1000 from the same
distribution:
y = normrnd(0.1,1,1,1000);
Test again if the sample comes from a normal distribution with mean 0:
[h,p,ci] = ttest(y,0)
h =
1
p =
0.0160
ci =
0.0142 0.1379
18-1430
ttest
This time the test rejects the null hypothesis at the default α = 0.05
significance level. The p value has fallen below α = 0.05 and the 95%
confidence interval on the mean does not contain 0.
Because the p value of the sample y is greater than 0.01, the test will
fail to reject the null hypothesis when the significance level is lowered
to α = 0.01:
[h,p,ci] = ttest(y,0,0.01)
h =
0
p =
0.0160
ci =
-0.0053 0.1574
Notice that at the lowered significance level the 99% confidence interval
on the mean widens to contain 0.
This example will produce slightly different results each time it is run,
because of the random sampling.
See Also ttest2, ztest
18-1431
ttest2
Purpose Two-sample t-test
Syntax h = ttest2(x,y)
h = ttest2(x,y,alpha)
h = ttest2(x,y,alpha,tail)
h = ttest2(x,y,alpha,tail,vartype)
h = ttest2(x,y,alpha,tail,vartype,dim)
[h,p] = ttest2(...)
[h,p,ci] = ttest2(...)
[h,p,ci,stats] = ttest2(...)
Description h = ttest2(x,y) performs a t-test of the null hypothesis that data
in the vectors x and y are independent random samples from normal
distributions with equal means and equal but unknown variances,
against the alternative that the means are not equal. The result of the
test is returned in h. h = 1 indicates a rejection of the null hypothesis
at the 5% significance level. h = 0 indicates a failure to reject the null
hypothesis at the 5% significance level. x and y need not be vectors
of the same length.
x and y can also be matrices or N-dimensional arrays. Matrices x and y
must have the same number of columns, in which case ttest2 performs
separate t-tests along each column and returns a vector of results.
N-dimensional arrays x and y must have the same size along all but
the first non-singleton dimension, in which case ttest2 works along
the first non-singleton dimension.
The test treats NaN values as missing data, and ignores them.
h = ttest2(x,y,alpha) performs the test at the (100*alpha)%
significance level. The default, when unspecified, is alpha = 0.05.
h = ttest2(x,y,alpha,tail) performs the test against the alternative
specified by the string tail. There are three options for tail:
• 'both' — Means are not equal (two-tailed test). This is the default,
when tail is unspecified.
• 'right' — Mean of x is greater than mean of y (right-tail test)
18-1432
ttest2
• 'left' — Mean of x is less than mean of y (left-tail test)
tail must be a single string, even when x is a matrix or an
N-dimensional array.
h = ttest2(x,y,alpha,tail,vartype) performs the test under the
assumption of equal or unequal population variances, as specified by
the string vartype. There are two options for vartype:
• 'equal' — Assumes equal variances. This is the default, when
vartype is unspecified.
• 'unequal' — Does not assume equal variances. This is the
Behrens-Fisher problem.
vartype must be a single string, even when x is a matrix or an
N-dimensional array.
If vartype is 'equal', the test computes a pooled sample standard
deviation using
(n − 1) sx2 + (m − 1) s2y
s=
n+m−2
where sx and sy are the sample standard deviations of x and y,
respectively, and n and m are the sample sizes of x and y, respectively.
h = ttest2(x,y,alpha,tail,vartype,dim) works along dimension
dim of x and y. Use [] to pass in default values for alpha, tail, or
vartype.
[h,p] = ttest2(...) returns the p value of the test. The p value
is the probability, under the null hypothesis, of observing a value as
extreme or more extreme of the test statistic
18-1433
ttest2
x−y
t=
2
sx2 s y
+
n m
where x and y are the sample means, sx and sy are the sample standard
deviations (replaced by the pooled standard deviation s in the default
case where vartype is 'equal'), and n and m are the sample sizes.
In the default case where vartype is 'equal', the test statistic, under
the null hypothesis, has Student’s t distribution with n + m – 2 degrees
of freedom.
In the case where vartype is 'unequal', the test statistic, under the
null hypothesis, has an approximate Student’s t distribution with a
number of degrees of freedom given by Satterthwaite’s approximation.
[h,p,ci] = ttest2(...) returns a 100*(1 – alpha)% confidence
interval on the difference of population means.
[h,p,ci,stats] = ttest2(...) returns structure stats with the
following fields:
• tstat — Value of the test statistic
• df — Degrees of freedom of the test
• sd — Pooled sample standard deviation (in the default case where
vartype is 'equal') or a vector with the sample standard deviations
(in the case where vartype is 'unequal').
Examples Simulate random samples of size 1000 from normal distributions
with means 0 and 0.1, respectively, and standard deviations 1 and 2,
respectively:
x = normrnd(0,1,1,1000);
y = normrnd(0.1,2,1,1000);
18-1434
ttest2
Test the null hypothesis that the samples come from populations with
equal means, against the alternative that the means are unequal.
Perform the test assuming unequal variances:
[h,p,ci] = ttest2(x,y,[],[],'unequal')
h =
1
p =
0.0102
ci =
-0.3227 -0.0435
The test rejects the null hypothesis at the default α = 0.05 significance
level. Under the null hypothesis, the probability of observing a value
as extreme or more extreme of the test statistic, as indicated by the p
value, is less than α. The 95% confidence interval on the mean of the
difference does not contain 0.
This example will produce slightly different results each time it is run,
because of the random sampling.
See Also ttest, ztest
18-1435
classregtree.type
Purpose Tree type
Syntax ttype = type(t)
Description ttype = type(t) returns the type of the tree t. ttype is 'regression'
for regression trees and 'classification' for classification trees.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
18-1436
classregtree.type
ttype = type(t)
ttype =
classification
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
See Also classregtree
18-1437
cvpartition.Type property
Purpose Type of partition
Description The type of validation partition. It is 'kfold', 'holdout', 'leaveout',
or 'resubstitution'.
See Also trainsize
18-1438
qrandset.Type property
Purpose Name of sequence on which point set P is based
Description P.Type returns a string that contains the name of the sequence on
which the point set P is based, for example 'Sobol'. You cannot change
the Type property for a point set.
18-1439
categorical.uint8
Purpose Convert categorical array to unsigned 8-bit integers
Syntax B = uint8(A)
Description B = uint8(A) converts the categorical array A to unsigned 8-bit
integers. Each element of B contains the internal categorical level code
for the corresponding element of A. Undefined elements of A are assigned
the value 0 in B. If A contains more than intmax('uint8') levels, the
internal codes will saturate to intmax('uint8')when cast to int8.
See Also double, int8
18-1440
categorical.uint16
Purpose Convert categorical array to unsigned 16-bit integers
Syntax B = uint16(A)
Description B = uint16(A) converts the categorical array A to unsigned 16-bit
integers. Each element of B contains the internal categorical level code
for the corresponding element of A.
Undefined elements of A are assigned the value 0 in B.
See Also double, int16
18-1441
categorical.uint32
Purpose Convert categorical array to unsigned 32-bit integers
Syntax B = uint32(A)
Description B = uint32(A) converts the categorical array A to unsigned 32-bit
integers. Each element of B contains the internal categorical level code
for the corresponding element of A.
Undefined elements of A are assigned the value 0 in B.
See Also double, int32
18-1442
categorical.uint64
Purpose Convert categorical array to unsigned 64-bit integers
Syntax B = uint64(A)
Description B = uint64(A) converts the categorical array A to unsigned 64-bit
integers. Each element of B contains the internal categorical level code
for the corresponding element of A.
Undefined elements of A are assigned the value 0 in B.
See Also double, int64
18-1443
categorical.undeflabel property
Purpose Text label for undefined levels
Description Text label for undefined levels. Constant property with value
'<undefined>'.
18-1444
dataset.unique
Purpose Unique observations in dataset array
Syntax B = unique(A)
B = unique(A,vars)
[B,i,j] = unique(A)
[...] = unique(A,vars,'first')
Description B = unique(A) returns a copy of the dataset A that contains only the
sorted unique observations. A must contain only variables whose class
has a unique method, including:
• numeric
• character
• logical
• categorical
• cell arrays of strings
For a variable with multiple columns, its class’s unique method must
support the 'rows' flag.
B = unique(A,vars) returns a dataset that contains only one
observation for each unique combination of values for the variables in A
specified in vars. vars is a positive integer, a vector of positive integers,
a variable name, a cell array containing one or more variable names,
or a logical vector. B includes all variables from A. The values in B for
the variables not specified in vars are taken from the last occurrence
among observations in A with each unique combination of values for the
variables specified in vars.
[B,i,j] = unique(A) also returns index vectors i and j such that B =
A(i,:) and A = B(j,:).
[...] = unique(A,vars,'first') returns the vector i to
index the first occurrence of each unique observation in A.
unique(A,vars,'last'), the default, returns the vector i to index
18-1445
dataset.unique
the last occurrence. Specify vars as [] to use the default value of all
variables.
See Also dataset, set, subsasgn
18-1446
dataset.Units property
Purpose Units of variables in data set
Description A cell array of strings giving the units of the variables in the data set.
This property may be empty, but if not empty, the number of strings
must equal the number of variables. Any individual string may be
empty for a variable that does not have units defined. The default is
an empty cell array.
18-1447
unidcdf
Purpose Discrete uniform cumulative distribution function
Syntax P = unidcdf(X,N)
Description P = unidcdf(X,N) computes the discrete uniform cdf at each of the
values in X using the corresponding maximum observable value in N. X
and N can be vectors, matrices, or multidimensional arrays that have
the same size. A scalar input is expanded to a constant array with the
same dimensions as the other inputs. The maximum observable values
in N must be positive integers.
The discrete uniform cdf is
floor ( x )
p = F( x| N ) = I(1,..., N ) ( x)
N
The result, p, is the probability that a single observation from the
discrete uniform distribution with maximum N will be a positive integer
less than or equal to x. The values x do not need to be integers.
Examples What is the probability of drawing a number 20 or less from a hat with
the numbers from 1 to 50 inside?
probability = unidcdf(20,50)
probability =
0.4000
See Also cdf, unidpdf, unidinv, unidstat, unidrnd, mle
“Uniform Distribution (Discrete)” on page B-101
18-1448
unidinv
Purpose Discrete uniform inverse cumulative distribution function
Syntax X = unidinv(P,N)
Description X = unidinv(P,N) returns the smallest positive integer X such that the
discrete uniform cdf evaluated at X is equal to or exceeds P. You can
think of P as the probability of drawing a number as large as X out of a
hat with the numbers 1 through N inside.
P and N can be vectors, matrices, or multidimensional arrays that have
the same size, which is also the size of X. A scalar input for N or P is
expanded to a constant array with the same dimensions as the other
input. The values in P must lie on the interval [0 1] and the values in N
must be positive integers.
Examples x = unidinv(0.7,20)
x =
14
y = unidinv(0.7 + eps,20)
y =
15
A small change in the first parameter produces a large jump in output.
The cdf and its inverse are both step functions. The example shows
what happens at a step.
See Also icdf, unidcdf, unidpdf, unidstat, unidrnd
“Uniform Distribution (Discrete)” on page B-101
18-1449
unidpdf
Purpose Discrete uniform probability density function
Syntax Y = unidpdf(X,N)
Description Y = unidpdf(X,N) computes the discrete uniform pdf at each of the
values in X using the corresponding maximum observable value in N. X
and N can be vectors, matrices, or multidimensional arrays that have
the same size. A scalar input is expanded to a constant array with the
same dimensions as the other inputs. The parameters in N must be
positive integers.
The discrete uniform pdf is
1
y = f ( x| N ) = I ( x)
N (1,..., N )
You can think of y as the probability of observing any one number
between 1 and n.
Examples For fixed n, the uniform discrete pdf is a constant.
y = unidpdf(1:6,10)
y =
0.1000 0.1000 0.1000 0.1000 0.1000 0.1000
Now fix x, and vary n.
likelihood = unidpdf(5,4:9)
likelihood =
0 0.2000 0.1667 0.1429 0.1250 0.1111
See Also pdf, unidcdf, unidinv, unidstat, unidrnd
“Uniform Distribution (Discrete)” on page B-101
18-1450
unidrnd
Purpose Discrete uniform random numbers
Syntax R = unidrnd(N)
R = unidrnd(N,v)
R = unidrnd(N,m,n)
Description R = unidrnd(N) generates random numbers for the discrete uniform
distribution with maximum N. The parameters in N must be positive
integers. N can be a vector, a matrix, or a multidimensional array. The
size of R is the size of N. The discrete uniform distribution arises from
experiments equivalent to drawing a number from one to N out of a hat.
R = unidrnd(N,v) generates random numbers for the discrete uniform
distribution with maximum N, where v is a row vector. If v is a 1-by-2
vector, R is a matrix with v(1) rows and v(2) columns. If v is 1-by-n, R
is an n-dimensional array.
R = unidrnd(N,m,n) generates random numbers for the discrete
uniform distribution with maximum N, where scalars m and n are the
row and column dimensions of R.
Examples In the Massachusetts lottery, a player chooses a four-digit number.
Generate random numbers for Monday through Saturday.
numbers = unidrnd(10000,1,6)-1
numbers =
4564 185 8214 4447 6154 7919
See Also random, unidpdf, unidcdf, unidinv, unidstat
“Uniform Distribution (Discrete)” on page B-101
18-1451
unidstat
Purpose Discrete uniform mean of and variance
Syntax [M,V] = unidstat(N)
Description [M,V] = unidstat(N) returns the mean of and variance for the
discrete uniform distribution with corresponding maximum observable
value in N.
The mean of the discrete uniform distribution with parameter N is
(N+1)/2. The variance is (N2+1)/12.
Examples [m,v] = unidstat(1:6)
m =
1.0000 1.5000 2.0000 2.5000 3.0000 3.5000
v =
0 0.2500 0.6667 1.2500 2.0000 2.9167
See Also unidpdf, unidcdf, unidinv, unidrnd
“Uniform Distribution (Discrete)” on page B-101
18-1452
unifcdf
Purpose Continuous uniform cumulative distribution function
Syntax P = unifcdf(X,A,B)
Description P = unifcdf(X,A,B) computes the uniform cdf at each of the values
in X using the corresponding lower endpoint (minimum), A and upper
endpoint (maximum), B. X, A, and B can be vectors, matrices, or
multidimensional arrays that all have the same size. A scalar input is
expanded to a constant matrix with the same dimensions as the other
inputs.
The uniform cdf is
x−a
p = F ( x | a, b) = I ( x)
b − a [ a,b]
The standard uniform distribution has A = 0 and B = 1.
Examples What is the probability that an observation from a standard uniform
distribution will be less than 0.75?
probability = unifcdf(0.75)
probability =
0.7500
What is the probability that an observation from a uniform distribution
with a = -1 and b = 1 will be less than 0.75?
probability = unifcdf(0.75,-1,1)
probability =
0.8750
See Also cdf, unifpdf, unifinv, unifstat, unifit, unifrnd
“Uniform Distribution (Continuous)” on page B-99
18-1453
unifinv
Purpose Continuous uniform inverse cumulative distribution function
Syntax X = unifinv(P,A,B)
Description X = unifinv(P,A,B) computes the inverse of the uniform cdf with
parameters A and B (the minimum and maximum values, respectively)
at the corresponding probabilities in P. P, A, and B can be vectors,
matrices, or multidimensional arrays that all have the same size. A
scalar input is expanded to a constant array with the same dimensions
as the other inputs.
The inverse of the uniform cdf is
x = F −1 ( p| a, b) = a + p ( a − b) I[0,1] ( p)
The standard uniform distribution has A = 0 and B = 1.
Examples What is the median of the standard uniform distribution?
median_value = unifinv(0.5)
median_value =
0.5000
What is the 99th percentile of the uniform distribution between -1
and 1?
percentile = unifinv(0.99,-1,1)
percentile =
0.9800
See Also icdf, unifcdf, unifpdf, unifstat, unifit, unifrnd
“Uniform Distribution (Continuous)” on page B-99
18-1454
unifit
Purpose Continuous uniform parameter estimates
Syntax [ahat,bhat] = unifit(data)
[ahat,bhat,ACI,BCI] = unifit(data)
[ahat,bhat,ACI,BCI] = unifit(data,alpha)
Description [ahat,bhat] = unifit(data) returns the maximum likelihood
estimates (MLEs) of the parameters of the uniform distribution given
the data in data.
[ahat,bhat,ACI,BCI] = unifit(data) also returns 95% confidence
intervals, ACI and BCI, which are matrices with two rows. The first
row contains the lower bound of the interval for each column of the
matrix data. The second row contains the upper bound of the interval.
[ahat,bhat,ACI,BCI] = unifit(data,alpha) enables you to control
of the confidence level alpha. For example, if alpha = 0.01 then ACI
and BCI are 99% confidence intervals.
Examples r = unifrnd(10,12,100,2);
[ahat,bhat,aci,bci] = unifit(r)
ahat =
10.0154 10.0060
bhat =
11.9989 11.9743
aci =
9.9551 9.9461
10.0154 10.0060
bci =
11.9989 11.9743
12.0592 12.0341
See Also mle, unifpdf, unifcdf, unifinv, unifstat, unifrnd
“Uniform Distribution (Continuous)” on page B-99
18-1455
unifpdf
Purpose Continuous uniform probability density function
Syntax Y = unifpdf(X,A,B)
Description Y = unifpdf(X,A,B) computes the continuous uniform pdf at each of
the values in X using the corresponding lower endpoint (minimum), A
and upper endpoint (maximum), B. X, A, and B can be vectors, matrices,
or multidimensional arrays that all have the same size. A scalar input
is expanded to a constant array with the same dimensions as the other
inputs. The parameters in B must be greater than those in A.
The continuous uniform distribution pdf is
1
y = f ( x | a, b) = I ( x)
b − a [ a,b]
The standard uniform distribution has A = 0 and B = 1.
Examples For fixed a and b, the uniform pdf is constant.
x = 0.1:0.1:0.6;
y = unifpdf(x)
y =
1 1 1 1 1 1
What if x is not between a and b?
y = unifpdf(-1,0,1)
y =
0
See Also pdf, unifcdf, unifinv, unifstat, unifit, unifrnd
“Uniform Distribution (Continuous)” on page B-99
18-1456
unifrnd
Purpose Continuous uniform random numbers
Syntax R = unifrnd(A,B)
R = unifrnd(A,B,m,n,...)
R = unifrnd(A,B,[m,n,...])
Description R = unifrnd(A,B) returns an array R of random numbers generated
from the continuous uniform distributions with lower and upper
endpoints specified by A and B, respectively. If A and B are arrays,
R(i,j) is generated from the distribution specified by the corresponding
elements of A and B. If either A or B is a scalar, it is expanded to the
size of the other input.
R = unifrnd(A,B,m,n,...) or R = unifrnd(A,B,[m,n,...]) returns
an m-by-n-by-... array. If A and B are scalars, all elements of R are
generated from the same distribution. If either A or B is an array, they
must be m-by-n-by-... .
Examples Generate one random number each from the continuous uniform
distributions on the intervals (0,1), (0,2), ..., (0,5):
a = 0; b = 1:5;
r1 = unifrnd(a,b)
r1 =
0.8147 1.8116 0.3810 3.6535 3.1618
Generate five random numbers each from the same distributions:
B = repmat(b,5,1);
R = unifrnd(a,B)
R =
0.0975 0.3152 0.4257 2.6230 3.7887
0.2785 1.9412 1.2653 0.1428 3.7157
0.5469 1.9143 2.7472 3.3965 1.9611
0.9575 0.9708 2.3766 3.7360 3.2774
0.9649 1.6006 2.8785 2.7149 0.8559
18-1457
unifrnd
Generate five random numbers from the continuous uniform
distribution on (0,2):
r2 = unifrnd(a,b(2),1,5)
r2 =
1.4121 0.0637 0.5538 0.0923 0.1943
See Also rand, random, unifpdf, unifcdf, unifinv, unifstat, unifit
“Uniform Distribution (Continuous)” on page B-99
18-1458
unifstat
Purpose Continuous uniform mean and variance
Syntax [M,V] = unifstat(A,B)
Description [M,V] = unifstat(A,B) returns the mean of and variance for the
continuous uniform distribution using the corresponding lower endpoint
(minimum), A and upper endpoint (maximum), B. Vector or matrix
inputs for A and B must have the same size, which is also the size of M
and V. A scalar input for A or B is expanded to a constant matrix with
the same dimensions as the other input.
The mean of the continuous uniform distribution with parameters a and
b is (a + b)/2, and the variance is (a – b)2/12.
Examples a = 1:6;
b = 2.*a;
[m,v] = unifstat(a,b)
m =
1.5000 3.0000 4.5000 6.0000 7.5000 9.0000
v =
0.0833 0.3333 0.7500 1.3333 2.0833 3.0000
See Also unifpdf, unifcdf, unifinv, unifit, unifrnd
“Uniform Distribution (Continuous)” on page B-99
18-1459
categorical.union
Purpose Set union for categorical arrays
Syntax C = union(A,B)
[C,IA,IB] = union(A,B)
Description C = union(A,B) when A and B are categorical arrays returns a
categorical vector C containing the combined values from A and B but
with no repetitions. The result C is sorted. The set of categorical levels
for C is the sorted union of the sets of levels of the inputs, as determined
by their labels.
[C,IA,IB] = union(A,B) also returns index vectors IA and IB such
that C is a sorted combination of the elements A(IA) and B(IB).
See Also intersect, ismember, setdiff, setxor, unique
18-1460
categorical.unique
Purpose Unique values in categorical array
Syntax B = unique(A)
[B,I,J] = unique(A)
[B,I,J] = unique(A,'first')
Description B = unique(A) returns a categorical array containing the unique
elements of A, sorted by the order of A’s levels.
[B,I,J] = unique(A) also returns index vectors I and J such that
B = A(I) and A = B(J).
[B,I,J] = unique(A,'first') returns the vector I to index the first
occurrence of each unique value in A. unique(A,'last'), the default,
returns the vector I to index the last occurrence.
See Also intersect, ismember, setdiff, setxor, union
18-1461
dataset.unstack
Purpose Unstack data from single variable into multiple variables
Syntax wide = unstack(tall,datavar,indvar)
[wide,itall] = unstack(tall,datavar,indvar)
wide = unstack(tall,datavar,indvar,’Parameter’,value)
Description wide = unstack(tall,datavar,indvar) converts a dataset array
tall to an equivalent dataset array wide that is in wide format, by
unstacking a single variable in tall into multiple variables in wide. In
general wide contains more variables, but fewer observations, than
tall.
datavar specifies the data variable in tall to unstack. indvar specifies
an indicator variable in tall that determines which variable in wide
each value in datavar is unstacked into. unstack treats the remaining
variables in tall as grouping variables. Each unique combination
of their values defines a group of observations in tall that will be
unstacked into a single observation in wide.
unstack creates m data variables in wide, where m is the number of
group levels in indvar. The values in indvar indicate which of those m
variables receive which values from datavar. The j-th data variable in
wide contains the values from datavar that correspond to observations
whose indvar value was the j-th of the m possible levels. Elements of
those m variables for which no corresponding data value in tall exists
contain a default value.
datavar is a positive integer, a variable name, or a logical vector
containing a single true value. indvar is a positive integer, a variable
name, or a logical vector containing a single true value.
[wide,itall] = unstack(tall,datavar,indvar) returns an index
vector itall indicating the correspondence between observations in
wide and those in tall. For each observation in wide, itall contains
the index of the first in the corresponding group of observations in tall.
For more information on grouping variables, see “Grouping Variables”
on page 2-34.
18-1462
dataset.unstack
Input wide = unstack(tall,datavar,indvar,’Parameter’,value)uses the
Arguments following parameter name/value pairs to control how unstack converts
variables in tall to variables in wide:
'GroupVars' Grouping variables in tall that define groups of
observations. groupvars is a positive integer, a
vector of positive integers, a variable name, a cell
array containing one or more variable names, or a
logical vector. The default is all variables in tall
not listed in datavar or indvar.
A cell array of strings containing names for the
'NewDataVarNames'
data variables unstack should create in wide.
Default is the group names of the grouping variable
specified in indvar.
'AggregationFun'A function handle that accepts a subset of values
from datavar and returns a single value. stack
applies this function to observations from the same
group that have the same value of indvar. The
function must aggregate the data values into a
single value, and in such cases it is not possible to
recover tall from wide using stack. The default is
@sum for numeric data variables. For non-numeric
variables, there is no default, and you must specify
'AggregationFun' if multiple observations in the
same group have the same values of indvar.
'ConstVars' Variables in tall to copy to wide without
unstacking. The values for these variables in
wide are taken from the first observation in each
group in tall, so these variables should typically
be constant within each group. ConstVars is a
positive integer, a vector of positive integers, a
variable name, a cell array containing one or more
variable names, or a logical vector. The default is
no variables.
18-1463
dataset.unstack
You can also specify more than one data variable in tall, each of
which becomes a set of m variables in wide. In this case, specify
datavar as a vector of positive integers, a cell array containing variable
names, or a logical vector. You may specify only one variable with
indvar. The names of each set of data variables in wide are the
name of the corresponding data variable in tall concatenated with
the names specified in 'NewDataVarNames'. The function specified in
'AggregationFun' must return a value with a single row.
Examples Convert a "wide format" data set to "tall format", and then back to
a different "wide format":
load flu
% FLU has a 'Date' variable, and 10 variables for estimated
% influenza rates (in 9 different regions, estimated from
% Google searches, plus a nationwide extimate from the
% CDC). Combine those 10 variables into a "tall" array
% that has a single data variable, 'FluRate', and an
% indicator variable, 'Region', that says which region
% each estimate is from.
flu2 = stack(flu, 2:11, 'NewDataVarName','FluRate',...
'IndVarName','Region')
dateNames = cellstr(datestr(flu.Date,'mmm_DD_YYYY'));
% Use the 'Date' variable from that tall array to split
% 'FluRate' into 52 separate variables, each containing
% the estimated influenza rates for each unique date.
% The new "wide" array has one observation for each region.
% In effect, this is the original array FLU "on its side".
flu3 = unstack(flu2, 'FluRate', 'Date',...
'NewDataVarNames',dateNames)
See Also dataset.stack | dataset.join | dataset.grpstats
How To • “Grouping Variables” on page 2-34
18-1464
paretotails.upperparams
Purpose Upper Pareto tails parameters
Syntax params = upperparams(obj)
Description params = upperparams(obj) returns the 2-element vector params of
shape and scale parameters, respectively, of the upper tail of the Pareto
tails object obj. upperparams does not return a location parameter.
Examples Fit Pareto tails to a t distribution at cumulative probabilities 0.1 and
0.9:
t = trnd(3,100,1);
obj = paretotails(t,0.1,0.9);
lowerparams(obj)
ans =
-0.1901 1.1898
upperparams(obj)
ans =
0.3646 0.5103
See Also paretotails, lowerparams
18-1465
dataset.UserData property
Purpose Variable containing additional information associated with data set
Description Any variable containing additional information to be associated with
the data set. The default is an empty array.
18-1466
ProbDistUnivParam.var
Purpose Return variance of ProbDistUnivParam object
Syntax V = var(PD)
Description V = var(PD) returns V, the variance of the ProbDistUnivParam object
PD.
Input PD An object of the class ProbDistUnivParam.
Arguments
Output V The variance of the ProbDistUnivParam
Arguments object PD.
See Also var
18-1467
dataset.VarDescription property
Purpose Cell array of strings giving descriptions of variables in data set
Description A cell array of strings giving the descriptions of the variables in the
data set. This property may be empty, but if not empty, the number
of strings must equal the number of variables. Any individual string
may be empty for a variable that does not have a description defined.
The default is an empty cell array.
18-1468
classregtree.varimportance
Purpose Compute embedded estimates of input feature importance
Syntax imp = varimportance(t)
Description imp = varimportance(t) computes estimates of input feature
importance for tree t by summing changes in the risk due to splits on
every feature. The returned vector imp has one element for each input
variable in the data used to train this tree. At each node, the risk is
estimated as node impurity if impurity was used to split nodes and
node error otherwise. This risk is weighted by the node probability.
Variable importance associated with this split is computed as the
difference between the risk for the parent node and the total risk for
the two children.
See Also risk
18-1469
CompactTreeBagger.VarNames property
Purpose Variable names
Description The VarNames property is a cell array containing the names of the
predictor variables (features). These names are taken from the optional
'names' parameter that supplied to TreeBagger. The default names
are 'x1', 'x2', etc.
18-1470
dataset.VarNames property
Purpose Cell array giving names of variables in data set
Description A cell array of nonempty, distinct strings giving the names of the
variables in the data set. The number of strings must equal the number
of variables. The default is the cell array of string names for the
variables used to create the data set.
18-1471
TreeBagger.VarNames property
Purpose Variable names
Description The VarNames property is a cell array containing the names of the
predictor variables (features). TreeBagger takes these names from the
optional 'names' parameter. The default names are 'x1', 'x2', etc.
18-1472
vartest
Purpose Chi-square variance test
Syntax H = vartest(X,V)
H = vartest(X,V,alpha)
H = vartest(X,V,alpha,tail)
[H,P] = vartest(...)
[H,P,CI] = vartest(...)
[H,P,CI,STATS] = vartest(...)
[...] = vartest(X,V,alpha,tail,dim)
Description H = vartest(X,V) performs a chi-square test of the hypothesis that
the data in the vector X comes from a normal distribution with variance
V, against the alternative that X comes from a normal distribution with
a different variance. The result is H = 0 if the null hypothesis (variance
is V) cannot be rejected at the 5% significance level, or H = 1 if the null
hypothesis can be rejected at the 5% level.
X may also be a matrix or an n-dimensional array. For matrices,
vartest performs separate tests along each column of X, and returns a
row vector of results. For n-dimensional arrays, vartest works along
the first nonsingleton dimension of X. V must be a scalar.
H = vartest(X,V,alpha) performs the test at the significance level
(100*alpha)%. alpha has a default value of 0.05 and must be a scalar.
H = vartest(X,V,alpha,tail) performs the test against the
alternative hypothesis specified by tail, where tail is a single string
from the following choices:
• 'both' — Variance is not V (two-tailed test). This is the default.
• 'right' — Variance is greater than V (right-tailed test).
• 'left' — Variance is less than V (left-tailed test).
[H,P] = vartest(...) returns the p value, i.e., the probability of
observing the given result, or one more extreme, by chance if the null
hypothesis is true. Small values of P cast doubt on the validity of the
null hypothesis.
18-1473
vartest
[H,P,CI] = vartest(...) returns a 100*(1-alpha)% confidence
interval for the true variance.
[H,P,CI,STATS] = vartest(...) returns the structure STATS with
the following fields:
• 'chisqstat' — Value of the test statistic
• 'df' — Degrees of freedom of the test
[...] = vartest(X,V,alpha,tail,dim) works along dimension dim
of X. Pass in [] for alpha or tail to use their default values.
Examples Determine whether the standard deviation is significantly different
from 7?
load carsmall
[h,p,ci] = vartest(MPG,7^2)
See Also ttest, ztest,vartest2
18-1474
vartest2
Purpose Two-sample F-test for equal variances
Syntax H = vartest2(X,Y)
H = vartest2(X,Y,alpha)
H = vartest2(X,Y,alpha,tail)
[H,P] = vartest2(...)
[H,P,CI] = vartest2(...)
[H,P,CI,STATS] = vartest2(...)
[...] = vartest2(X,Y,alpha,tail,dim)
Description H = vartest2(X,Y) performs an F test of the hypothesis that two
independent samples, in the vectors X and Y, come from normal
distributions with the same variance, against the alternative that they
come from normal distributions with different variances. The result is H
= 0 if the null hypothesis (variances are equal) cannot be rejected at the
5% significance level, or H = 1 if the null hypothesis can be rejected at
the 5% level. X and Y can have different lengths. X and Y can also be
matrices or n-dimensional arrays.
For matrices, vartest2 performs separate tests along each column,
and returns a vector of results. X and Y must have the same number
of columns. For n-dimensional arrays, vartest2 works along the first
nonsingleton dimension. X and Y must have the same size along all
the remaining dimensions.
H = vartest2(X,Y,alpha) performs the test at the significance level
(100*alpha)%. alpha must be a scalar.
H = vartest2(X,Y,alpha,tail) performs the test against the
alternative hypothesis specified by tail, where tail is one of the
following single strings:
• 'both' — Variance is not Y (two-tailed test). This is the default.
• 'right' — Variance is greater than Y (right-tailed test).
• 'left' — Variance is less than Y (left-tailed test).
18-1475
vartest2
[H,P] = vartest2(...) returns the p value, i.e., the probability of
observing the given result, or one more extreme, by chance if the null
hypothesis is true. Small values of P cast doubt on the validity of the
null hypothesis.
[H,P,CI] = vartest2(...) returns a 100*(1-alpha)% confidence
interval for the true variance ratio var(X)/var(Y).
[H,P,CI,STATS] = vartest2(...) returns a structure with the
following fields:
• 'fstat' — Value of the test statistic
• 'df1' — Numerator degrees of freedom of the test
• 'df2' — Denominator degrees of freedom of the test
[...] = vartest2(X,Y,alpha,tail,dim) works along dimension
dim of X. To pass in the default values for alpha or tail use [].
Examples Is the variance significantly different for two model years, and what is a
confidence interval for the ratio of these variances?
load carsmall
[H,P,CI] =
vartest2(MPG(Model_Year==82),MPG(Model_Year==76))
See Also ansaribradley, vartest, vartestn, ttest2
18-1476
vartestn
Purpose Bartlett multiple-sample test for equal variances
Syntax vartestn(X)
vartestn(X,group)
p = vartestn(...)
[p,STATS] = vartestn(...)
[...] = vartestn(...,displayopt)
[...] = vartestn(...,testtype)
Description vartestn(X) performs Bartlett’s test for equal variances for the
columns of the matrix X. This is a test of the null hypothesis that the
columns of X come from normal distributions with the same variance,
against the alternative that they come from normal distributions with
different variances. The result is a display of a box plot of the groups,
and a summary table of statistics.
vartestn(X,group) requires a vector X, and a group argument that is
a categorical variable, vector, string array, or cell array of strings with
one row for each element of X. The X values corresponding to the same
value of group are placed in the same group. (See “Grouped Data” on
page 2-34.) The function tests for equal variances across groups.
vartestn treats NaNs as missing values and ignores them.
p = vartestn(...) returns the p value, i.e., the probability of
observing the given result, or one more extreme, by chance if the null
hypothesis of equal variances is true. Small values of p cast doubt on
the validity of the null hypothesis.
[p,STATS] = vartestn(...) returns a structure with the following
fields:
• 'chistat' — Value of the test statistic
• 'df' — Degrees of freedom of the test
[...] = vartestn(...,displayopt) determines if a box plot and
table are displayed. displayopt may be 'on' (the default) or 'off' .
18-1477
vartestn
[...] = vartestn(...,testtype) sets the test type. When testtype
is 'robust', vartestn performs Levene’s test in place of Bartlett’s
test, which is a useful alternative when the sample distributions are
not normal, and especially when they are prone to outliers. For this
test the STATS output structure has a field named 'fstat' containing
the test statistic, and 'df1' and 'df2' containing its numerator and
denominator degrees of freedom. When testtype is 'classical'
vartestn performs Bartlett’s test.
Examples Does the variance of mileage measurements differ significantly from
one model year to another?
load carsmall
p = vartestn(MPG,Model_Year)
p =
0.8327
18-1478
vartestn
See Also “Grouped Data” on page 2-34
vartest, vartest2, anova1
18-1479
categorical.vertcat
Purpose Vertical concatenation for categorical arrays
Syntax C = vertcat(dim,A,B,...)
C = vertcat(A,B)
Description C = vertcat(dim,A,B,...) vertically concatenates the categorical
arrays A,B,... . For matrices, all inputs must have the same number
of rows. For n-D arrays, all inputs must have the same sizes except in
the second dimension. The set of categorical levels for C is the sorted
union of the sets of levels of the inputs, as determined by their labels.
C = vertcat(A,B) is called for the syntax [A B].
See Also cat, horzcat
18-1480
dataset.vertcat
Purpose Vertical concatenation for dataset arrays
Syntax ds = vertcat(ds1, ds2, ...)
Description ds = vertcat(ds1, ds2, ...) vertically concatenates the dataset
arrays ds1, ds2, ... . Observation names, when present, must be
unique across datasets. vertcat fills in default observation names for
the output when some of the inputs have names and some do not.
Variable names for all dataset arrays must be identical except for
order. vertcat concatenates by matching variable names. vertcat
assigns values for the "per-variable" properties (e.g., Units and
VarDescription) in ds from the corresponding property values in ds1.
See Also cat, horzcat
18-1481
classregtree.view
Purpose Plot tree
Syntax view(t)
view(t,param1,val1,param2,val2,...)
Description view(t) displays the decision tree t as computed by classregtree in
a figure window. Each branch in the tree is labeled with its decision
rule, and each terminal node is labeled with the predicted value for that
node. Click any node to get more information about it. The information
displayed is specified by the Click to display pop-up menu at the top
of the figure.
view(t,param1,val1,param2,val2,...) specifies optional parameter
name/value pairs:
• 'names' — A cell array of names for the predictor variables, in the
order in which they appear in the matrix X from which the tree was
created. (See classregtree.)
• 'prunelevel' — Initial pruning level to display.
For each branch node, the left child node corresponds to the points that
satisfy the condition, and the right child node corresponds to the points
that do not satisfy the condition.
Examples Create a classification tree for Fisher’s iris data:
load fisheriris;
t = classregtree(meas,species,...
'names',{'SL' 'SW' 'PL' 'PW'})
t =
Decision tree for classification
1 if PL<2.45 then node 2 elseif PL>=2.45 then node 3 else setosa
2 class = setosa
3 if PW<1.75 then node 4 elseif PW>=1.75 then node 5 else versicolor
4 if PL<4.95 then node 6 elseif PL>=4.95 then node 7 else versicolor
5 class = virginica
18-1482
classregtree.view
6 if PW<1.65 then node 8 elseif PW>=1.65 then node 9 else versicolor
7 class = virginica
8 class = versicolor
9 class = virginica
view(t)
References [1] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification
and Regression Trees. Boca Raton, FL: CRC Press, 1984.
18-1483
classregtree.view
See Also classregtree, eval, prune, test
18-1484
wblcdf
Purpose Weibull cumulative distribution function
Syntax P = wblcdf(X,A,B)
[P,PLO,PUP] = wblcdf(X,A,B,PCOV,alpha)
Description P = wblcdf(X,A,B) computes the cdf of the Weibull distribution with
scale parameter A and shape parameter B, at each of the values in X. X,
A, and B can be vectors, matrices, or multidimensional arrays that all
have the same size. A scalar input is expanded to a constant array of
the same size as the other inputs. The default values for A and B are
both 1. The parameters A and B must be positive.
[P,PLO,PUP] = wblcdf(X,A,B,PCOV,alpha) returns confidence
bounds for P when the input parameters A and B are estimates. PCOV is
the 2-by-2 covariance matrix of the estimated parameters. alpha has a
default value of 0.05, and specifies 100(1 - alpha)% confidence bounds.
PLO and PUP are arrays of the same size as P containing the lower and
upper confidence bounds.
The function wblcdf computes confidence bounds for P using a normal
approximation to the distribution of the estimate
bˆ ( log x − log aˆ )
and then transforms those bounds to the scale of the output P. The
computed bounds give approximately the desired confidence level when
you estimate mu, sigma, and PCOV from large samples, but in smaller
samples other methods of computing the confidence bounds might be
more accurate.
The Weibull cdf is
b b
⎛t⎞ ⎛ x⎞
− −⎜ ⎟
x − b b−1 ⎜⎝ a ⎟⎠
p = F ( x | a, b) = ∫0 ba t e dt = 1− e ⎝a⎠ I( 0,∞ ) ( x )
Examples What is the probability that a value from a Weibull distribution with
parameters a = 0.15 and b = 0.8 is less than 0.5?
18-1485
wblcdf
probability = wblcdf(0.5, 0.15, 0.8)
probability =
0.9272
How sensitive is this result to small changes in the parameters?
[A, B] = meshgrid(0.1:0.05:0.2,0.2:0.05:0.3);
probability = wblcdf(0.5, A, B)
probability =
0.7484 0.7198 0.6991
0.7758 0.7411 0.7156
0.8022 0.7619 0.7319
See Also cdf, wblpdf, wblinv, wblstat, wblfit, wbllike, wblrnd
“Weibull Distribution” on page B-103
18-1486
wblfit
Purpose Weibull parameter estimates
Syntax parmhat = wblfit(data)
[parmhat,parmci] = wblfit(data)
parmhat,parmci] = wblfit(data,alpha)
[...] = wblfit(data,alpha,censoring)
[...] = wblfit(data,alpha,censoring,freq)
[...] = wblfit(...,options)
Description parmhat = wblfit(data) returns the maximum likelihood estimates,
parmhat, of the parameters of the Weibull distribution given the values
in the vector data, which must be positive. parmhat is a two-element
row vector: parmhat(1) estimates the Weibull parameter a, and
parmhat(2) estimates the Weibull parameter b, in the pdf
b
⎛ x⎞
−
− b b−1 ⎜⎝ a ⎟⎠
y = f ( x | a, b) = ba x e I( 0,∞ ) ( x )
[parmhat,parmci] = wblfit(data) returns 95% confidence intervals
for the estimates of a and b in the 2-by-2 matrix parmci. The first
row contains the lower bounds of the confidence intervals for the
parameters, and the second row contains the upper bounds of the
confidence intervals.
[parmhat,parmci] = wblfit(data,alpha) returns 100(1 - alpha)%
confidence intervals for the parameter estimates.
[...] = wblfit(data,alpha,censoring) accepts a Boolean vector,
censoring, of the same size as data, which is 1 for observations that
are right-censored and 0 for observations that are observed exactly.
[...] = wblfit(data,alpha,censoring,freq) accepts a frequency
vector, freq, of the same size as data. The vector freq typically
contains integer frequencies for the corresponding elements in data, but
can contain any non-negative values. Pass in [] for alpha, censoring,
or freq to use their default values.
18-1487
wblfit
[...] = wblfit(...,options) accepts a structure, options, that
specifies control parameters for the iterative algorithm the function
uses to compute maximum likelihood estimates. The Weibull fit
function accepts an options structure that can be created using the
function statset. Enter statset ('wblfit') to see the names and
default values of the parameters that lognfit accepts in the options
structure. See the reference page for statset for more information
about these options.
Examples data = wblrnd(0.5,0.8,100,1);
[parmhat, parmci] = wblfit(data)
parmhat =
0.5861 0.8567
parmci =
0.4606 0.7360
0.7459 0.9973
See Also mle, wbllike, wblpdf, wblcdf, wblinv, wblstat, wblrnd
“Weibull Distribution” on page B-103
18-1488
wblinv
Purpose Weibull inverse cumulative distribution function
Syntax X = wblinv(P,A,B)
[X,XLO,XUP] = wblinv(P,A,B,PCOV,alpha)
Description X = wblinv(P,A,B) returns the inverse cumulative distribution
function (cdf) for a Weibull distribution with scale parameter A and
shape parameter B, evaluated at the values in P. P, A, and B can be
vectors, matrices, or multidimensional arrays that all have the same
size. A scalar input is expanded to a constant array of the same size as
the other inputs. The default values for A and B are both 1.
[X,XLO,XUP] = wblinv(P,A,B,PCOV,alpha) returns confidence
bounds for X when the input parameters A and B are estimates.
PCOV is a 2-by-2 matrix containing the covariance matrix of the
estimated parameters. alpha has a default value of 0.05, and specifies
100(1 - alpha)% confidence bounds. XLO and XUP are arrays of the same
size as X containing the lower and upper confidence bounds.
The function wblinv computes confidence bounds for X using a normal
approximation to the distribution of the estimate
log q
log a +
b
where q is the Pth quantile from a Weibull distribution with scale
and shape parameters both equal to 1. The computed bounds give
approximately the desired confidence level when you estimate mu,
sigma, and PCOV from large samples, but in smaller samples other
methods of computing the confidence bounds might be more accurate.
The inverse of the Weibull cdf is
1/ b
x = F −1 ( p| a, b) = − a ⎡⎣ ln (1 − p ) ⎤⎦ I[ 0,1] ( p)
Examples The lifetimes (in hours) of a batch of light bulbs has a Weibull
distribution with parameters a = 200 and b = 6.
18-1489
wblinv
Find the median lifetime of the bulbs:
life = wblinv(0.5, 200, 6)
life =
188.1486
Generate 100 random values from this distribution, and estimate the
90th percentile (with confidence bounds) from the random sample
x = wblrnd(200,6,100,1);
p = wblfit(x)
[nlogl,pcov] = wbllike(p,x)
[q90,q90lo,q90up] = wblinv(0.9,p(1),p(2),pcov)
p =
204.8918 6.3920
nlogl =
496.8915
pcov =
11.3392 0.5233
0.5233 0.2573
q90 =
233.4489
q90lo =
18-1490
wblinv
226.0092
q90up =
241.1335
See Also icdf, wblcdf, wblpdf, wblstat, wblfit, wbllike, wblrnd
“Weibull Distribution” on page B-103
18-1491
wbllike
Purpose Weibull negative log-likelihood
Syntax nlogL = wbllike(params,data)
[logL,AVAR] = wbllike(params,data)
[...] = wbllike(params,data,censoring)
[...] = wbllike(params,data,censoring,freq)
Description nlogL = wbllike(params,data) returns the Weibull log-likelihood.
params(1) is the scale parameter, A, and params(2) is the shape
parameter, B.
[logL,AVAR] = wbllike(params,data) also returns AVAR, which is the
asymptotic variance-covariance matrix of the parameter estimates if
the values in params are the maximum likelihood estimates. AVAR is the
inverse of Fisher’s information matrix. The diagonal elements of AVAR
are the asymptotic variances of their respective parameters.
[...] = wbllike(params,data,censoring) accepts a Boolean
vector, censoring, of the same size as data, which is 1 for observations
that are right-censored and 0 for observations that are observed exactly.
[...] = wbllike(params,data,censoring,freq) accepts a
frequency vector, freq, of the same size as data. freq typically contains
integer frequencies for the corresponding elements in data, but can
contain any nonnegative values. Pass in [] for censoring to use its
default value.
The Weibull negative log-likelihood for uncensored data is
n
( − log L ) = − log ∏ f ( a, b| xi ) = −∑ log f ( a, b| xi )
i=1 i=1
where f is the Weibull pdf.
wbllike is a utility function for maximum likelihood estimation.
Examples This example continues the example from wblfit.
r = wblrnd(0.5,0.8,100,1);
18-1492
wbllike
[logL, AVAR] = wbllike(wblfit(r),r)
logL =
47.3349
AVAR =
0.0048 0.0014
0.0014 0.0040
References [1] Patel, J. K., C. H. Kapadia, and D. B. Owen. Handbook of Statistical
Distributions. New York: Marcel Dekker, 1976.
See Also wblfit, wblpdf, wblcdf, wblinv, wblstat, wblrnd
“Weibull Distribution” on page B-103
18-1493
wblpdf
Purpose Weibull probability density function
Syntax Y = wblpdf(X,A,B)
Description Y = wblpdf(X,A,B) computes the Weibull pdf at each of the values in
X using the corresponding scale parameter, A and shape parameter, B.
X, A, and B can be vectors, matrices, or multidimensional arrays that
all have the same size. A scalar input is expanded to a constant array
of the same size as the other inputs. The parameters in A and B must
be positive.
The Weibull pdf is
b
⎛ x⎞
−
− b b−1 ⎜⎝ a ⎟⎠
y = f ( x | a, b) = ba x e I( 0,∞ ) ( x )
Some references refer to the Weibull distribution with a single
parameter. This corresponds to wblpdf with A = 1.
Examples The exponential distribution is a special case of the Weibull distribution.
lambda = 1:6;
y = wblpdf(0.1:0.1:0.6,lambda,1)
y =
0.9048 0.4524 0.3016 0.2262 0.1810 0.1508
y1 = exppdf(0.1:0.1:0.6,lambda)
y1 =
0.9048 0.4524 0.3016 0.2262 0.1810 0.1508
References [1] Devroye, L. Non-Uniform Random Variate Generation. New York:
Springer-Verlag, 1986.
See Also pdf, wblcdf, wblfit, wblinv, wbllike, wblplot, wblrnd, wblstat
“Weibull Distribution” on page B-103
18-1494
wblplot
Purpose Weibull probability plot
Syntax wblplot(X)
h = wblplot(X)
Description wblplot(X) displays a Weibull probability plot of the data in X. If X is a
matrix, wblplot displays a plot for each column.
h = wblplot(X) returns handles to the plotted lines.
The purpose of a Weibull probability plot is to graphically assess
whether the data in X could come from a Weibull distribution. If the
data are Weibull the plot will be linear. Other distribution types might
introduce curvature in the plot. wblplot uses midpoint probability
plotting positions. Use probplot when the data included censored
observations.
Examples r = wblrnd(1.2,1.5,50,1);
wblplot(r)
18-1495
wblplot
See Also probplot, normplot, wblcdf, wblfit, wblinv, wbllike, wblpdf,
wblrnd, wblstat
“Weibull Distribution” on page B-103
18-1496
wblrnd
Purpose Weibull random numbers
Syntax R = wblrnd(A,B)
R = wblrnd(A,B,v)
R = wblrnd(A,B,m,n)
Description R = wblrnd(A,B) generates random numbers for the Weibull
distribution with scale parameter, A and shape parameter, B. The input
arguments A and B can be either scalars or matrices. A and B, can be
vectors, matrices, or multidimensional arrays that all have the same
size. A scalar input is expanded to a constant array of the same size
as the other input.
R = wblrnd(A,B,v) generates random numbers for the Weibull
distribution with parameters A and B, where v is a row vector. If v is a
1-by-2 vector, R is a matrix with v(1) rows and v(2) columns. If v is
1-by-n, R is an n-dimensional array.
R = wblrnd(A,B,m,n) generates random numbers for the Weibull
distribution with parameters A and B, where scalars m and n are the
row and column dimensions of R.
Devroye [1] refers to the Weibull distribution with a single parameter;
this is wblrnd with A = 1.
Examples n1 = wblrnd(0.5:0.5:2,0.5:0.5:2)
n1 =
0.0178 0.0860 2.5216 0.9124
n2 = wblrnd(1/2,1/2,[1 6])
n2 =
0.0046 1.7214 2.2108 0.0367 0.0531 0.0917
References [1] Devroye, L. Non-Uniform Random Variate Generation. New York:
Springer-Verlag, 1986.
See Also random, wblpdf, wblcdf, wblinv, wblstat, wblfit, wbllike
18-1497
wblrnd
“Weibull Distribution” on page B-103
18-1498
wblstat
Purpose Weibull mean and variance
Syntax [M,V] = wblstat(A,B)
Description [M,V] = wblstat(A,B) returns the mean of and variance for the
Weibull distribution with scale parameter, A and shape parameter, B.
Vector or matrix inputs for A and B must have the same size, which
is also the size of M and V. A scalar input for A or B is expanded to a
constant matrix with the same dimensions as the other input.
The mean of the Weibull distribution with parameters a and b is
( )
a ⎡⎢Γ 1 + b−1 ⎤⎥
⎣ ⎦
and the variance is
⎡
( ) ( )
2⎤
a2 ⎢Γ 1 + 2b−1 − Γ 1 + b−1 ⎥
⎣ ⎦
Examples [m,v] = wblstat(1:4,1:4)
m =
1.0000 1.7725 2.6789 3.6256
v =
1.0000 0.8584 0.9480 1.0346
wblstat(0.5,0.7)
ans =
0.6329
See Also wblpdf, wblcdf, wblinv, wblfit, wbllike, wblrnd
“Weibull Distribution” on page B-103
18-1499
wishrnd
Purpose Wishart random numbers
Syntax W = wishrnd(Sigma,df)
W = wishrnd(Sigma,df,D)
[W,D] = wishrnd(Sigma,df)
Description W = wishrnd(Sigma,df) generates a random matrix W having the
Wishart distribution with covariance matrix Sigma and with df degrees
of freedom. The inverse of W has the Inverse Wishart distribution with
parameters Tau = inv(Sigma) and df degrees of freedom.
W = wishrnd(Sigma,df,D) expects D to be the Cholesky factor of Sigma.
If you call wishrnd multiple times using the same value of Sigma, it’s
more efficient to supply D instead of computing it each time.
[W,D] = wishrnd(Sigma,df) returns D so you can provide it as input
in future calls to wishrnd.
This function defines the parameter Sigma so that the mean of the
output matrix is Sigma*df
See Also iwishrnd
“Wishart Distribution” on page B-106
18-1500
TreeBagger.X property
Purpose X data used to create ensemble
Description The X property is a numeric matrix of size Nobs-by-Nvars, where Nobs is
the number of observations (rows) and Nvars is the number of variables
(columns) in the training data. This matrix contains the predictor (or
feature) values.
18-1501
xptread
Purpose Create dataset array from data stored in SAS XPORT format file
Syntax data = xptread
data = xptread(filename)
[data,missing] = xptread(filename)
xptread(...,'ReadObsNames',true)
Description data = xptread displays a dialog box for selecting a file, then reads
data from the file into a dataset array. The file must be in the SAS
XPORT format.
data = xptread(filename) retrieves data from a SAS XPORT format
file filename. The XPORT format allows for 28 missing data types,
represented in the file by an upper case letter, '.' or '_'. xptread
converts All missing data to NaN values in data. However, if you need
the specific missing types then you can recover this information by
specifying a second output.
[data,missing] = xptread(filename) returns a nominal array,
missing, of the same size as data containing the missing data type
information from the xport format file. The entries are undefined for
values that are not present and are one of '.', '_', 'A',...,'Z' for
missing values.
xptread(...,'ReadObsNames',true) treats the first variable in the
file as observation names. The default value is false.
xptread only supports single data sets per file.
Examples Read in a SAS XPORT format dataset:
data = xptread('sample.xpt')
See Also dataset | dataset.export
18-1502
x2fx
Purpose Convert predictor matrix to design matrix
Syntax D = x2fx(X,model)
D = x2fx(X,model,categ)
D = x2fx(X,model,categ,catlevels)
Description D = x2fx(X,model) converts a matrix of predictors X to a design matrix
D for regression analysis. Distinct predictor variables should appear
in different columns of X.
The optional input model controls the regression model. By default,
x2fx returns the design matrix for a linear additive model with a
constant term. model is one of the following strings:
• 'linear' — Constant and linear terms. This is the default.
• 'interaction' — Constant, linear, and interaction terms
• 'quadratic' — Constant, linear, interaction, and squared terms
• 'purequadratic' — Constant, linear, and squared terms
If X has n columns, the order of the columns of D for a full quadratic
model is:
1 The constant term
2 The linear terms (the columns of X, in order 1, 2, ..., n)
3 The interaction terms (pairwise products of the columns of X, in order
(1, 2), (1, 3), ..., (1, n), (2, 3), ..., (n–1, n))
4 The squared terms (in order 1, 2, ..., n)
Other models use a subset of these terms, in the same order.
Alternatively, model can be a matrix specifying polynomial terms of
arbitrary order. In this case, model should have one column for each
column in X and one row for each term in the model. The entries in
any row of model are powers for the corresponding columns of X. For
18-1503
x2fx
example, if X has columns X1, X2, and X3, then a row [0 1 2] in model
specifies the term (X1.^0).*(X2.^1).*(X3.^2). A row of all zeros in
model specifies a constant term, which can be omitted.
D = x2fx(X,model,categ) treats columns with numbers listed in
the vector categ as categorical variables. Terms involving categorical
variables produce dummy variable columns in D. Dummy variables
are computed under the assumption that possible categorical levels
are completely enumerated by the unique values that appear in the
corresponding column of X.
D = x2fx(X,model,categ,catlevels) accepts a vector catlevels
the same length as categ, specifying the number of levels in each
categorical variable. In this case, values in the corresponding column of
X must be integers in the range from 1 to the specified number of levels.
Not all of the levels need to appear in X.
Examples Example 1
The following converts 2 predictors X1 and X2 (the columns of X) into a
design matrix for a full quadratic model with terms constant, X1, X2,
X1.*X2, X1.^2, and X2.^2.
X = [1 10
2 20
3 10
4 20
5 15
6 15];
D = x2fx(X,'quadratic')
D =
1 1 10 10 1 100
1 2 20 40 4 400
1 3 10 30 9 100
1 4 20 80 16 400
1 5 15 75 25 225
1 6 15 90 36 225
18-1504
x2fx
Example 2
The following converts 2 predictors X1 and X2 (the columns of X) into
a design matrix for a quadratic model with terms constant, X1, X2,
X1.*X2, and X1.^2.
X = [1 10
2 20
3 10
4 20
5 15
6 15];
model = [0 0
1 0
0 1
1 1
2 0];
D = x2fx(X,model)
D =
1 1 10 10 1
1 2 20 40 4
1 3 10 30 9
1 4 20 80 16
1 5 15 75 25
1 6 15 90 36
See Also regstats, rstool, candexch, candgen, cordexch, and rowexch.
18-1505
TreeBagger.Y property
Purpose Y data used to create ensemble
Description The Y property is an array of true class labels for classification, or
response values for regression. Y can be a numeric column vector, a
character matrix, or a cell array of strings.
18-1506
zscore
Purpose Standardized z-scores
Syntax Z = zscore(X)
[Z,mu,sigma] = zscore(X)
[...] = zscore(X,1)
[...] = zscore(X,flag,dim)
Description Z = zscore(X) returns a centered, scaled version of X, the same
size as X. For vector input x, output is the vector of z-scores z =
(x mean(x))./std(x). For matrix input X, z-scores are computed
using the mean and standard deviation along each column of X. For
higher-dimensional arrays, z-scores are computed using the mean and
standard deviation along the first non-singleton dimension.
The columns of Z have mean zero and standard deviation one (unless a
column of X is constant, in which case that column of Z is constant at 0).
z-scores are used to put data on the same scale before further analysis.
[Z,mu,sigma] = zscore(X) also returns mean(X) in mu and std(X) in
sigma.
[...] = zscore(X,1) normalizes X using std(X,1), that is, by
computing the standard deviation(s) using n rather than n–1, where n
is the length of the dimension along which zscore works. zscore(X,0)
is the same as zscore(X).
[...] = zscore(X,flag,dim) standardizes X by working along the
dimension dim of X. Set flag to 0 to use the default normalization by
n–1; set flag to 1 to use n.
Examples Compare the predictors in the Moore data on original and standardized
scales:
load moore
predictors = moore(:,1:5);
subplot(2,1,1),plot(predictors)
subplot(2,1,2),plot(zscore(predictors))
18-1507
zscore
See Also mean, std
18-1508
ztest
Purpose z-test
Syntax h = ztest(x,m,sigma)
h = ztest(...,alpha)
h = ztest(...,alpha,tail)
h = ztest(...,alpha,tail,dim)
[h,p] = ztest(...)
[h,p,ci] = ztest(...)
[h,p,ci,zval] = ztest(...)
Description h = ztest(x,m,sigma) performs a z-test of the null hypothesis that
data in the vector x are a random sample from a normal distribution
with mean m and standard deviation sigma, against the alternative
that the mean is not m. The result of the test is returned in h. h =
1 indicates a rejection of the null hypothesis at the 5% significance
level. h = 0 indicates a failure to reject the null hypothesis at the 5%
significance level.
x can also be a matrix or an N-dimensional array. For matrices, ztest
performs separate z-tests along each column of x and returns a vector
of results. For N-dimensional arrays, ztest works along the first
non-singleton dimension of x.
The test treats NaN values as missing data, and ignores them.
h = ztest(...,alpha) performs the test at the (100*alpha)%
significance level. The default, when unspecified, is alpha = 0.05.
h = ztest(...,alpha,tail) performs the test against the alternative
specified by the string tail. There are three options for tail:
• 'both' — Mean is not m (two-tailed test). This is the default, when
tail is unspecified.
• 'right' — Mean is greater than m (right-tail test)
• 'left' — Mean is less than m (left-tail test)
tail must be a single string, even when x is a matrix or an
N-dimensional array.
18-1509
ztest
h = ztest(...,alpha,tail,dim) works along dimension dim of x. Use
[] to pass in default values for alpha or tail.
[h,p] = ztest(...) returns the p value of the test. The p value is the
probability, under the null hypothesis, of observing a value as extreme
or more extreme of the test statistic
x−
z=
/ n
where x is the sample mean, μ = m is the hypothesized population
mean, σ is the population standard deviation, and n is the sample
size. Under the null hypothesis, the test statistic will have a standard
normal distribution, N(0,1).
[h,p,ci] = ztest(...) returns a 100*(1 – alpha)% confidence
interval on the population mean.
[h,p,ci,zval] = ztest(...) returns the value of the test statistic.
Examples Simulate a random sample of size 100 from a normal distribution with
mean 0.1 and standard deviation 1:
x = normrnd(0.1,1,1,100);
Test the null hypothesis that the sample comes from a standard normal
distribution:
[h,p,ci] = ztest(x,0,1)
h =
0
p =
0.1391
ci =
-0.0481 0.3439
The test fails to reject the null hypothesis at the default α = 0.05
significance level. Under the null hypothesis, the probability of
observing a value as extreme or more extreme of the test statistic, as
18-1510
ztest
indicated by the p value, is greater than α. The 95% confidence interval
on the mean contains 0.
Simulate a larger random sample of size 1000 from the same
distribution:
y = normrnd(0.1,1,1,1000);
Test again if the sample comes from a normal distribution with mean 0:
[h,p,ci] = ztest(y,0,1)
h =
1
p =
5.5160e-005
ci =
0.0655 0.1895
This time the test rejects the null hypothesis at the default α = 0.05
significance level. The p value has fallen below α = 0.05 and the 95%
confidence interval on the mean does not contain 0.
Because the p value of the sample y is less than 0.01, the test will still
reject the null hypothesis when the significance level is lowered to α
= 0.01:
[h,p,ci] = ztest(y,0,1,0.01)
h =
1
p =
5.5160e-005
ci =
0.0461 0.2090
This example will produce slightly different results each time it is run,
because of the random sampling.
See Also ttest, ttest2
18-1511
ztest
18-1512
A
Data Sets
A Data Sets
Statistics Toolbox software includes the sample data sets in the following
table.
To load a data set into the MATLAB workspace, type:
load filename
where filename is one of the files listed in the table.
Data sets contain individual data variables, description variables with
references, and dataset arrays encapsulating the data set and its description,
as appropriate.
File Description of Data Set
acetylene.mat Chemical reaction data with correlated predictors
arrhythmia.mat Cardiac arrhythmia data from the UCI machine
learning repository
carbig.mat Measurements of cars, 1970–1982
carsmall.mat Subset of carbig.mat. Measurements of cars, 1970,
1976, 1982
cereal.mat Breakfast cereal ingredients
cities.mat Quality of life ratings for U.S. metropolitan areas
discrim.mat A version of cities.mat used for discriminant
analysis
examgrades.mat Exam grades on a scale of 0–100
fisheriris.mat Fisher’s 1936 iris data
flu.mat Google Flu Trends estimated ILI (influenza-like
illness) percentage for various regions of the US,
and CDC weighted ILI percentage based on sentinel
provider reports
gas.mat Gasoline prices around the state of Massachusetts
in 1993
hald.mat Heat of cement vs. mix of ingredients
hogg.mat Bacteria counts in different shipments of milk
A-2
Data Sets
File Description of Data Set
hospital.mat Simulated hospital data
imports-85.mat 1985 Auto Imports Database from the UCI
repository
ionosphere.mat Ionosphere dataset from the UCI machine learning
repository
kmeansdata.mat Four-dimensional clustered data
lawdata.mat Grade point average and LSAT scores from 15 law
schools
mileage.mat Mileage data for three car models from two factories
moore.mat Biochemical oxygen demand on five predictors
morse.mat Recognition of Morse code distinctions by non-coders
ovariancancer.mat Grouped observations on 4000 predictors
parts.mat Dimensional run-out on 36 circular parts
polydata.mat Sample data for polynomial fitting
popcorn.mat Popcorn yield by popper type and brand
reaction.mat Reaction kinetics for Hougen-Watson model
sat.dat Scholastic Aptitude Test averages by gender and
test (table)
sat2.dat Scholastic Aptitude Test averages by gender and
test (csv)
spectra.mat NIR spectra and octane numbers of 60 gasoline
samples
stockreturns.mat Simulated stock returns
A-3
A Data Sets
A-4
B
Distribution Reference
• “Bernoulli Distribution” on page B-3
• “Beta Distribution” on page B-4
• “Binomial Distribution” on page B-7
• “Birnbaum-Saunders Distribution” on page B-10
• “Chi-Square Distribution” on page B-12
• “Copulas” on page B-14
• “Custom Distributions” on page B-15
• “Exponential Distribution” on page B-16
• “Extreme Value Distribution” on page B-19
• “F Distribution” on page B-25
• “Gamma Distribution” on page B-27
• “Gaussian Distribution” on page B-30
• “Gaussian Mixture Distributions” on page B-31
• “Generalized Extreme Value Distribution” on page B-32
• “Generalized Pareto Distribution” on page B-37
• “Geometric Distribution” on page B-41
• “Hypergeometric Distribution” on page B-43
• “Inverse Gaussian Distribution” on page B-45
• “Inverse Wishart Distribution” on page B-46
• “Johnson System” on page B-48
• “Logistic Distribution” on page B-49
B Distribution Reference
• “Loglogistic Distribution” on page B-50
• “Lognormal Distribution” on page B-51
• “Multinomial Distribution” on page B-54
• “Multivariate Gaussian Distribution” on page B-57
• “Multivariate Normal Distribution” on page B-58
• “Multivariate t Distribution” on page B-64
• “Nakagami Distribution” on page B-70
• “Negative Binomial Distribution” on page B-72
• “Noncentral Chi-Square Distribution” on page B-76
• “Noncentral F Distribution” on page B-78
• “Noncentral t Distribution” on page B-80
• “Nonparametric Distributions” on page B-82
• “Normal Distribution” on page B-83
• “Pareto Distribution” on page B-86
• “Pearson System” on page B-87
• “Piecewise Distributions” on page B-88
• “Poisson Distribution” on page B-89
• “Rayleigh Distribution” on page B-91
• “Rician Distribution” on page B-93
• “Student’s t Distribution” on page B-95
• “t Location-Scale Distribution” on page B-97
• “Uniform Distribution (Continuous)” on page B-99
• “Uniform Distribution (Discrete)” on page B-101
• “Weibull Distribution” on page B-103
• “Wishart Distribution” on page B-106
B-2
Bernoulli Distribution
Bernoulli Distribution
Definition of the Bernoulli Distribution
The Bernoulli distribution is a special case of the binomial distribution, with
n = 1.
See Also
“Discrete Distributions” on page 5-7
B-3
B Distribution Reference
Beta Distribution
In this section...
“Definition” on page B-4
“Background” on page B-4
“Parameters” on page B-5
“Example” on page B-6
“See Also” on page B-6
Definition
The beta pdf is
1
y = f ( x | a, b) = x a−1 (1 − x)b−1 I(0,1) ( x)
B(a, b)
where B( · ) is the Beta function. The indicator function I(0,1)(x) ensures that
only values of x in the range (0 1) have nonzero probability.
Background
The beta distribution describes a family of curves that are unique in that they
are nonzero only on the interval (0 1). A more general version of the function
assigns parameters to the endpoints of the interval.
The beta cdf is the same as the incomplete beta function.
The beta distribution has a functional relationship with the t distribution. If
Y is an observation from Student’s t distribution with ν degrees of freedom,
then the following transformation generates X, which is beta distributed.
1 1 Y
X= +
2 2 + Y2
B-4
Beta Distribution
⎛ ⎞
If Y~t(v), then X ∼ ⎜ , ⎟
⎝2 2⎠
This relationship is used to compute values of the t cdf and inverse function
as well as generating t distributed random numbers.
Parameters
Suppose you are collecting data that has hard lower and upper bounds of zero
and one respectively. Parameter estimation is the process of determining the
parameters of the beta distribution that fit this data best in some sense.
One popular criterion of goodness is to maximize the likelihood function. The
likelihood has the same form as the beta pdf. But for the pdf, the parameters
are known constants and the variable is x. The likelihood function reverses the
roles of the variables. Here, the sample values (the x’s) are already observed.
So they are the fixed constants. The variables are the unknown parameters.
Maximum likelihood estimation (MLE) involves calculating the values of the
parameters that give the highest likelihood given the particular set of data.
The function betafit returns the MLEs and confidence intervals for the
parameters of the beta distribution. Here is an example using random
numbers from the beta distribution with a = 5 and b = 0.2.
r = betarnd(5,0.2,100,1);
[phat, pci] = betafit(r)
phat =
4.5330 0.2301
pci =
2.8051 0.1771
6.2610 0.2832
The MLE for parameter a is 4.5330, compared to the true value of 5. The
95% confidence interval for a goes from 2.8051 to 6.2610, which includes
the true value.
Similarly the MLE for parameter b is 0.2301, compared to the true value
of 0.2. The 95% confidence interval for b goes from 0.1771 to 0.2832, which
B-5
B Distribution Reference
also includes the true value. In this made-up example you know the “true
value.” In experimentation you do not.
Example
The shape of the beta distribution is quite variable depending on the values of
the parameters, as illustrated by the plot below.
The constant pdf (the flat line) shows that the standard uniform distribution
is a special case of the beta distribution.
See Also
“Continuous Distributions (Data)” on page 5-4
B-6
Binomial Distribution
Binomial Distribution
In this section...
“Definition” on page B-7
“Background” on page B-7
“Parameters” on page B-8
“Example” on page B-9
“See Also” on page B-9
Definition
The binomial pdf is
⎛ n⎞
f (k| n, p) = ⎜ ⎟ pk (1 − p)n− k
⎝ k⎠
where k is the number of successes in n trials of a Bernoulli process with
probability of success p.
The binomial distribution is discrete, defined for integers k = 0, 1, 2, ... n,
where it is nonzero.
Background
The binomial distribution models the total number of successes in repeated
trials from an infinite population under the following conditions:
• Only two outcomes are possible on each of n trials.
• The probability of success for each trial is constant.
• All trials are independent of each other.
The binomial distribution is a generalization of the Bernoulli distribution; it
generalizes to the multinomial distribution.
B-7
B Distribution Reference
Parameters
Suppose you are collecting data from a widget manufacturing process, and
you record the number of widgets within specification in each batch of 100.
You might be interested in the probability that an individual widget is
within specification. Parameter estimation is the process of determining the
parameter, p, of the binomial distribution that fits this data best in some
sense.
One popular criterion of goodness is to maximize the likelihood function.
The likelihood has the same form as the binomial pdf above. But for the pdf,
the parameters (n and p) are known constants and the variable is x. The
likelihood function reverses the roles of the variables. Here, the sample values
(the x’s) are already observed. So they are the fixed constants. The variables
are the unknown parameters. MLE involves calculating the value of p that
give the highest likelihood given the particular set of data.
The function binofit returns the MLEs and confidence intervals for the
parameters of the binomial distribution. Here is an example using random
numbers from the binomial distribution with n = 100 and p = 0.9.
r = binornd(100,0.9)
r =
88
[phat, pci] = binofit(r,100)
phat =
0.8800
pci =
0.7998
0.9364
The MLE for parameter p is 0.8800, compared to the true value of 0.9. The
95% confidence interval for p goes from 0.7998 to 0.9364, which includes
the true value. In this made-up example you know the “true value” of p. In
experimentation you do not.
B-8
Binomial Distribution
Example
The following commands generate a plot of the binomial pdf for n = 10 and
p = 1/2.
x = 0:10;
y = binopdf(x,10,0.5);
plot(x,y,'+')
See Also
“Discrete Distributions” on page 5-7
B-9
B Distribution Reference
Birnbaum-Saunders Distribution
In this section...
“Definition” on page B-10
“Background” on page B-10
“Parameters” on page B-11
“See Also” on page B-11
Definition
The Birnbaum-Saunders distribution has the density function
⎧ ⎛ ⎞ ⎫⎛ ⎛
2
⎞⎞
⎪ ⎜ x − ⎟ ⎪⎜ ⎜ x + ⎟ ⎟
⎪ x ⎠ ⎪⎜ ⎝ x
exp ⎨− ⎝ ⎠⎟
1
⎬
2 ⎪ 2 2
⎪ ⎜ 2 x ⎟
⎪ ⎪⎝⎜ ⎟
⎩ ⎭ ⎠
with scale parameter β > 0 and shape parameter γ > 0, for x > 0.
If x has a Birnbaum-Saunders distribution with parameters β and γ, then
⎛ x ⎞
⎜ − ⎟
⎝ x ⎠
has a standard normal distribution.
Background
The Birnbaum-Saunders distribution was originally proposed as a lifetime
model for materials subject to cyclic patterns of stress and strain, where the
ultimate failure of the material comes from the growth of a prominent flaw.
In materials science, Miner’s Rule suggests that the damage occurring after n
cycles, at a stress level with an expected lifetime of N cycles, is proportional
B-10
Birnbaum-Saunders Distribution
to n / N. Whenever Miner’s Rule applies, the Birnbaum-Saunders model is a
reasonable choice for a lifetime distribution model.
Parameters
See mle, dfittool.
See Also
“Continuous Distributions (Data)” on page 5-4
B-11
B Distribution Reference
Chi-Square Distribution
In this section...
“Definition” on page B-12
“Background” on page B-12
“Example” on page B-13
“See Also” on page B-13
Definition
The χ2 pdf is
x(
−2 ) / 2 − x / 2
e
y = f ( x | ) =
22 Γ ( / 2)
where Γ( · ) is the Gamma function, and ν is the degrees of freedom.
Background
The χ2 distribution is a special case of the gamma distribution where b = 2 in
the equation for gamma distribution below.
x
1
y = f ( x | a, b) = x a−1 e b
ba Γ ( a )
The χ2 distribution gets special attention because of its importance in normal
sampling theory. If a set of n observations is normally distributed with
variance σ2, and s2 is the sample standard deviation, then
( n − 1) s2
∼ 2 ( n − 1)
2
B-12
Chi-Square Distribution
This relationship is used to calculate confidence intervals for the estimate of
the normal parameter σ2 in the function normfit.
Example
The χ2 distribution is skewed to the right especially for few degrees of freedom
(ν). The plot shows the χ2 distribution with four degrees of freedom.
x = 0:0.2:15;
y = chi2pdf(x,4);
plot(x,y)
See Also
“Continuous Distributions (Statistics)” on page 5-6
B-13
B Distribution Reference
Copulas
See “Copulas” on page 5-107.
B-14
Custom Distributions
Custom Distributions
User-defined custom distributions, created using files and function handles,
are supported by the Statistics Toolbox functions pdf, cdf, icdf, and mle, and
the Statistics Toolbox GUI dfittool.
B-15
B Distribution Reference
Exponential Distribution
In this section...
“Definition” on page B-16
“Background” on page B-16
“Parameters” on page B-16
“Example” on page B-17
“See Also” on page B-18
Definition
The exponential pdf is
−x
1
y = f ( x| ) = e
Background
Like the chi-square distribution, the exponential distribution is a special case
of the gamma distribution (obtained by setting a = 1)
x
1
y = f ( x | a, b) = x a−1 e b
ba Γ ( a )
where Γ( · ) is the Gamma function.
The exponential distribution is special because of its utility in modeling
events that occur randomly over time. The main application area is in studies
of lifetimes.
Parameters
Suppose you are stress testing light bulbs and collecting data on their
lifetimes. You assume that these lifetimes follow an exponential distribution.
B-16
Exponential Distribution
You want to know how long you can expect the average light bulb to last.
Parameter estimation is the process of determining the parameters of the
exponential distribution that fit this data best in some sense.
One popular criterion of goodness is to maximize the likelihood function. The
likelihood has the same form as the exponential pdf above. But for the pdf,
the parameters are known constants and the variable is x. The likelihood
function reverses the roles of the variables. Here, the sample values (the x’s)
are already observed. So they are the fixed constants. The variables are the
unknown parameters. MLE involves calculating the values of the parameters
that give the highest likelihood given the particular set of data.
The function expfit returns the MLEs and confidence intervals for the
parameters of the exponential distribution. Here is an example using random
numbers from the exponential distribution with µ = 700.
lifetimes = exprnd(700,100,1);
[muhat, muci] = expfit(lifetimes)
muhat =
672.8207
muci =
547.4338
810.9437
The MLE for parameter µ is 672, compared to the true value of 700. The 95%
confidence interval for µ goes from 547 to 811, which includes the true value.
In the life tests you do not know the true value of µ so it is nice to have a
confidence interval on the parameter to give a range of likely values.
Example
For exponentially distributed lifetimes, the probability that an item will
survive an extra unit of time is independent of the current age of the item.
The example shows a specific case of this special property.
l = 10:10:60;
B-17
B Distribution Reference
lpd = l+0.1;
deltap = (expcdf(lpd,50)-expcdf(l,50))./(1-expcdf(l,50))
deltap =
0.0020 0.0020 0.0020 0.0020 0.0020 0.0020
The following commands generate a plot of the exponential pdf with its
parameter (and mean), µ, set to 2.
x = 0:0.1:10;
y = exppdf(x,2);
plot(x,y)
See Also
“Continuous Distributions (Data)” on page 5-4
B-18
Extreme Value Distribution
Extreme Value Distribution
In this section...
“Definition” on page B-19
“Background” on page B-19
“Parameters” on page B-21
“Example” on page B-22
“See Also” on page B-24
Definition
The probability density function for the extreme value distribution with
location parameter µ and scale parameter σ is
⎛x−⎞ ⎛ ⎛ x − ⎞⎞
y = f ( x | , ) = −1 exp ⎜ ⎟ exp ⎜ − exp ⎜ ⎟ ⎟
⎝ ⎠ ⎝ ⎝ ⎠⎠
If T has a Weibull distribution with parameters a and b, then log T has an
extreme value distribution with parameters µ = log a and σ = 1/b.
Background
Extreme value distributions are often used to model the smallest or largest
value among a large set of independent, identically distributed random values
representing measurements or observations. The extreme value distribution
is appropriate for modeling the smallest value from a distribution whose tails
decay exponentially fast, for example, the normal distribution. It can also
model the largest value from a distribution, such as the normal or exponential
distributions, by using the negative of the original values.
For example, the following fits an extreme value distribution to minimum
values taken over 1000 sets of 500 observations from a normal distribution:
xMinima = min(randn(1000,500), [], 2);
paramEstsMinima = evfit(xMinima);
y = linspace(-5,-1.5,1001);
B-19
B Distribution Reference
hist(xMinima,-4.75:.25:-1.75);
p = evpdf(y,paramEstsMinima(1),paramEstsMinima(2));
line(y,.25*length(xMinima)*p,'color','r')
The following fits an extreme value distribution to the maximum values in
each set of observations:
xMaxima = max(randn(1000,500), [], 2);
paramEstsMaxima = evfit(-xMaxima);
y = linspace(1.5,5,1001);
hist(xMaxima,1.75:.25:4.75);
p = evpdf(-y,paramEstsMaxima(1),paramEstsMaxima(2));
line(y,.25*length(xMaxima)*p,'color','r')
B-20
Extreme Value Distribution
Although the extreme value distribution is most often used as a model for
extreme values, you can also use it as a model for other types of continuous
data. For example, extreme value distributions are closely related to the
Weibull distribution. If T has a Weibull distribution, then log(T) has a type 1
extreme value distribution.
Parameters
The function evfit returns the maximum likelihood estimates (MLEs) and
confidence intervals for the parameters of the extreme value distribution. The
following example shows how to fit some sample data using evfit, including
estimates of the mean and variance from the fitted distribution.
Suppose you want to model the size of the smallest washer in each batch
of 1000 from a manufacturing process. If you believe that the sizes are
B-21
B Distribution Reference
independent within and between each batch, you can fit an extreme value
distribution to measurements of the minimum diameter from a series of eight
experimental batches. The following code returns the MLEs of the distribution
parameters as parmhat and the confidence intervals as the columns of parmci.
x = [19.774 20.141 19.44 20.511 21.377 19.003 19.66 18.83];
[parmhat, parmci] = evfit(x)
parmhat =
20.2506 0.8223
parmci =
19.644 0.49861
20.857 1.3562
You can find mean and variance of the extreme value distribution with these
parameters using the function evstat.
[meanfit, varfit] = evstat(parmhat(1),parmhat(2))
meanfit =
19.776
varfit =
1.1123
Example
The following code generates a plot of the pdf for the extreme value
distribution.
t = [-5:.01:2];
y = evpdf(t);
plot(t,y)
B-22
Extreme Value Distribution
The extreme value distribution is skewed to the left, and its general shape
remains the same for all parameter values. The location parameter, mu, shifts
the distribution along the real line, and the scale parameter, sigma, expands
or contracts the distribution. This example plots the probability function for
different combinations of mu and sigma.
x = -15:.01:5;
plot(x,evpdf(x,2,1),'-', ...
x,evpdf(x,0,2),':', ...
x,evpdf(x,-2,4),'-.');
legend({'mu = 2, sigma = 1', ...
'mu = 0, sigma = 2', ...
'mu = -2, sigma = 4'}, ...
'Location','NW')
xlabel('x')
ylabel('f(x|mu,sigma)')
B-23
B Distribution Reference
See Also
“Continuous Distributions (Data)” on page 5-4
B-24
F Distribution
F Distribution
In this section...
“Definition” on page B-25
“Background” on page B-25
“Example” on page B-26
“See Also” on page B-26
Definition
The pdf for the F distribution is
⎡ ( + 2 ) ⎤ 1 1 −2
Γ⎢ 1 ⎥
y = f ( x | 1 , 2 ) = ⎣
2 ⎦ ⎛ 1 ⎞ 2 x 2
⎜ ⎟ 1 + 2
⎛ ⎞ ⎛ ⎞
Γ⎜ 1 ⎟Γ⎜ 2 ⎟ ⎝ 2 ⎠ ⎡ ⎛ ⎞ ⎤ 2
⎝ 2⎠ ⎝ 2 ⎠ 1
⎢1 + ⎜ ⎟ x ⎥
⎣ ⎝ 2 ⎠ ⎦
where Γ( · ) is the Gamma function.
Background
The F distribution has a natural relationship with the chi-square distribution.
If χ1 and χ2 are both chi-square with ν1 and ν2 degrees of freedom respectively,
then the statistic F below is F-distributed.
1
1
F ( 1 , 2 ) =
2
2
The two parameters, ν1 and ν2, are the numerator and denominator degrees
of freedom. That is, ν1 and ν2 are the number of independent pieces of
information used to calculate χ1 and χ2, respectively.
B-25
B Distribution Reference
Example
The most common application of the F distribution is in standard tests of
hypotheses in analysis of variance and regression.
The plot shows that the F distribution exists on the positive real numbers
and is skewed to the right.
x = 0:0.01:10;
y = fpdf(x,5,3);
plot(x,y)
See Also
“Continuous Distributions (Statistics)” on page 5-6
B-26
Gamma Distribution
Gamma Distribution
In this section...
“Definition” on page B-27
“Background” on page B-27
“Parameters” on page B-28
“Example” on page B-29
“See Also” on page B-29
Definition
The gamma pdf is
−x
1 a −1
y = f ( x | a, b) = x eb
ba Γ(a)
where Γ( · ) is the Gamma function.
Background
The gamma distribution models sums of exponentially distributed random
variables.
The gamma distribution family is based on two parameters. The chi-square
and exponential distributions, which are children of the gamma distribution,
are one-parameter distributions that fix one of the two gamma parameters.
The gamma distribution has the following relationship with the incomplete
Gamma function.
⎛x ⎞
f ( x | a, b) = gammainc ⎜ , a ⎟
⎝b ⎠
B-27
B Distribution Reference
When a is large, the gamma distribution closely approximates a normal
distribution with the advantage that the gamma distribution has density
only for positive real numbers.
Parameters
Suppose you are stress testing computer memory chips and collecting data on
their lifetimes. You assume that these lifetimes follow a gamma distribution.
You want to know how long you can expect the average computer memory chip
to last. Parameter estimation is the process of determining the parameters of
the gamma distribution that fit this data best in some sense.
One popular criterion of goodness is to maximize the likelihood function.
The likelihood has the same form as the gamma pdf above. But for the pdf,
the parameters are known constants and the variable is x. The likelihood
function reverses the roles of the variables. Here, the sample values (the x’s)
are already observed. So they are the fixed constants. The variables are the
unknown parameters. MLE involves calculating the values of the parameters
that give the highest likelihood given the particular set of data.
The function gamfit returns the MLEs and confidence intervals for the
parameters of the gamma distribution. Here is an example using random
numbers from the gamma distribution with a = 10 and b = 5.
lifetimes = gamrnd(10,5,100,1);
[phat, pci] = gamfit(lifetimes)
phat =
10.9821 4.7258
pci =
7.4001 3.1543
14.5640 6.2974
Note phat(1) = â and phat(2) = b̂ . The MLE for parameter a is 10.98,
compared to the true value of 10. The 95% confidence interval for a goes from
7.4 to 14.6, which includes the true value.
B-28
Gamma Distribution
Similarly the MLE for parameter b is 4.7, compared to the true value of 5.
The 95% confidence interval for b goes from 3.2 to 6.3, which also includes
the true value.
In the life tests you do not know the true value of a and b so it is nice to have
a confidence interval on the parameters to give a range of likely values.
Example
In the example the gamma pdf is plotted with the solid line. The normal
pdf has a dashed line type.
x = gaminv((0.005:0.01:0.995),100,10);
y = gampdf(x,100,10);
y1 = normpdf(x,1000,100);
plot(x,y,'-',x,y1,'-.')
See Also
“Continuous Distributions (Data)” on page 5-4
B-29
B Distribution Reference
Gaussian Distribution
See “Normal Distribution” on page B-83.
B-30
Gaussian Mixture Distributions
Gaussian Mixture Distributions
See the discussion of the gmdistribution class in the “Gaussian Mixture
Models” on page 5-99 section of “Probability Distributions Used for
Multivariate Modeling” on page 5-99 and the “Gaussian Mixture Models” on
page 11-28 section of Chapter 11, “Cluster Analysis”.
B-31
B Distribution Reference
Generalized Extreme Value Distribution
In this section...
“Definition” on page B-32
“Background” on page B-32
“Parameters” on page B-33
“Example” on page B-34
“See Also” on page B-36
Definition
The probability density function for the generalized extreme value distribution
with location parameter µ, scale parameter σ, and shape parameter k ≠ 0 is
⎛ − ⎞
1
−1−
1
⎛1⎞ ⎜ ⎛ ( x − μ) ⎞ k ⎟ ⎛ ( x − μ) ⎞ k
y = f ( x |k, μ, σ ) = ⎜ ⎟ exp ⎜ − ⎜ 1 + k 1+k
⎝σ⎠ ⎜ ⎝ σ ⎟⎠ ⎟⎟ ⎜⎝ σ ⎟⎠
⎝ ⎠
for
(x-μ)
1+k >0
σ
k > 0 corresponds to the Type II case, while k < 0 corresponds to the Type
III case. For k = 0, corresponding to the Type I case, the density is
⎛1⎞ ⎛ ⎛ ( x − μ) ⎞ ( x − μ) ⎞
y = f ( x |0, μ, σ ) = ⎜ ⎟ exp ⎜ − exp ⎜ − − ⎟
⎝σ⎠ ⎝ ⎝ σ ⎟⎠ σ ⎠
Background
Like the extreme value distribution, the generalized extreme value
distribution is often used to model the smallest or largest value among a
large set of independent, identically distributed random values representing
measurements or observations. For example, you might have batches of 1000
B-32
Generalized Extreme Value Distribution
washers from a manufacturing process. If you record the size of the largest
washer in each batch, the data are known as block maxima (or minima if you
record the smallest). You can use the generalized extreme value distribution
as a model for those block maxima.
The generalized extreme value combines three simpler distributions into a
single form, allowing a continuous range of possible shapes that includes all
three of the simpler distributions. You can use any one of those distributions
to model a particular dataset of block maxima. The generalized extreme
value distribution allows you to “let the data decide” which distribution is
appropriate.
The three cases covered by the generalized extreme value distribution are
often referred to as the Types I, II, and III. Each type corresponds to the
limiting distribution of block maxima from a different class of underlying
distributions. Distributions whose tails decrease exponentially, such as the
normal, lead to the Type I. Distributions whose tails decrease as a polynomial,
such as Student’s t, lead to the Type II. Distributions whose tails are finite,
such as the beta, lead to the Type III.
Types I, II, and III are sometimes also referred to as the Gumbel, Frechet,
and Weibull types, though this terminology can be slightly confusing. The
Type I (Gumbel) and Type III (Weibull) cases actually correspond to the
mirror images of the usual Gumbel and Weibull distributions, for example,
as computed by the functions evcdf and evfit , or wblcdf and wblfit,
respectively. Finally, the Type II (Frechet) case is equivalent to taking the
reciprocal of values from a standard Weibull distribution.
Parameters
If you generate 250 blocks of 1000 random values drawn from Student’s t
distribution with 5 degrees of freedom, and take their maxima, you can fit a
generalized extreme value distribution to those maxima.
blocksize = 1000;
nblocks = 250;
t = trnd(5,blocksize,nblocks);
x = max(t); % 250 column maxima
paramEsts = gevfit(x)
paramEsts =
B-33
B Distribution Reference
0.2438 1.1760 5.8045
Notice that the shape parameter estimate (the first element) is positive,
which is what you would expect based on block maxima from a Student’s t
distribution.
hist(x,2:20);
set(get(gca,'child'),'FaceColor',[.8 .8 1])
xgrid = linspace(2,20,1000);
line(xgrid,nblocks*...
gevpdf(xgrid,paramEsts(1),paramEsts(2),paramEsts(3)));
Example
The following code generates examples of probability density functions for the
three basic forms of the generalized extreme value distribution.
B-34
Generalized Extreme Value Distribution
x = linspace(-3,6,1000);
y1 = gevpdf(x,-.5,1,0);
y2 = gevpdf(x,0,1,0);
y3 = gevpdf(x,.5,1,0)
plot(x,y1,'-', x,y2,'-', x,y3,'-')
legend({'K<0, Type III' 'K=0, Type I' 'K>0, Type II'});
Notice that for k > 0, the distribution has zero probability density for x such
that
σ
x < - + μ
k
For k < 0, the distribution has zero probability density for
B-35
B Distribution Reference
σ
x > - + μ
k
For k = 0, there is no upper or lower bound.
See Also
“Continuous Distributions (Data)” on page 5-4
B-36
Generalized Pareto Distribution
Generalized Pareto Distribution
In this section...
“Definition” on page B-37
“Background” on page B-37
“Parameters” on page B-38
“Example” on page B-39
“See Also” on page B-40
Definition
The probability density function for the generalized Pareto distribution with
shape parameter k ≠ 0, scale parameter σ, and threshold parameter θ, is
1
−1 −
⎛ 1 ⎞⎛ (x − ) ⎞ k
y = f ( x| k, , ) = ⎜ ⎟ ⎜ 1 + k
⎝ ⎠⎝ ⎟⎠
for θ < x, when k > 0, or for θ < x < –σ/k when k < 0.
For k = 0, the density is
( x − )
⎛1⎞ −
y = f ( x|0, , ) = ⎜ ⎟ e
⎝ ⎠
for θ < x.
If k = 0 and θ = 0, the generalized Pareto distribution is equivalent to
the exponential distribution. If k > 0 and θ = σ/k, the generalized Pareto
distribution is equivalent to the Pareto distribution.
Background
Like the exponential distribution, the generalized Pareto distribution is often
used to model the tails of another distribution. For example, you might
have washers from a manufacturing process. If random influences in the
process lead to differences in the sizes of the washers, a standard probability
B-37
B Distribution Reference
distribution, such as the normal, could be used to model those sizes. However,
while the normal distribution might be a good model near its mode, it might
not be a good fit to real data in the tails and a more complex model might
be needed to describe the full range of the data. On the other hand, only
recording the sizes of washers larger (or smaller) than a certain threshold
means you can fit a separate model to those tail data, which are known as
exceedences. You can use the generalized Pareto distribution in this way, to
provide a good fit to extremes of complicated data.
The generalized Pareto distribution allows a continuous range of possible
shapes that includes both the exponential and Pareto distributions as special
cases. You can use either of those distributions to model a particular dataset
of exceedences. The generalized Pareto distribution allows you to “let the data
decide” which distribution is appropriate.
The generalized Pareto distribution has three basic forms, each corresponding
to a limiting distribution of exceedence data from a different class of
underlying distributions.
• Distributions whose tails decrease exponentially, such as the normal, lead
to a generalized Pareto shape parameter of zero.
• Distributions whose tails decrease as a polynomial, such as Student’s t,
lead to a positive shape parameter.
• Distributions whose tails are finite, such as the beta, lead to a negative
shape parameter.
The generalized Pareto distribution is used in the tails of distribution fit
objects of the paretotails class.
Parameters
If you generate a large number of random values from a Student’s t
distribution with 5 degrees of freedom, and then discard everything less than
2, you can fit a generalized Pareto distribution to those exceedences.
t = trnd(5,5000,1);
y = t(t > 2) - 2;
paramEsts = gpfit(y)
paramEsts =
B-38
Generalized Pareto Distribution
0.1267 0.8134
Notice that the shape parameter estimate (the first element) is positive, which
is what you would expect based on exceedences from a Student’s t distribution.
hist(y+2,2.25:.5:11.75);
set(get(gca,'child'),'FaceColor',[.8 .8 1])
xgrid = linspace(2,12,1000);
line(xgrid,.5*length(y)*...
gppdf(xgrid,paramEsts(1),paramEsts(2),2));
Example
The following code generates examples of the probability density functions for
the three basic forms of the generalized Pareto distribution.
x = linspace(0,10,1000);
B-39
B Distribution Reference
y1 = gppdf(x,-.25,1,0);
y2 = gppdf(x,0,1,0);
y3 = gppdf(x,1,1,0)
plot(x,y1,'-', x,y2,'-', x,y3,'-')
legend({'K<0' 'K=0' 'K>0'});
σ
Notice that for k < 0, the distribution has zero probability density for x > - ,
while for k ≥ 0, there is no upper bound. k
See Also
“Continuous Distributions (Data)” on page 5-4
B-40
Geometric Distribution
Geometric Distribution
In this section...
“Definition” on page B-41
“Background” on page B-41
“Example” on page B-41
“See Also” on page B-42
Definition
The geometric pdf is
y = f ( x | p) = pq x I(0,1,...) ( x)
where q = 1 – p. The geometric distribution is a special case of the negative
binomial distribution, with r = 1.
Background
The geometric distribution is discrete, existing only on the nonnegative
integers. It is useful for modeling the runs of consecutive successes (or
failures) in repeated independent trials of a system.
The geometric distribution models the number of successes before one failure
in an independent succession of tests where each test results in success or
failure.
Example
Suppose the probability of a five-year-old battery failing in cold weather is
0.03. What is the probability of starting 25 consecutive days during a long
cold snap?
1 - geocdf(25,0.03)
ans =
B-41
B Distribution Reference
0.4530
The plot shows the cdf for this scenario.
x = 0:25;
y = geocdf(x,0.03);
stairs(x,y)
See Also
“Discrete Distributions” on page 5-7
B-42
Hypergeometric Distribution
Hypergeometric Distribution
In this section...
“Definition” on page B-43
“Background” on page B-43
“Example” on page B-44
“See Also” on page B-44
Definition
The hypergeometric pdf is
⎛ K ⎞⎛ M − K ⎞
⎜ ⎟⎜ ⎟
x n− x ⎠
y = f ( x | M , K , n) = ⎝ ⎠ ⎝
⎛M⎞
⎜ ⎟
⎝n⎠
Background
The hypergeometric distribution models the total number of successes in a
fixed-size sample drawn without replacement from a finite population.
The distribution is discrete, existing only for nonnegative integers less than
the number of samples or the number of possible successes, whichever is
greater. The hypergeometric distribution differs from the binomial only in
that the population is finite and the sampling from the population is without
replacement.
The hypergeometric distribution has three parameters that have direct
physical interpretations.
• M is the size of the population.
• K is the number of items with the desired characteristic in the population.
• n is the number of samples drawn.
B-43
B Distribution Reference
Sampling “without replacement” means that once a particular sample
is chosen, it is removed from the relevant population for all subsequent
selections.
Example
The plot shows the cdf of an experiment taking 20 samples from a group of
1000 where there are 50 items of the desired type.
x = 0:10;
y = hygecdf(x,1000,50,20);
stairs(x,y)
See Also
“Discrete Distributions” on page 5-7
B-44
Inverse Gaussian Distribution
Inverse Gaussian Distribution
In this section...
“Definition” on page B-45
“Background” on page B-45
“Parameters” on page B-45
“See Also” on page B-45
Definition
The inverse Gaussian distribution has the density function
⎪⎧ ⎫
exp ⎨−
3 2
( x − )2 ⎪⎬
2 x ⎪⎩ 2 x ⎪⎭
Background
Also known as the Wald distribution, the inverse Gaussian is used to model
nonnegative positively skewed data. The distribution originated in the theory
of Brownian motion, but has been used to model diverse phenomena. Inverse
Gaussian distributions have many similarities to standard Gaussian (normal)
distributions, which lead to applications in inferential statistics.
Parameters
See mle, dfittool.
See Also
“Continuous Distributions (Data)” on page 5-4
B-45
B Distribution Reference
Inverse Wishart Distribution
Definition
The probability density function of the d-dimensional Inverse Wishart
distribution is given by
⎛ 1 ⎞
ν /2 ⎜ ( )
- trace ΤX −1 ⎟
Τ ( ) e⎝ 2 ⎠
y = f(Χ, Σ, ν) =
(νd)/2 (d(d-1))/4 (ν + d +1)/ 2
where X and T are2d-by-d π symmetric X positiveΓ (definite
ν / 2 ) ...Γmatrices,
(ν-(d-1))/2and ν is a
scalar greater than or equal to d. While it is possible to define the Inverse
Wishart for singular Τ, the density cannot be written as above.
If a random matrix has a Wishart distribution with parameters T-1 and ν, then
the inverse of that random matrix has an inverse Wishart distribution with
parameters Τ and ν. The mean of the distribution is given by
1
Τ
ν − d −1
where d is the number of rows and columns in T.
Only random matrix generation is supported for the inverse Wishart,
including both singular and nonsingular T.
Background
The inverse Wishart distribution is based on the Wishart distribution. In
Bayesian statistics it is used as the conjugate prior for the covariance matrix
of a multivariate normal distribution.
Example
Notice that the sampling variability is quite large when the degrees of
freedom is small.
Tau = [1 .5; .5 2];
df = 10; S1 = iwishrnd(Tau,df)*(df-2-1)
B-46
Inverse Wishart Distribution
S1 =
1.7959 0.64107
0.64107 1.5496
df = 1000; S2 = iwishrnd(Tau,df)*(df-2-1)
S2 =
0.9842 0.50158
0.50158 2.1682
See Also
“Multivariate Distributions” on page 5-8
B-47
B Distribution Reference
Johnson System
See “Pearson and Johnson Systems” on page 6-27.
B-48
Logistic Distribution
Logistic Distribution
In this section...
“Definition” on page B-49
“Background” on page B-49
“Parameters” on page B-49
“See Also” on page B-49
Definition
The logistic distribution has the density function
x−
e
2
⎛ x− ⎞
⎜
1+ e ⎟
⎜ ⎟
⎝ ⎠
with location parameter µ and scale parameter σ > 0, for all real x.
Background
The logistic distribution originated with Verhulst’s work on demography in
the early 1800s. The distribution has been used for various growth models,
and is used in logistic regression. It has longer tails and a higher kurtosis
than the normal distribution.
Parameters
See mle, dfittool.
See Also
“Continuous Distributions (Data)” on page 5-4
B-49
B Distribution Reference
Loglogistic Distribution
In this section...
“Definition” on page B-50
“Parameters” on page B-50
“See Also” on page B-50
Definition
The variable x has a loglogistic distribution with location parameter µ and
scale parameter σ > 0 if ln x has a logistic distribution with parameters µ
and σ. The relationship is similar to that between the lognormal and normal
distribution.
Parameters
See mle, dfittool.
See Also
“Continuous Distributions (Data)” on page 5-4
B-50
Lognormal Distribution
Lognormal Distribution
In this section...
“Definition” on page B-51
“Background” on page B-51
“Example” on page B-52
“See Also” on page B-53
Definition
The lognormal pdf is
−( ln x − )
2
1 2 2
y = f ( x| , ) = e
x 2
Background
The normal and lognormal distributions are closely related. If X is distributed
lognormally with parameters µ and σ, then log(X) is distributed normally
with mean µ and standard deviation σ.
The mean m and variance v of a lognormal random variable are functions of µ
and σ that can be calculated with the lognstat function. They are:
(
m = exp + 2 / 2 )
( )( ( ) )
v = exp 2 + 2 exp 2 − 1
A lognormal distribution with mean m and variance v has parameters
= log ⎛⎜ m2 / v + m2 ⎞⎟
⎝ ⎠
(
= log v / m2 + 1 )
B-51
B Distribution Reference
The lognormal distribution is applicable when the quantity of interest must
be positive, since log(X) exists only when X is positive.
Example
Suppose the income of a family of four in the United States follows a lognormal
distribution with µ = log(20,000) and σ2 = 1.0. Plot the income density.
x = (10:1000:125010)';
y = lognpdf(x,log(20000),1.0);
plot(x,y)
set(gca,'xtick',[0 30000 60000 90000 120000])
set(gca,'xticklabel',{'0','$30,000','$60,000',...
'$90,000','$120,000'})
B-52
Lognormal Distribution
See Also
“Continuous Distributions (Data)” on page 5-4
B-53
B Distribution Reference
Multinomial Distribution
In this section...
“Definition” on page B-54
“Background” on page B-54
“Example” on page B-54
Definition
The multinomial pdf is
n!
f ( x | n, p) =
x x
p 1 pk k
x1 ! xk ! 1
where x = (x1,..., xk) gives the number of each of k outcomes in n trials of a
process with fixed probabilities p = (p1,...,pk) of individual outcomes in any one
trial. The vector x has non-negative integer components that sum to n. The
vector p has non-negative integer components that sum to 1.
Background
The multinomial distribution is a generalization of the binomial distribution.
The binomial distribution gives the probability of the number of “successes”
and “failures” in n independent trials of a two-outcome process. The
probability of “success” and “failure” in any one trial is given by the fixed
probabilities p and q = 1–p. The multinomial distribution gives the probability
of each combination of outcomes in n independent trials of a k-outcome
process. The probability of each outcome in any one trial is given by the fixed
probabilities p1,...,pk.
The expected value of outcome i is npi. The variance of outcome i is npi(1 – pi).
The covariance of outcomes i and j is –npipj for distinct i and j.
Example
The following uses mnpdf to produce a visualization of a trinomial distribution:
B-54
Multinomial Distribution
% Compute the distribution
p = [1/2 1/3 1/6]; % Outcome probabilities
n = 10; % Sample size
x1 = 0:n;
x2 = 0:n;
[X1,X2] = meshgrid(x1,x2);
X3 = n-(X1+X2);
Y = mnpdf([X1(:),X2(:),X3(:)],repmat(p,(n+1)^2,1));
% Plot the distribution
Y = reshape(Y,n+1,n+1);
bar3(Y)
set(gca,'XTickLabel',0:n)
set(gca,'YTickLabel',0:n)
xlabel('x_1')
ylabel('x_2')
zlabel('Probability Mass')
B-55
B Distribution Reference
Note that the visualization does not show x3, which is determined by the
constraint x1 + x2 + x3 = n.
B-56
Multivariate Gaussian Distribution
Multivariate Gaussian Distribution
See “Multivariate Normal Distribution” on page B-58.
B-57
B Distribution Reference
Multivariate Normal Distribution
In this section...
“Definition” on page B-58
“Background” on page B-58
“Example” on page B-59
“See Also” on page B-63
Definition
The probability density function of the d-dimensional multivariate normal
distribution is given by
1
− (x- ) Σ-1 (x- )′
1 2
y = f ( x, , Σ) = e
Σ (2 ) d
where x and μ are 1-by-d vectors and Σ is a d-by-d symmetric positive definite
matrix. While it is possible to define the multivariate normal for singular Σ,
the density cannot be written as above. Only random vector generation is
supported for the singular case. Note that while most textbooks define the
multivariate normal with x and μ oriented as column vectors, for the purposes
of data analysis software, it is more convenient to orient them as row vectors,
and Statistics Toolbox software uses that orientation.
Background
The multivariate normal distribution is a generalization of the univariate
normal to two or more variables. It is a distribution for random vectors
of correlated variables, each element of which has a univariate normal
distribution. In the simplest case, there is no correlation among variables, and
elements of the vectors are independent univariate normal random variables.
The multivariate normal distribution is parameterized with a mean vector, μ,
and a covariance matrix, Σ. These are analogous to the mean μ and variance
σ2 parameters of a univariate normal distribution. The diagonal elements of Σ
B-58
Multivariate Normal Distribution
contain the variances for each variable, while the off-diagonal elements of Σ
contain the covariances between variables.
The multivariate normal distribution is often used as a model for multivariate
data, primarily because it is one of the few multivariate distributions that is
tractable to work with.
Example
This example shows the probability density function (pdf) and cumulative
distribution function (cdf) for a bivariate normal distribution with unequal
standard deviations. You can use the multivariate normal distribution in a
higher number of dimensions as well, although visualization is not easy.
mu = [0 0];
Sigma = [.25 .3; .3 1];
x1 = -3:.2:3; x2 = -3:.2:3;
[X1,X2] = meshgrid(x1,x2);
F = mvnpdf([X1(:) X2(:)],mu,Sigma);
F = reshape(F,length(x2),length(x1));
surf(x1,x2,F);
caxis([min(F(:))-.5*range(F(:)),max(F(:))]);
axis([-3 3 -3 3 0 .4])
xlabel('x1'); ylabel('x2'); zlabel('Probability Density');
B-59
B Distribution Reference
F = mvncdf([X1(:) X2(:)],mu,Sigma);
F = reshape(F,length(x2),length(x1));
surf(x1,x2,F);
caxis([min(F(:))-.5*range(F(:)),max(F(:))]);
axis([-3 3 -3 3 0 1])
xlabel('x1'); ylabel('x2'); zlabel('Cumulative Probability');
B-60
Multivariate Normal Distribution
Since the bivariate normal distribution is defined on the plane, you can also
compute cumulative probabilities over rectangular regions. For example,
this contour plot illustrates the computation that follows, of the probability
contained within the unit square.
contour(x1,x2,F,[.0001 .001 .01 .05:.1:.95 .99 .999 .9999]);
xlabel('x'); ylabel('y');
line([0 0 1 1 0],[1 0 0 1 1],'linestyle','--','color','k');
B-61
B Distribution Reference
mvncdf([0 0],[1 1],mu,Sigma)
ans =
0.20974
Computing a multivariate cumulative probability requires significantly
more work than computing a univariate probability. By default, the mvncdf
function computes values to less than full machine precision, and returns an
estimate of the error as an optional second output:
[F,err] = mvncdf([0 0],[1 1],mu,Sigma)
F =
0.20974
err =
1e-008
B-62
Multivariate Normal Distribution
See Also
“Multivariate Distributions” on page 5-8
B-63
B Distribution Reference
Multivariate t Distribution
In this section...
“Definition” on page B-64
“Background” on page B-64
“Example” on page B-65
“See Also” on page B-69
Definition
The probability density function of the d-dimensional multivariate Student’s t
distribution is given by
-(ν+d)/2
1 1 Γ((ν+d)/2) ⎛ x ′ Ρ-1 x ⎞
y = f(x, Ρ, ν) = ⎜1 + ⎟
⎜ ν ⎟⎠
(νπ)d Γ(ν /2) ⎝
1/2
Σ
where x is a 1-by-d vector, P is a d-by-d symmetric, positive definite matrix,
and ν is a positive scalar. While it is possible to define the multivariate
Student’s t for singular P, the density cannot be written as above. For the
singular case, only random number generation is supported. Note that while
most textbooks define the multivariate Student’s t with x oriented as a column
vector, for the purposes of data analysis software, it is more convenient to
orient x as a row vector, and Statistics Toolbox software uses that orientation.
Background
The multivariate Student’s t distribution is a generalization of the univariate
Student’s t to two or more variables. It is a distribution for random vectors
of correlated variables, each element of which has a univariate Student’s t
distribution. In the same way as the univariate Student’s t distribution can
be constructed by dividing a standard univariate normal random variable by
the square root of a univariate chi-square random variable, the multivariate
Student’s t distribution can be constructed by dividing a multivariate
normal random vector having zero mean and unit variances by a univariate
chi-square random variable.
B-64
Multivariate t Distribution
The multivariate Student’s t distribution is parameterized with a correlation
matrix, P, and a positive scalar degrees of freedom parameter, ν. ν is
analogous to the degrees of freedom parameter of a univariate Student’s t
distribution. The off-diagonal elements of P contain the correlations between
variables. Note that when P is the identity matrix, variables are uncorrelated;
however, they are not independent.
The multivariate Student’s t distribution is often used as a substitute for
the multivariate normal distribution in situations where it is known that
the marginal distributions of the individual variables have fatter tails than
the normal.
Example
This example shows the probability density function (pdf) and cumulative
distribution function (cdf) for a bivariate Student’s t distribution. You can use
the multivariate Student’s t distribution in a higher number of dimensions as
well, although visualization is not easy.
Rho = [1 .6; .6 1];
nu = 5;
x1 = -3:.2:3; x2 = -3:.2:3;
[X1,X2] = meshgrid(x1,x2);
F = mvtpdf([X1(:) X2(:)],Rho,nu);
F = reshape(F,length(x2),length(x1));
surf(x1,x2,F);
caxis([min(F(:))-.5*range(F(:)),max(F(:))]);
axis([-3 3 -3 3 0 .2])
xlabel('x1'); ylabel('x2'); zlabel('Probability Density');
B-65
B Distribution Reference
F = mvtcdf([X1(:) X2(:)],Rho,nu);
F = reshape(F,length(x2),length(x1));
surf(x1,x2,F);
caxis([min(F(:))-.5*range(F(:)),max(F(:))]);
axis([-3 3 -3 3 0 1])
xlabel('x1'); ylabel('x2'); zlabel('Cumulative Probability');
B-66
Multivariate t Distribution
Since the bivariate Student’s t distribution is defined on the plane, you can
also compute cumulative probabilities over rectangular regions. For example,
this contour plot illustrates the computation that follows, of the probability
contained within the unit square.
contour(x1,x2,F,[.0001 .001 .01 .05:.1:.95 .99 .999 .9999]);
xlabel('x'); ylabel('y');
line([0 0 1 1 0],[1 0 0 1 1],'linestyle','--','color','k');
B-67
B Distribution Reference
mvtcdf([0 0],[1 1],Rho,nu)
ans =
0.14013
Computing a multivariate cumulative probability requires significantly
more work than computing a univariate probability. By default, the mvtcdf
function computes values to less than full machine precision, and returns an
estimate of the error as an optional second output:
[F,err] = mvtcdf([0 0],[1 1],Rho,nu)
F =
0.14013
err =
1e-008
B-68
Multivariate t Distribution
See Also
“Multivariate Distributions” on page 5-8
B-69
B Distribution Reference
Nakagami Distribution
In this section...
“Definition” on page B-70
“Background” on page B-70
“Parameters” on page B-70
“See Also” on page B-71
Definition
The Nakagami distribution has the density function
−
⎛⎞ x2
x(
1 2 −1)
2⎜ ⎟ e
⎝ ⎠ Γ ( )
with shape parameter µ and scale parameter ω > 0, for x > 0. If x has a
Nakagami distribution with parameters µ and ω, then x2 has a gamma
distribution with shape parameter µ and scale parameter ω/µ.
Background
In communications theory, Nakagami distributions, Rician distributions,
and Rayleigh distributions are used to model scattered signals that reach
a receiver by multiple paths. Depending on the density of the scatter, the
signal will display different fading characteristics. Rayleigh and Nakagami
distributions are used to model dense scatters, while Rician distributions
model fading with a stronger line-of-sight. Nakagami distributions can be
reduced to Rayleigh distributions, but give more control over the extent
of the fading.
Parameters
See mle, dfittool.
B-70
Nakagami Distribution
See Also
“Continuous Distributions (Data)” on page 5-4
B-71
B Distribution Reference
Negative Binomial Distribution
In this section...
“Definition” on page B-72
“Background” on page B-72
“Parameters” on page B-73
“Example” on page B-75
“See Also” on page B-75
Definition
When the r parameter is an integer, the negative binomial pdf is
⎛ r + x − 1⎞ r x
y = f ( x | r, p) = ⎜ ⎟ p q I(0,1,...) ( x)
⎝ x ⎠
where q = 1 – p. When r is not an integer, the binomial coefficient in the
definition of the pdf is replaced by the equivalent expression
Γ(r + x)
Γ(r)Γ( x + 1)
Background
In its simplest form (when r is an integer), the negative binomial distribution
models the number of failures x before a specified number of successes is
reached in a series of independent, identical trials. Its parameters are the
probability of success in a single trial, p, and the number of successes, r. A
special case of the negative binomial distribution, when r = 1, is the geometric
distribution, which models the number of failures before the first success.
More generally, r can take on non-integer values. This form of the negative
binomial distribution has no interpretation in terms of repeated trials, but,
like the Poisson distribution, it is useful in modeling count data. The negative
binomial distribution is more general than the Poisson distribution because it
has a variance that is greater than its mean, making it suitable for count data
B-72
Negative Binomial Distribution
that do not meet the assumptions of the Poisson distribution. In the limit,
as r increases to infinity, the negative binomial distribution approaches the
Poisson distribution.
Parameters
Suppose you are collecting data on the number of auto accidents on a busy
highway, and would like to be able to model the number of accidents per day.
Because these are count data, and because there are a very large number of
cars and a small probability of an accident for any specific car, you might
think to use the Poisson distribution. However, the probability of having an
accident is likely to vary from day to day as the weather and amount of traffic
change, and so the assumptions needed for the Poisson distribution are not
met. In particular, the variance of this type of count data sometimes exceeds
the mean by a large amount. The data below exhibit this effect: most days
have few or no accidents, and a few days have a large number.
accident = [2 3 4 2 3 1 12 8 14 31 23 1 10 7 0];
mean(accident)
ans =
8.0667
var(accident)
ans =
79.352
The negative binomial distribution is more general than the Poisson, and is
often suitable for count data when the Poisson is not. The function nbinfit
returns the maximum likelihood estimates (MLEs) and confidence intervals
for the parameters of the negative binomial distribution. Here are the results
from fitting the accident data:
[phat,pci] = nbinfit(accident)
phat =
1.0060 0.1109
pci =
0.2152 0.0171
1.7968 0.2046
B-73
B Distribution Reference
It is difficult to give a physical interpretation in this case to the individual
parameters. However, the estimated parameters can be used in a model
for the number of daily accidents. For example, a plot of the estimated
cumulative probability function shows that while there is an estimated 10%
chance of no accidents on a given day, there is also about a 10% chance that
there will be 20 or more accidents.
plot(0:50,nbincdf(0:50,phat(1),phat(2)),'.-');
xlabel('Accidents per Day')
ylabel('Cumulative Probability')
B-74
Negative Binomial Distribution
Example
The negative binomial distribution can take on a variety of shapes ranging
from very skewed to nearly symmetric. This example plots the probability
function for different values of r, the desired number of successes: .1, 1, 3, 6.
x = 0:10;
plot(x,nbinpdf(x,.1,.5),'s-', ...
x,nbinpdf(x,1,.5),'o-', ...
x,nbinpdf(x,3,.5),'d-', ...
x,nbinpdf(x,6,.5),'^-');
legend({'r = .1' 'r = 1' 'r = 3' 'r = 6'})
xlabel('x')
ylabel('f(x|r,p)')
See Also
“Discrete Distributions” on page 5-7
B-75
B Distribution Reference
Noncentral Chi-Square Distribution
In this section...
“Definition” on page B-76
“Background” on page B-76
“Example” on page B-77
Definition
There are many equivalent formulas for the noncentral chi-square distribution
function. One formulation uses a modified Bessel function of the first
kind. Another uses the generalized Laguerre polynomials. The cumulative
distribution function is computed using a weighted sum of χ2 probabilities
with the weights equal to the probabilities of a Poisson distribution.
The Poisson parameter is one-half of the noncentrality parameter of the
noncentral chi-square
⎛⎛1 ⎞j ⎞
∞ ⎜ ⎜ ⎟ − ⎟
F ( x | , ) = ∑ ⎜⎜ ⎝ ⎟ Pr ⎡ 2
2 ⎠ ⎤
j !
e2 ⎟ ⎣ +2 j ≤ x ⎦
j =0 ⎜ ⎟
⎜ ⎟
⎝ ⎠
where δ is the noncentrality parameter.
Background
The χ2 distribution is actually a simple special case of the noncentral
chi-square distribution. One way to generate random numbers with a χ2
distribution (with ν degrees of freedom) is to sum the squares of ν standard
normal random numbers (mean equal to zero.)
What if the normally distributed quantities have a mean other than zero? The
sum of squares of these numbers yields the noncentral chi-square distribution.
The noncentral chi-square distribution requires two parameters: the degrees
of freedom and the noncentrality parameter. The noncentrality parameter is
the sum of the squared means of the normally distributed quantities.
B-76
Noncentral Chi-Square Distribution
The noncentral chi-square has scientific application in thermodynamics and
signal processing. The literature in these areas may refer to it as the “Rician
Distribution” on page B-93 or generalized “Rayleigh Distribution” on page
B-91.
Example
The following commands generate a plot of the noncentral chi-square pdf.
x = (0:0.1:10)';
p1 = ncx2pdf(x,4,2);
p = chi2pdf(x,4);
plot(x,p,'-',x,p1,'-')
B-77
B Distribution Reference
Noncentral F Distribution
In this section...
“Definition” on page B-78
“Background” on page B-78
“Example” on page B-79
“See Also” on page B-79
Definition
Similar to the noncentral χ2 distribution, the toolbox calculates noncentral
F distribution probabilities as a weighted sum of incomplete beta functions
using Poisson probabilities as the weights.
⎛⎛1 ⎞j ⎞
∞ ⎜⎜ ⎟ − ⎟
F ( x | 1 , 2 , ) = ∑ ⎜⎜ ⎝
2 ⎠ ⎟I ⎛ 1 ⋅ x 1 + j, 2 ⎞
j!
e2 ⎟ ⎜⎜ + ⋅ x 2 ⎟
2 ⎟⎠
j =0 ⎜ ⎟ ⎝ 2 1
⎜ ⎟
⎝ ⎠
I(x|a,b) is the incomplete beta function with parameters a and b, and δ is
the noncentrality parameter.
Background
As with the χ2 distribution, the F distribution is a special case of the
noncentral F distribution. The F distribution is the result of taking the ratio
of χ2 random variables each divided by its degrees of freedom.
If the numerator of the ratio is a noncentral chi-square random variable
divided by its degrees of freedom, the resulting distribution is the noncentral
F distribution.
The main application of the noncentral F distribution is to calculate the power
of a hypothesis test relative to a particular alternative.
B-78
Noncentral F Distribution
Example
The following commands generate a plot of the noncentral F pdf.
x = (0.01:0.1:10.01)';
p1 = ncfpdf(x,5,20,10);
p = fpdf(x,5,20);
plot(x,p,'-',x,p1,'-')
See Also
“Continuous Distributions (Statistics)” on page 5-6
B-79
B Distribution Reference
Noncentral t Distribution
In this section...
“Definition” on page B-80
“Background” on page B-80
“Example” on page B-81
“See Also” on page B-81
Definition
The most general representation of the noncentral t distribution is quite
complicated. Johnson and Kotz [60] give a formula for the probability that a
noncentral t variate falls in the range [–t, t].
⎛⎛1 ⎞j ⎞
∞ ⎜⎜ ⎟ − ⎟ ⎛ 2
⎞
P r ( ( −t ) < x < t |( , ) ) = ∑ ⎜⎜ ⎝ ⎟I ⎜ x
2 ⎠ 1
e2 + j , ⎟
⎟ ⎜ 2 2 2⎟
j =0 ⎜
j!
⎟ ⎝ + x ⎠
⎜ ⎟
⎝ ⎠
I(x|a,b) is the incomplete beta function with parameters a and b, δ is the
noncentrality parameter, and ν is the number of degrees of freedom.
Background
The noncentral t distribution is a generalization of Student’s t distribution.
Student’s t distribution with n – 1 degrees of freedom models the t-statistic
x−
t=
s/ n
where x is the sample mean and s is the sample standard deviation of a
random sample of size n from a normal population with mean μ. If the
population mean is actually μ0, then the t-statistic has a noncentral t
distribution with noncentrality parameter
B-80
Noncentral t Distribution
0 −
=
/ n
The noncentrality parameter is the normalized difference between μ0 and μ.
The noncentral t distribution gives the probability that a t test will correctly
reject a false null hypothesis of mean μ when the population mean is actually
μ0; that is, it gives the power of the t test. The power increases as the
difference μ0 – μ increases, and also as the sample size n increases.
Example
The following commands generate a plot of the noncentral t pdf.
x = (-5:0.1:5)';
p1 = nctcdf(x,10,1);
p = tcdf(x,10);
plot(x,p,'-',x,p1,'-')
See Also
“Continuous Distributions (Statistics)” on page 5-6
B-81
B Distribution Reference
Nonparametric Distributions
See the discussion of ksdensity in “Estimating PDFs without Parameters”
on page 5-55.
B-82
Normal Distribution
Normal Distribution
In this section...
“Definition” on page B-83
“Background” on page B-83
“Parameters” on page B-84
“Example” on page B-85
“See Also” on page B-85
Definition
The normal pdf is
−( x − )2
1
y = f ( x| , ) = e 2
2
2
Background
The normal distribution is a two-parameter family of curves. The first
parameter, µ, is the mean. The second, σ, is the standard deviation. The
standard normal distribution (written Φ(x)) sets µ to 0 and σ to 1.
Φ(x) is functionally related to the error function, erf.
(
erf ( x ) = 2Φ x 2 − 1 )
The first use of the normal distribution was as a continuous approximation
to the binomial.
The usual justification for using the normal distribution for modeling is the
Central Limit Theorem, which states (roughly) that the sum of independent
samples from any distribution with finite mean and variance converges to the
normal distribution as the sample size goes to infinity.
B-83
B Distribution Reference
Parameters
To use statistical parameters such as mean and standard deviation reliably,
you need to have a good estimator for them. The maximum likelihood
estimates (MLEs) provide one such estimator. However, an MLE might be
biased, which means that its expected value of the parameter might not
equal the parameter being estimated. For example, an MLE is biased for
estimating the variance of a normal distribution. An unbiased estimator
that is commonly used to estimate the parameters of the normal distribution
is the minimum variance unbiased estimator (MVUE). The MVUE has the
minimum variance of all unbiased estimators of a parameter.
The MVUEs of parameters µ and σ2 for the normal distribution are the sample
mean and variance. The sample mean is also the MLE for µ. The following
are two common formulas for the variance.
1 n
∑ ( xi − x )
2
s2 =
n i=1 (B-1)
1 n
s2 = ∑
n − 1 i=1
( xi − x )2
(B-2)
where
n
xi
x=∑
i=1
n
Equation 1 is the maximum likelihood estimator for σ2, and equation 2 is
the MVUE.
As an example, suppose you want to estimate the mean, µ, and the variance,
σ2, of the heights of all fourth grade children in the United States. The
function normfit returns the MVUE for µ, the square root of the MVUE for
σ2, and confidence intervals for µ and σ2. Here is a playful example modeling
the heights in inches of a randomly chosen fourth grade class.
height = normrnd(50,2,30,1); % Simulate heights.
[mu,s,muci,sci] = normfit(height)
B-84
Normal Distribution
mu =
50.2025
s =
1.7946
muci =
49.5210
50.8841
sci =
1.4292
2.4125
Note that s^2 is the MVUE of the variance.
s^2
ans =
3.2206
Example
The plot shows the bell curve of the standard normal pdf, with µ = 0 and σ = 1.
See Also
“Continuous Distributions (Data)” on page 5-4
B-85
B Distribution Reference
Pareto Distribution
See “Generalized Pareto Distribution” on page B-37.
B-86
Pearson System
Pearson System
See “Pearson and Johnson Systems” on page 6-27.
B-87
B Distribution Reference
Piecewise Distributions
See the discussion of the @piecewisedistribution class in “Fitting Piecewise
Distributions” on page 5-72.
B-88
Poisson Distribution
Poisson Distribution
In this section...
“Definition” on page B-89
“Background” on page B-89
“Parameters” on page B-90
“Example” on page B-90
“See Also” on page B-90
Definition
The Poisson pdf is
x −
y = f ( x|) = e I(0,1,...) ( x)
x!
Background
The Poisson distribution is appropriate for applications that involve counting
the number of times a random event occurs in a given amount of time,
distance, area, etc. Sample applications that involve Poisson distributions
include the number of Geiger counter clicks per second, the number of people
walking into a store in an hour, and the number of flaws per 1000 feet of
video tape.
The Poisson distribution is a one-parameter discrete distribution that takes
nonnegative integer values. The parameter, λ, is both the mean and the
variance of the distribution. Thus, as the size of the numbers in a particular
sample of Poisson random numbers gets larger, so does the variability of
the numbers.
The Poisson distribution is the limiting case of a binomial distribution where
N approaches infinity and p goes to zero while Np = λ.
B-89
B Distribution Reference
The Poisson and exponential distributions are related. If the number of
counts follows the Poisson distribution, then the interval between individual
counts follows the exponential distribution.
Parameters
The MLE and the MVUE of the Poisson parameter, λ, is the sample mean.
The sum of independent Poisson random variables is also Poisson distributed
with the parameter equal to the sum of the individual parameters. This
is used to calculate confidence intervals λ. As λ gets large the Poisson
distribution can be approximated by a normal distribution with µ = λ and σ2
= λ. This approximation is used to calculate confidence intervals for values
of λ greater than 100.
Example
The plot shows the probability for each nonnegative integer when λ = 5.
x = 0:15;
y = poisspdf(x,5);
plot(x,y,'+')
See Also
“Discrete Distributions” on page 5-7
B-90
Rayleigh Distribution
Rayleigh Distribution
In this section...
“Definition” on page B-91
“Background” on page B-91
“Parameters” on page B-92
“Example” on page B-92
“See Also” on page B-92
Definition
The Rayleigh pdf is
⎛ − x2 ⎞
x ⎜ 2⎟
2b ⎠
y = f ( x | b) = e⎝
b2
Background
The Rayleigh distribution is a special case of the Weibull distribution. If
A and B are the parameters of the Weibull distribution, then the Rayleigh
distribution with parameter b is equivalent to the Weibull distribution with
parameters A = 2b and B = 2.
If the component velocities of a particle in the x and y directions are two
independent normal random variables with zero means and equal variances,
then the distance the particle travels per unit time is distributed Rayleigh.
In communications theory, Nakagami distributions, Rician distributions,
and Rayleigh distributions are used to model scattered signals that reach
a receiver by multiple paths. Depending on the density of the scatter, the
signal will display different fading characteristics. Rayleigh and Nakagami
distributions are used to model dense scatters, while Rician distributions
model fading with a stronger line-of-sight. Nakagami distributions can be
reduced to Rayleigh distributions, but give more control over the extent
of the fading.
B-91
B Distribution Reference
Parameters
The raylfit function returns the MLE of the Rayleigh parameter. This
estimate is
1 n 2
b= ∑ xi
2n i=1
Example
The following commands generate a plot of the Rayleigh pdf.
x = [0:0.01:2];
p = raylpdf(x,0.5);
plot(x,p)
See Also
“Continuous Distributions (Data)” on page 5-4
B-92
Rician Distribution
Rician Distribution
In this section...
“Definition” on page B-93
“Background” on page B-93
“Parameters” on page B-93
“See Also” on page B-94
Definition
The Rician distribution has the density function
⎛ x 2 + s2 ⎞
−⎜ ⎟
⎛ xs ⎞ x 2 2 ⎠
I0 ⎜ ⎟ e ⎝
⎝2 ⎠2
with noncentrality parameter s ≥ 0 and scale parameter σ > 0, for x > 0. I0
is the zero-order modified Bessel function of the first kind. If x has a Rician
distribution with parameters s and σ, then (x/σ)2 has a noncentral chi-square
distribution with two degrees of freedom and noncentrality parameter (s/σ)2.
Background
In communications theory, Nakagami distributions, Rician distributions,
and Rayleigh distributions are used to model scattered signals that reach
a receiver by multiple paths. Depending on the density of the scatter, the
signal will display different fading characteristics. Rayleigh and Nakagami
distributions are used to model dense scatters, while Rician distributions
model fading with a stronger line-of-sight. Nakagami distributions can be
reduced to Rayleigh distributions, but give more control over the extent
of the fading.
Parameters
See mle, dfittool.
B-93
B Distribution Reference
See Also
“Continuous Distributions (Data)” on page 5-4
B-94
Student’s t Distribution
Student’s t Distribution
In this section...
“Definition” on page B-95
“Background” on page B-95
“Example” on page B-96
“See Also” on page B-96
Definition
Student’s t pdf is
⎛ + 1 ⎞
Γ⎜
2 ⎟⎠ 1
y = f ( x | ) = ⎝
1
⎛ ⎞ +1
Γ⎜ ⎟ ⎛ x 2⎞ 2
⎝2⎠ ⎜1 + ⎟
⎜ ⎟⎠
⎝
where Γ( · ) is the Gamma function.
Background
The t distribution is a family of curves depending on a single parameter ν (the
degrees of freedom). As ν goes to infinity, the t distribution approaches the
standard normal distribution.
W. S. Gossett discovered the distribution through his work at the Guinness
brewery. At the time, Guinness did not allow its staff to publish, so Gossett
used the pseudonym “Student.”
If x is a random sample of size n from a normal distribution with mean μ,
then the statistic
x−
t=
s/ n
B-95
B Distribution Reference
where x is the sample mean and s is the sample standard deviation, has
Student’s t distribution with n – 1 degrees of freedom.
Example
The plot compares the t distribution with ν = 5 (solid line) to the shorter
tailed, standard normal distribution (dashed line).
x = -5:0.1:5;
y = tpdf(x,5);
z = normpdf(x,0,1);
plot(x,y,'-',x,z,'-.')
See Also
“Continuous Distributions (Statistics)” on page 5-6
B-96
t Location-Scale Distribution
t Location-Scale Distribution
In this section...
“Definition” on page B-97
“Background” on page B-97
“Parameters” on page B-97
“See Also” on page B-98
Definition
The t location-scale distribution has the density function
⎛ +1 ⎞
−⎜ ⎟
⎛ + 1 ⎞ ⎡ ⎛x−⎞ ⎤
2 ⎝ 2 ⎠
Γ⎜ ⎟ ⎢ + ⎜ ⎟ ⎥
⎝ 2 ⎠ ⎢ ⎝ ⎠ ⎥
⎛ ⎞ ⎢ ⎥
Γ ⎜ ⎟ ⎢ ⎥
⎝ 2 ⎠ ⎢⎣ ⎥⎦
with location parameter µ, scale parameter σ > 0, and shape parameter ν > 0.
If x has a t location-scale distribution, with parameters µ, σ, and ν, then
x−
has a Student’s t distribution with ν degrees of freedom.
Background
The t location-scale distribution is useful for modeling data distributions
with heavier tails (more prone to outliers) than the normal distribution. It
approaches the normal distribution as ν approaches infinity, and smaller
values of ν yield heavier tails.
Parameters
See mle, dfittool.
B-97
B Distribution Reference
See Also
“Continuous Distributions (Statistics)” on page 5-6
B-98
Uniform Distribution (Continuous)
Uniform Distribution (Continuous)
In this section...
“Definition” on page B-99
“Background” on page B-99
“Parameters” on page B-99
“Example” on page B-99
“See Also” on page B-100
Definition
The uniform cdf is
x−a
p = F ( x | a, b) = I ( x)
b − a [ a,b]
Background
The uniform distribution (also called rectangular) has a constant pdf between
its two parameters a (the minimum) and b (the maximum). The standard
uniform distribution (a = 0 and b = 1) is a special case of the beta distribution,
obtained by setting both of its parameters to 1.
The uniform distribution is appropriate for representing the distribution of
round-off errors in values tabulated to a particular number of decimal places.
Parameters
The sample minimum and maximum are the MLEs of a and b respectively.
Example
The example illustrates the inversion method for generating normal random
numbers using rand and norminv. Note that the MATLAB function, randn,
does not use inversion since it is not efficient for this case.
u = rand(1000,1);
B-99
B Distribution Reference
x = norminv(u,0,1);
hist(x)
See Also
“Continuous Distributions (Data)” on page 5-4
B-100
Uniform Distribution (Discrete)
Uniform Distribution (Discrete)
In this section...
“Definition” on page B-101
“Background” on page B-101
“Example” on page B-101
“See Also” on page B-102
Definition
The discrete uniform pdf is
1
y = f ( x| N ) = I ( x)
N (1,..., N )
Background
The discrete uniform distribution is a simple distribution that puts equal
weight on the integers from one to N.
Example
As for all discrete distributions, the cdf is a step function. The plot shows
the discrete uniform cdf for N = 10.
x = 0:10;
y = unidcdf(x,10);
stairs(x,y)
set(gca,'Xlim',[0 11])
B-101
B Distribution Reference
Pick a random sample of 10 from a list of 553 items:
numbers = unidrnd(553,1,10)
numbers =
293 372 5 213 37 231 380 326 515 468
See Also
“Discrete Distributions” on page 5-7
B-102
Weibull Distribution
Weibull Distribution
In this section...
“Definition” on page B-103
“Background” on page B-103
“Parameters” on page B-104
“Example” on page B-104
“See Also” on page B-105
Definition
The Weibull pdf is
b
⎛ x⎞
−
− b b−1 ⎜⎝ a ⎟⎠
y = f ( x | a, b) = ba x e I( 0,∞ ) ( x )
Background
Waloddi Weibull offered the distribution that bears his name as an
appropriate analytical tool for modeling the breaking strength of materials.
Current usage also includes reliability and lifetime modeling. The Weibull
distribution is more flexible than the exponential for these purposes.
To see why, consider the hazard rate function (instantaneous failure rate). If
f(t) and F(t) are the pdf and cdf of a distribution, then the hazard rate is
f (t)
h (t) =
1 − F (t)
Substituting the pdf and cdf of the exponential distribution for f(t) and F(t)
above yields a constant. The example below shows that the hazard rate for
the Weibull distribution can vary.
B-103
B Distribution Reference
Parameters
Suppose you want to model the tensile strength of a thin filament using
the Weibull distribution. The function wblfit gives maximum likelihood
estimates and confidence intervals for the Weibull parameters.
strength = wblrnd(0.5,2,100,1); % Simulated strengths.
[p,ci] = wblfit(strength)
p =
0.4715 1.9811
ci =
0.4248 1.7067
0.5233 2.2996
The default 95% confidence interval for each parameter contains the true
value.
Example
The exponential distribution has a constant hazard function, which is not
generally the case for the Weibull distribution.
The plot shows the hazard functions for exponential (dashed line) and Weibull
(solid line) distributions having the same mean life. The Weibull hazard rate
here increases with age (a reasonable assumption).
t = 0:0.1:4.5;
h1 = exppdf(t,0.6267) ./ (1-expcdf(t,0.6267));
h2 = wblpdf(t,2,2) ./ (1-wblcdf(t,2,2));
plot(t,h1,'-',t,h2,'-')
B-104
Weibull Distribution
See Also
“Continuous Distributions (Data)” on page 5-4
B-105
B Distribution Reference
Wishart Distribution
In this section...
“Definition” on page B-106
“Background” on page B-106
“Example” on page B-107
“See Also” on page B-107
Definition
The probability density function of the d-dimensional Wishart distribution is
given by
⎛ 1 ⎞
(ν-d-1)/2 ) ⎜⎝ ( )
- trace Σ −1 Χ ⎟
Χ( e 2 ⎠
y = f(Χ, Σ, ν) =
ν/2
2(νd)/2 π(d(d-1))/4 ∑ Γ ( ν / 2 ) ...Γ(ν-(d-1))/2
where X and Σ are d-by-d symmetric positive definite matrices, and ν is
a scalar greater than d – 1. While it is possible to define the Wishart for
singular Σ, the density cannot be written as above.
Only random matrix generation is supported for the Wishart distribution,
including both singular and nonsingular Σ.
Background
The Wishart distribution is a generalization of the univariate chi-square
distribution to two or more variables. It is a distribution for symmetric
positive semidefinite matrices, typically covariance matrices, the diagonal
elements of which are each chi-square random variables. In the same way
as the chi-square distribution can be constructed by summing the squares of
independent, identically distributed, zero-mean univariate normal random
variables, the Wishart distribution can be constructed by summing the inner
products of independent, identically distributed, zero-mean multivariate
normal random vectors.
B-106
Wishart Distribution
The Wishart distribution is parameterized with a symmetric, positive
semidefinite matrix, Σ, and a positive scalar degrees of freedom parameter, ν.
ν is analogous to the degrees of freedom parameter of a univariate chi-square
distribution, and Σν is the mean of the distribution.
The Wishart distribution is often used as a model for the distribution of the
sample covariance matrix for multivariate normal random data, after scaling
by the sample size.
Example
If x is a bivariate normal random vector with mean zero and covariance matrix
⎛ 1 .5 ⎞
Σ=⎜ ⎟
⎝ .5 2 ⎠
then you can use the Wishart distribution to generate a sample covariance
matrix without explicitly generating x itself. Notice how the sampling
variability is quite large when the degrees of freedom is small.
Sigma = [1 .5; .5 2];
df = 10; S1 = wishrnd(Sigma,df)/df
S1 =
1.7959 0.64107
0.64107 1.5496
df = 1000; S2 = wishrnd(Sigma,df)/df
S2 =
0.9842 0.50158
0.50158 2.1682
See Also
“Multivariate Distributions” on page 5-8
B-107
B Distribution Reference
B-108
C
Bibliography
[1] Atkinson, A. C., and A. N. Donev. Optimum Experimental Designs. New
York: Oxford University Press, 1992.
[2] Bates, D. M., and D. G. Watts. Nonlinear Regression Analysis and Its
Applications. Hoboken, NJ: John Wiley & Sons, Inc., 1988.
[3] Belsley, D. A., E. Kuh, and R. E. Welsch. Regression Diagnostics.
Hoboken, NJ: John Wiley & Sons, Inc., 1980.
[4] Berry, M. W., et al. “Algorithms and Applications for Approximate
Nonnegative Matrix Factorization.” Computational Statistics and Data
Analysis. Vol. 52, No. 1, 2007, pp. 155–173.
[5] Bookstein, Fred L. Morphometric Tools for Landmark Data. Cambridge,
UK: Cambridge University Press, 1991.
[6] Bouye, E., V. Durrleman, A. Nikeghbali, G. Riboulet, and T. Roncalli.
“Copulas for Finance: A Reading Guide and Some Applications.” Working
Paper. Groupe de Recherche Operationnelle, Credit Lyonnais, 2000.
[7] Bowman, A. W., and A. Azzalini. Applied Smoothing Techniques for Data
Analysis. New York: Oxford University Press, 1997.
[8] Box, G. E. P., and N. R. Draper. Empirical Model-Building and Response
Surfaces. Hoboken, NJ: John Wiley & Sons, Inc., 1987.
[9] Box, G. E. P., W. G. Hunter, and J. S. Hunter. Statistics for Experimenters.
Hoboken, NJ: Wiley-Interscience, 1978.
C Bibliography
[10] Bratley, P., and B. L. Fox. “ALGORITHM 659 Implementing Sobol’s
Quasirandom Sequence Generator.” ACM Transactions on Mathematical
Software. Vol. 14, No. 1, 1988, pp. 88–100.
[11] Breiman, L., J. Friedman, R. Olshen, and C. Stone. Classification and
Regression Trees. Boca Raton, FL: CRC Press, 1984.
[12] Breiman, L., et al., Classification and Regression Trees, Chapman &
Hall, Boca Raton, 1993.
[13] Bulmer, M. G. Principles of Statistics. Mineola, NY: Dover Publications,
Inc., 1979.
[14] Bury, K.. Statistical Distributions in Engineering. Cambridge, UK:
Cambridge University Press, 1999.
[15] Chatterjee, S., and A. S. Hadi. “Influential Observations, High Leverage
Points, and Outliers in Linear Regression.” Statistical Science. Vol. 1, 1986,
pp. 379–416.
[16] Collett, D. Modeling Binary Data. New York: Chapman & Hall, 2002.
[17] Conover, W. J. Practical Nonparametric Statistics. Hoboken, NJ: John
Wiley & Sons, Inc., 1980.
[18] Cook, R. D., and S. Weisberg. Residuals and Influence in Regression.
New York: Chapman & Hall/CRC Press, 1983.
[19] Cox, D. R., and D. Oakes. Analysis of Survival Data. London: Chapman
& Hall, 1984.
[20] Davidian, M., and D. M. Giltinan. Nonlinear Models for Repeated
Measurements Data. New York: Chapman & Hall, 1995.
[21] Deb, P., and M. Sefton. “The Distribution of a Lagrange Multiplier Test
of Normality.” Economics Letters. Vol. 51, 1996, pp. 123–130.
[22] de Jong, S. “SIMPLS: An Alternative Approach to Partial Least Squares
Regression.” Chemometrics and Intelligent Laboratory Systems. Vol. 18, 1993,
pp. 251–263.
C-2
Bibliography
[23] Demidenko, E. Mixed Models: Theory and Applications. Hoboken, NJ:
John Wiley & Sons, Inc., 2004.
[24] Delyon, B., M. Lavielle, and E. Moulines, Convergence of a stochastic
approximation version of the EM algorithm, Annals of Statistics, 27, 94-128,
1999.
[25] Dempster, A. P., N. M. Laird, and D. B. Rubin. “Maximum Likelihood
from Incomplete Data via the EM Algorithm.” Journal of the Royal Statistical
Society. Series B, Vol. 39, No. 1, 1977, pp. 1–37.
[26] Devroye, L. Non-Uniform Random Variate Generation. New York:
Springer-Verlag, 1986.
[27] Dobson, A. J. An Introduction to Generalized Linear Models. New York:
Chapman & Hall, 1990.
[28] Draper, N. R., and H. Smith. Applied Regression Analysis. Hoboken,
NJ: Wiley-Interscience, 1998.
[29] Drezner, Z. “Computation of the Trivariate Normal Integral.”
Mathematics of Computation. Vol. 63, 1994, pp. 289–294.
[30] Drezner, Z., and G. O. Wesolowsky. “On the Computation of the Bivariate
Normal Integral.” Journal of Statistical Computation and Simulation. Vol.
35, 1989, pp. 101–107.
[31] DuMouchel, W. H., and F. L. O’Brien. “Integrating a Robust Option
into a Multiple Regression Computing Environment.” Computer Science and
Statistics: Proceedings of the 21st Symposium on the Interface. Alexandria,
VA: American Statistical Association, 1989.
[32] Durbin, R., S. Eddy, A. Krogh, and G. Mitchison. Biological Sequence
Analysis. Cambridge, UK: Cambridge University Press, 1998.
[33] Efron, B., and R. J. Tibshirani. An Introduction to the Bootstrap. New
York: Chapman & Hall, 1993.
[34] Embrechts, P., C. Klüppelberg, and T. Mikosch. Modelling Extremal
Events for Insurance and Finance. New York: Springer, 1997.
C-3
C Bibliography
[35] Evans, M., N. Hastings, and B. Peacock. Statistical Distributions. 2nd
ed., Hoboken, NJ: John Wiley & Sons, Inc., 1993, pp. 50–52, 73–74, 102–105,
147, 148.
[36] Genz, A. “Numerical Computation of Rectangular Bivariate and
Trivariate Normal and t Probabilities.” Statistics and Computing. Vol. 14,
No. 3, 2004, pp. 251–260.
[37] Genz, A., and F. Bretz. “Comparison of Methods for the Computation
of Multivariate t Probabilities.” Journal of Computational and Graphical
Statistics. Vol. 11, No. 4, 2002, pp. 950–971.
[38] Genz, A., and F. Bretz. “Numerical Computation of Multivariate t
Probabilities with Application to Power Calculation of Multiple Contrasts.”
Journal of Statistical Computation and Simulation. Vol. 63, 1999, pp.
361–378.
[39] Gibbons, J. D. Nonparametric Statistical Inference. New York: Marcel
Dekker, 1985.
[40] Goodall, C. R. “Computation Using the QR Decomposition.” Handbook in
Statistics. Vol. 9, Amsterdam: Elsevier/North-Holland, 1993.
[41] Hahn, Gerald J., and S. S. Shapiro. Statistical Models in Engineering.
Hoboken, NJ: John Wiley & Sons, Inc., 1994, p. 95.
[42] Hald, A. Statistical Theory with Engineering Applications. Hoboken,
NJ: John Wiley & Sons, Inc., 1960.
[43] Harman, H. H. Modern Factor Analysis. 3rd Ed. Chicago: University of
Chicago Press, 1976.
[44] Hastie, T., R. Tibshirani, and J. Friedman. The Elements of Statistical
Learning. New York: Springer, 2001.
[45] Hochberg, Y., and A. C. Tamhane. Multiple Comparison Procedures.
Hoboken, NJ: John Wiley & Sons, 1987.
[46] Hoerl, A. E., and R. W. Kennard. “Ridge Regression: Applications to
Nonorthogonal Problems.” Technometrics. Vol. 12, No. 1, 1970, pp. 69–82.
C-4
Bibliography
[47] Hoerl, A. E., and R. W. Kennard. “Ridge Regression: Biased Estimation
for Nonorthogonal Problems.” Technometrics. Vol. 12, No. 1, 1970, pp. 55–67.
[48] Hogg, R. V., and J. Ledolter. Engineering Statistics. New York:
MacMillan, 1987.
[49] Holland, P. W., and R. E. Welsch. “Robust Regression Using Iteratively
Reweighted Least-Squares.” Communications in Statistics: Theory and
Methods, A6, 1977, pp. 813–827.
[50] Hollander, M., and D. A. Wolfe. Nonparametric Statistical Methods.
Hoboken, NJ: John Wiley & Sons, Inc., 1999.
[51] Hong, H. S., and F. J. Hickernell. “ALGORITHM 823: Implementing
Scrambled Digital Sequences.” ACM Transactions on Mathematical Software.
Vol. 29, No. 2, 2003, pp. 95–109.
[52] Huber, P. J. Robust Statistics. Hoboken, NJ: John Wiley & Sons, Inc.,
1981.
[53] Jackson, J. E. A User’s Guide to Principal Components. Hoboken, NJ:
John Wiley and Sons, 1991.
[54] Jain, A., and R. Dubes. Algorithms for Clustering Data. Upper Saddle
River, NJ: Prentice-Hall, 1988.
[55] Jarque, C. M., and A. K. Bera. “A test for normality of observations
and regression residuals.” International Statistical Review. Vol. 55, No. 2,
1987, pp. 163–172.
[56] Joe, S., and F. Y. Kuo. “Remark on Algorithm 659: Implementing Sobol’s
Quasirandom Sequence Generator.” ACM Transactions on Mathematical
Software. Vol. 29, No. 1, 2003, pp. 49–57.
[57] Johnson, N., and S. Kotz. Distributions in Statistics: Continuous
Univariate Distributions-2. Hoboken, NJ: John Wiley & Sons, Inc., 1970, pp.
130–148, 189–200, 201–219.
[58] Johnson, N. L., N. Balakrishnan, and S. Kotz. Continuous Multivariate
Distributions. Vol. 1. Hoboken, NJ: Wiley-Interscience, 2000.
C-5
C Bibliography
[59] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous Univariate
Distributions. Vol. 1, Hoboken, NJ: Wiley-Interscience, 1993.
[60] Johnson, N. L., S. Kotz, and N. Balakrishnan. Continuous Univariate
Distributions. Vol. 2, Hoboken, NJ: Wiley-Interscience, 1994.
[61] Johnson, N. L., S. Kotz, and N. Balakrishnan. Discrete Multivariate
Distributions. Hoboken, NJ: Wiley-Interscience, 1997.
[62] Johnson, N. L., S. Kotz, and A. W. Kemp. Univariate Discrete
Distributions. Hoboken, NJ: Wiley-Interscience, 1993.
[63] Jolliffe, I. T. Principal Component Analysis. 2nd ed., New York:
Springer-Verlag, 2002.
[64] Jöreskog, K. G. “Some Contributions to Maximum Likelihood Factor
Analysis.” Psychometrika. Vol. 32, 1967, pp. 443–482.
[65] Kaufman L., and P. J. Rousseeuw. Finding Groups in Data: An
Introduction to Cluster Analysis. Hoboken, NJ: John Wiley & Sons, Inc., 1990.
[66] Kendall, David G. “A Survey of the Statistical Theory of Shape.”
Statistical Science. Vol. 4, No. 2, 1989, pp. 87–99.
[67] Kocis, L., and W. J. Whiten. “Computational Investigations of
Low-Discrepancy Sequences.” ACM Transactions on Mathematical Software.
Vol. 23, No. 2, 1997, pp. 266–294.
[68] Kotz, S., and S. Nadarajah. Extreme Value Distributions: Theory and
Applications. London: Imperial College Press, 2000.
[69] Krzanowski, W. J. Principles of Multivariate Analysis: A User’s
Perspective. New York: Oxford University Press, 1988.
[70] Lawless, J. F. Statistical Models and Methods for Lifetime Data.
Hoboken, NJ: Wiley-Interscience, 2002.
[71] Lawley, D. N., and A. E. Maxwell. Factor Analysis as a Statistical
Method. 2nd ed. New York: American Elsevier Publishing, 1971.
C-6
Bibliography
[72] Lilliefors, H. W. “On the Kolmogorov-Smirnov test for normality
with mean and variance unknown.” Journal of the American Statistical
Association. Vol. 62, 1967, pp. 399–402.
[73] Lilliefors, H. W. “On the Kolmogorov-Smirnov test for the exponential
distribution with mean unknown.” Journal of the American Statistical
Association. Vol. 64, 1969, pp. 387–389.
[74] Lindstrom, M. J., and D. M. Bates. “Nonlinear mixed-effects models for
repeated measures data.” Biometrics. Vol. 46, 1990, pp. 673–687.
[75] Little, Roderick J. A., and Donald B. Rubin. Statistical Analysis with
Missing Data. 2nd ed., Hoboken, NJ: John Wiley & Sons, Inc., 2002.
[76] Mardia, K. V., J. T. Kent, and J. M. Bibby. Multivariate Analysis.
Burlington, MA: Academic Press, 1980.
[77] Marquardt, D.W. “Generalized Inverses, Ridge Regression, Biased
Linear Estimation, and Nonlinear Estimation.” Technometrics. Vol. 12, No.
3, 1970, pp. 591–612.
[78] Marquardt, D. W., and R.D. Snee. “Ridge Regression in Practice.” The
American Statistician. Vol. 29, No. 1, 1975, pp. 3–20.
[79] Marsaglia, G., and W. W. Tsang. “A Simple Method for Generating
Gamma Variables.” ACM Transactions on Mathematical Software. Vol. 26,
2000, pp. 363–372.
[80] Marsaglia, G., W. Tsang, and J. Wang. “Evaluating Kolmogorov’s
Distribution.” Journal of Statistical Software. Vol. 8, Issue 18, 2003.
[81] Martinez, W. L., and A. R. Martinez. Computational Statistics with
MATLAB. New York: Chapman & Hall/CRC Press, 2002.
[82] Massey, F. J. “The Kolmogorov-Smirnov Test for Goodness of Fit.”
Journal of the American Statistical Association. Vol. 46, No. 253, 1951, pp.
68–78.
[83] Matousek, J. “On the L2-Discrepancy for Anchored Boxes.” Journal of
Complexity. Vol. 14, No. 4, 1998, pp. 527–556.
C-7
C Bibliography
[84] McLachlan, G., and D. Peel. Finite Mixture Models. Hoboken, NJ: John
Wiley & Sons, Inc., 2000.
[85] McCullagh, P., and J. A. Nelder. Generalized Linear Models. New York:
Chapman & Hall, 1990.
[86] McGill, R., J. W. Tukey, and W. A. Larsen. “Variations of Boxplots.” The
American Statistician. Vol. 32, No. 1, 1978, pp. 12–16.
[87] Meeker, W. Q., and L. A. Escobar. Statistical Methods for Reliability
Data. Hoboken, NJ: John Wiley & Sons, Inc., 1998.
[88] Meng, Xiao-Li, and Donald B. Rubin. “Maximum Likelihood Estimation
via the ECM Algorithm.” Biometrika. Vol. 80, No. 2, 1993, pp. 267–278.
[89] Meyers, R. H., and D.C. Montgomery. Response Surface Methodology:
Process and Product Optimization Using Designed Experiments. Hoboken,
NJ: John Wiley & Sons, Inc., 1995.
[90] Miller, L. H. “Table of Percentage Points of Kolmogorov Statistics.”
Journal of the American Statistical Association. Vol. 51, No. 273, 1956, pp.
111–121.
[91] Milliken, G. A., and D. E. Johnson. Analysis of Messy Data, Volume 1:
Designed Experiments. Boca Raton, FL: Chapman & Hall/CRC Press, 1992.
[92] Montgomery, D. Introduction to Statistical Quality Control. Hoboken,
NJ: John Wiley & Sons, 1991, pp. 369–374.
[93] Montgomery, D. C. Design and Analysis of Experiments. Hoboken, NJ:
John Wiley & Sons, Inc., 2001.
[94] Mood, A. M., F. A. Graybill, and D. C. Boes. Introduction to the Theory of
Statistics. 3rd ed., New York: McGraw-Hill, 1974. pp. 540–541.
[95] Moore, J. Total Biochemical Oxygen Demand of Dairy Manures. Ph.D.
thesis. University of Minnesota, Department of Agricultural Engineering,
1975.
C-8
Bibliography
[96] Mosteller, F., and J. Tukey. Data Analysis and Regression. Upper Saddle
River, NJ: Addison-Wesley, 1977.
[97] Nelson, L. S. “Evaluating Overlapping Confidence Intervals.” Journal of
Quality Technology. Vol. 21, 1989, pp. 140–141.
[98] Patel, J. K., C. H. Kapadia, and D. B. Owen. Handbook of Statistical
Distributions. New York: Marcel Dekker, 1976.
[99] Pinheiro, J. C., and D. M. Bates. “Approximations to the log-likelihood
function in the nonlinear mixed-effects model.” Journal of Computational and
Graphical Statistics. Vol. 4, 1995, pp. 12–35.
[100] Rice, J. A. Mathematical Statistics and Data Analysis. Pacific Grove,
CA: Duxbury Press, 1994.
[101] Rosipal, R., and N. Kramer. “Overview and Recent Advances in
Partial Least Squares.” Subspace, Latent Structure and Feature Selection:
Statistical and Optimization Perspectives Workshop (SLSFS 2005), Revised
Selected Papers (Lecture Notes in Computer Science 3940). Berlin, Germany:
Springer-Verlag, 2006, pp. 34–51.
[102] Sachs, L. Applied Statistics: A Handbook of Techniques. New York:
Springer-Verlag, 1984, p. 253.
[103] Searle, S. R., F. M. Speed, and G. A. Milliken. “Population marginal
means in the linear model: an alternative to least-squares means.” American
Statistician. 1980, pp. 216–221.
[104] Seber, G. A. F. Linear Regression Analysis. Hoboken, NJ:
Wiley-Interscience, 2003.
[105] Seber, G. A. F. Multivariate Observations. Hoboken, NJ: John Wiley &
Sons, Inc., 1984.
[106] Seber, G. A. F., and C. J. Wild. Nonlinear Regression. Hoboken, NJ:
Wiley-Interscience, 2003.
[107] Sexton, Joe, and A. R. Swensen. “ECM Algorithms that Converge at the
Rate of EM.” Biometrika. Vol. 87, No. 3, 2000, pp. 651–662.
C-9
C Bibliography
[108] Snedecor, G. W., and W. G. Cochran. Statistical Methods. Ames, IA:
Iowa State Press, 1989.
[109] Spath, H. Cluster Dissection and Analysis: Theory, FORTRAN
Programs, Examples. Translated by J. Goldschmidt. New York: Halsted
Press, 1985.
[110] Stein, M. “Large sample properties of simulations using latin hypercube
sampling.” Technometrics. Vol. 29, No. 2, 1987, pp. 143–151. Correction,
Vol. 32, p. 367.
[111] Stephens, M. A. “Use of the Kolmogorov-Smirnov, Cramer-Von Mises
and Related Statistics Without Extensive Tables.” Journal of the Royal
Statistical Society. Series B, Vol. 32, No. 1, 1970, pp. 115–122.
[112] Street, J. O., R. J. Carroll, and D. Ruppert. “A Note on Computing
Robust Regression Estimates via Iteratively Reweighted Least Squares.” The
American Statistician. Vol. 42, 1988, pp. 152–154.
[113] Student. “On the Probable Error of the Mean.” Biometrika. Vol. 6,
No. 1, 1908, pp. 1–25.
[114] Vellemen, P. F., and D. C. Hoaglin. Application, Basics, and Computing
of Exploratory Data Analysis. Pacific Grove, CA: Duxbury Press, 1981.
[115] Weibull, W. “A Statistical Theory of the Strength of Materials.”
Ingeniors Vetenskaps Akademiens Handlingar. Stockholm: Royal Swedish
Institute for Engineering Research, No. 151, 1939.
[116] Zahn, C. T. “Graph-theoretical methods for detecting and describing
Gestalt clusters.” IEEE Transactions on Computers. Vol. C-20, Issue 1,
1971, pp. 68–86.
C-10
Index
A
Index computing with 2-31
absolute deviation 3-5 constructing 2-25
added variable plots creating 2-24
adding new term to model 9-23 removing observations from 2-31
from stepwise 9-29 multidimensional 2-6
addedvarplot 18-2 numerical 2-4
additive effects 8-9 statistical 2-11
adjacent value 18-90 average linkage 18-699
adjacent values 4-7
AIC. See Akaike Information Criterion B
Akaike Information Criterion (AIC) 5-105 18-8
bacteria counts 8-4
alternative hypotheses 7-3
Bartlett multiple-sample test 7-15
analysis of variance
barttest 18-42
F distribution B-26
batch updates 18-640
functions 16-32
Bayes classification
multivariate 8-39
objects 16-43 17-5
N-way 8-12
Bayes Information Criterion (BIC) 5-105 18-58
one-way 8-3
bbdesign 18-43
two-way 8-9
Bernoulli distribution B-3
visualization functions 16-12 16-32
Bernoulli random variables 18-66
andrewsplot 18-9
beta distribution B-4
ANOVA tables
betacdf 18-46
regression 9-13
betafit 18-47
anova1 18-13
betainv 18-49
anova2 18-19
betalike 18-51
anovan 18-23
betapdf 18-53
Ansari-Bradley test 7-13
betarnd 18-55
ansaribradley 18-33
betastat 18-57
aoctool 8-27 18-36
BIC. See Bayes Information Criterion
arrays
binocdf 18-59
categorical
binofit 18-61
accessing 2-18
binoinv 18-63
combining 2-19
binomial distribution B-7
computing with 2-20
negative B-72
constructing 2-16
binopdf 18-64
implementation 2-14
binornd 18-66
types 2-14
binostat 18-68
dataset
biplot 18-69
accessing 2-27
Birnbaum-Saunders distribution B-10
combining 2-29
Index-1
Index
bootci 18-72 central composite designs (CCDs)
bootstrapping 3-9 generating 18-129
bootstrp 18-76 types 14-9
box plots 4-6 Central Limit Theorem B-83
Box-Behnken designs 14-13 central tendency
generating 18-43 functions 16-8
Box-Wilson designs 14-9 centroid linkage 18-699
boxplot 18-84 chi-square distribution B-12
chi-square goodness-of-fit test 7-13
chi-square variance test, one-sample 7-14
C
chi2cdf 18-145
candexch 18-99 chi2gof 18-146
candgen 18-104 chi2inv 18-151
candidate sets 14-17 chi2pdf 18-153
canoncorr 18-108 chi2rnd 18-155
Canonical Maximum Likelihood (CML) 18-228 chi2stat 18-157
capability 18-111 cholcov 18-162
capability studies 15-6 circuit boards 18-64
capaplot 18-114 city block metric 12-15 18-1051 18-1057
case names classical multidimensional scaling
reading from file 18-116 cmdscale function 18-198
writing to file 18-117 overview 10-3
caseread 18-116 classification
casewrite 18-117 functions 16-42
categorical arrays objects 17-5
accessing 2-18 visualization functions 16-14 16-42
combining 2-19 Classification
computing with 2-20 naive bayes 12-6
constructing 2-16 performance curves 12-53
functions 16-3 tree bagging 12-30
implementation 2-14 trees 12-9
objects 17-2 classification trees
types 2-14 example 12-9
categorical data 2-13 functions 16-42
CCD. See central composite designs objects 17-5
ccdesign 18-129 classifiers 12-2
cdf 18-132 classify 18-169
cdfplot 18-140 cluster 18-188
cell arrays cluster analysis
storing heterogeneous data in 2-7 functions 16-40
Index-2
Index
hierarchical clustering 11-3 copulapdf 18-235
K-means clustering 11-21 copularnd 18-239
overview 11-2 copulas 5-108 B-14
visualization functions 16-13 16-40 copulastat 18-237
cluster trees cordexch 18-241
constructing clusters from 18-188 corr 18-245
creating 18-697 corrcov 18-248
creating, from data 18-195 correlation
inconsistency coefficient 18-585 functions 16-10
plotting 18-313 Cox proportional hazards fit 18-251
clusterdata 18-195 coxphfit 18-251
cmdscale 18-198 criterion function 10-23
CML. See Canonical Maximum Likelihood crosstab 18-259
coefficients crossval 18-262
linear model 9-3 cumulative distribution
combnk 18-202 functions 16-19
common factors 10-46 cumulative distribution function (cdf)
comparisons, multiple 8-6 empirical 5-63
complete linkage 18-699 for parametric estimation 5-62
confidence intervals graphing an estimate 4-12
communicating results of hypothesis curse of dimensionality 10-2
tests 7-4 cut variables 18-278
nonlinear regression 9-61
confounding effects 14-5
D
confounding patterns 14-7
confusionmat 18-209 D-optimal designs
container variables 2-2 creating from candidate set 18-99
continuous distributions functions 16-48
data 5-4 generating candidate set 18-104
statistics 5-6 overview 14-15
control charts 15-3 data
controlchart 18-212 categorical 2-13
controlrules 18-218 heterogeneous 2-7
Cook’s distance 18-1204 landmark 10-14
cophenet 18-223 statistical 2-23
cophenetic correlation coefficients 11-10 18-223 data containers 2-2
cophenetic distance 11-10 data organization
copulacdf 18-225 functions 16-3
copulafit 18-227 objects 17-2
copulaparam 18-233 data sets
Index-3
Index
normalizing 11-4 visualization functions 16-14 16-47
statistical examples A-2 dfittool 18-317
dataset arrays dimension reduction
accessing 2-27 common factor analysis 18-395
combining 2-29 multivariate statistical methods 10-2
computing with 2-31 PCA from covariance matrix 18-1032
constructing 2-25 PCA from raw data matrix 18-1114
creating 2-24 PCA residuals 18-1034
functions 16-6 to 16-7 discrete distributions 5-7
objects 17-2 discrete uniform distribution B-101
daugment 18-300 discriminant analysis 12-3
dcovary 18-304 functions 16-42
decision trees discriminant functions 12-3
computing error rate 18-1416 dispersion
computing response values 18-1420 functions 16-8
creating 18-1408 dissimilarity matrices
creating subtrees 18-1411 creating 11-4
displaying 18-1405 distance matrices
fitting 18-1408 creating 11-4
pruning 18-1411 distribution
dendrogram 18-313 objects 16-15 17-3
density estimation visualization functions 16-11 16-16
ksdensity function 18-663 distribution fitting
descriptive statistics functions 5-70 16-24
functions 16-8 tool 5-11
design matrices 9-5 distribution statistics
design matrix 9-66 functions 5-68 16-23
design of experiments distributions
basic factors 14-6 custom B-15
confounding effects 14-5 functions that support 5-52
D-optimal designs 14-15 disttool 18-340
fractional factorial designs 14-5 dummyvar 18-345
full factorial designs 14-3 Durbin-Watson test 7-13
functions 16-47 dwtest 18-348
generators 14-6
levels 14-3
E
Plackett-Burman designs 14-5
resolution 14-6 18-442 ecdf 18-350
response surface designs 14-9 ecdfhist 18-353
two-level designs 14-4 effects
Index-4
Index
fixed 9-64 F
random 9-64 F distribution B-25
statistical 9-64 F-test, one-sample 7-14
efinv 18-362 factor analysis
emission matrices functions 16-39
estimating 13-9 maximum likelihood 18-395
empirical cumulative distribution function 5-63 factoran 18-395
18-350 factorial designs
ensemble methods fractional 14-5
functions 16-36 16-44 full 14-3
objects 17-4 to 17-5 generating fractional 18-440
equal variances generating full 18-455
Bartlett multiple-sample test for 7-15 fcdf 18-410
F-test for 7-14 feature selection
erf B-83 functions 16-39
error function B-83 overview 10-23
Euclidean distance 12-14 18-1050 18-1056 sequential 10-23
evcdf 18-358 feature transformation
evfit 18-360 functions 16-39
evlike 18-373 overview 10-28
evpdf 18-374 ff2n 18-412
evrnd 18-375 file I/O
evstat 18-376 functions 16-2
expcdf 18-377 filter methods
expectation maximization (EM) algorithm feature selection 18-1272
cluster analysis 11-2 finv 18-416
Gaussian mixture models 5-99 11-28 fitdist 18-429
expfit 18-379 folds
expinv 18-386 partition 18-283
explike 18-388 fpdf 18-439
exponential distribution B-16 fracfact 18-440
exppdf 18-392 fracfactgen 18-442
exprnd 18-393 fractional factorial designs
expstat 18-394 functions 16-48
extrapolated 18-1163 generating 18-440
extreme value distribution B-19 overview 14-5
extreme value fit 18-360 friedman 18-445
Friedman’s test 8-37
frnd 18-449
fstat 18-451
Index-5
Index
fsurfht 18-452 gevpdf 18-488
full factorial designs gevrnd 18-489
functions 16-47 gevstat 18-490
generating 18-455 gline 18-491
overview 14-3 glmfit 18-493
fullfact 18-455 glmval 18-498
functions glyphplot 18-501
vectorized 2-9 gname 18-511
furthest neighbor linkage 18-699 gpcdf 18-513
gpfit 18-514
gpinv 18-516
G
gplike 18-517
gagerr 18-456 gplotmatrix 18-519
gamcdf 18-461 gppdf 18-518
gamfit 18-463 gprnd 18-522
gaminv 18-465 gpstat 18-523
gamlike 18-467 graphical user interfaces
gamma distribution B-27 functions 16-52
gampdf 18-469 group mean clusters, plot 8-44
gamrnd 18-470 grouped plot matrices 8-40
gamstat 18-471 grouping variables
Gauss-Markov theorem 9-5 functions for 2-35
Gaussian distribution B-30 use for computing statistics 2-34
Gaussian mixture distributions B-31 using 2-36
Gaussian mixture models grp2idx 18-527
functions 16-41 grpstats 18-529
objects 17-4 gscatter 18-537
generalized extreme value distribution B-32
generalized Pareto distribution B-37
geocdf 18-473 H
geoinv 18-474 harmmean 18-544
geomean 18-475 hat matrix 9-7
geometric distribution B-41 heterogeneous data
geopdf 18-476 storing, in MATLAB 2-7
geornd 18-477 hidden Markov models
geostat 18-478 functions 16-46
gevcdf 18-483 overview 13-5
gevfit 18-484 hierarchical clustering
gevinv 18-486 cluster analysis 11-3
gevlike 18-487 computing inconsistency coefficient 18-585
Index-6
Index
constructing clusters 18-188 IFM. See Inference Functions for Margins method
cophenetic correlation coefficients 18-223 incomplete beta function B-4
creating cluster trees 18-697 incomplete gamma function B-27
creating clusters 11-16 inconsistency coefficient 18-585
creating clusters from data 18-195 inconsistent 18-585
determining proximity 18-1048 18-1054 Inference Functions for Margins (IFM)
evaluating cluster formation 18-223 method 18-228
functions 16-40 initial state distribution
grouping objects 11-6 changing 13-12
inconsistency coefficient 18-585 interaction effects
plotting cluster trees 18-313 designed experiments 14-2
procedure 11-3 two-way ANOVA 8-9
hist3 18-545 interactionplot 18-593
histfit 18-553 interquartile range (iqr) 3-6
histogram fit 18-553 inverse cumulative distribution
hmmdecode 18-556 functions 5-66 16-21
hmmestimate 13-9 18-558 inverse Gaussian distribution B-45
hmmgenerate 18-561 inverse Wishart distribution B-46 B-106
hmmtrain 13-10 18-563 invpred 18-596
hmmviterbi 18-566 iqr 18-599
holdout iwishrnd 18-617
partition 18-283
Hotelling’s T-squared 10-42
J
hougen 18-570
hygecdf 18-571 jackknife 18-618
hygeinv 18-572 Jarque-Bera test 7-13 18-620
hygepdf 18-573 jbtest 18-620
hygernd 18-574 Johnson system of distributions 6-27 B-48
hygestat 18-575 johnsrnd 18-623
hypergeometric distribution B-43
hypotheses B-26 K
hypothesis tests
K-means clustering
assumptions 7-5
cluster separation 11-22
functions 16-31
functions 16-41
functions that support 7-13
local minima 11-26
power 7-4 18-1257
number of clusters 11-23
overview 11-21
I silhouette plot 18-1290
icdf 18-576
Index-7
Index
Kaplan-Meier cumulative distribution multiple 9-8
function 18-350 polynomial 9-37
kernel bandwidth 5-57 response surfaces 9-45
kernel smoothing functions ridge 9-29
specifying 5-59 robust 9-14
kmeans 18-637 stepwise 9-19
Kolmogorov-Smirnov test linear transformations
one-sample 7-13 Procrustes 18-1144
two-sample 7-13 linhyptest 18-695
Kruskal-Wallis test 8-36 link functions 9-53
kruskalwallis 18-659 linkage 18-697
ksdensity 18-663 average 18-699
kstest 18-669 centroid 18-699
kstest2 18-674 complete 18-699
kurtosis 18-678 furthest neighbor 18-699
nearest neighbor 18-699
single 18-699
L
ward 18-700
landmark data 10-14 loadings 10-36 10-46
latin hypercube designs logistic distribution B-49
functions 16-48 logistic models 9-54
latin hypercube sample 18-690 logistic regression
normal distribution 18-691 stepwise 9-56
least squares loglogistic distribution B-50
iteratively reweighted 9-14 logncdf 18-701
leverage 18-688 lognfit 18-703
leverage plots logninv 18-705
partial regression 9-23 lognlike 18-707
leverage, linear regression models 9-7 lognormal distribution B-51
lhsdesign 18-690 lognormal fit 18-703
lhsnorm 18-691 lognpdf 18-708
likelihood function 18-53 lognrnd 18-710
Lilliefors test 7-13 lognstat 18-712
example 7-7 loss
lillietest 18-692 prediction 18-1270
linear hypothesis test 7-13 lsline 18-716
linear models
generalized 9-52
linear regression M
functions 16-34 mad 18-718
Index-8
Index
mahal 18-720 mhsample 18-757
Mahalanobis distance Minkowski metric 12-15 18-1051 18-1057
computing 18-720 18-722 missing data 3-16
in cluster analysis 12-14 18-1050 18-1057 missing values
main effects 14-2 functions 16-9
maineffectsplot 18-726 mixed-effects models 9-65
Mann-Whitney U-test 18-1184 mle 18-762
MANOVA 8-39 MLE. See maximum likelihood — estimation
manova1 18-728 mlecov 18-768
manovacluster 18-732 mnpdf 18-772
Markov chains mnrfit 18-774
emission matrix 13-4 mnrnd 18-781
emissions 13-4 mnrval 18-783
initial state 13-4 model assessment
Monte Carlo simulations 6-15 functions 16-41
overview 13-3 objects 17-5
transition matrices 13-4 models
Markov models mixed-effects 9-65
hidden moment 18-789
functions for 13-7 MS. See mean squares
generating test sequences for 13-8 multcompare 18-792
overview 13-5 multicollinearity 18-1219
state diagram 13-3 addressed by ridge regression 9-29
maximum likelihood multidimensional arrays
coefficient estimates 9-5 classical (metric) scaling 18-198
estimation 5-70 multidimensional scaling (MDS)
factor analysis 18-395 classical (metric) 10-3
MCMC 6-15 functions 16-38
MDS. See multidimensional scaling multinomial distribution B-54
mdscale 18-738 multiple comparison procedure 18-792
mean multiple linear regression 9-8
of probability distribution 5-68 multivariate analysis of variance
mean absolute deviation 18-718 example 8-39
mean squares (MS) 18-14 multivariate distributions 5-8
measures of multivariate Gaussian distribution B-57
central tendency 3-3 multivariate normal distribution B-58
dispersion 3-5 multivariate regression 9-4 9-57
median absolute deviation 18-718 multivariate statistics
metric multidimensional scaling 10-3 analysis of variance 8-39
See also classical multidimensional scaling functions 16-38
Index-9
Index
principal component analysis 10-31 nctrnd 18-872
visualization functions 16-13 16-38 nctstat 18-874
multivariate t distribution B-64 ncx2cdf 18-876
multivarichart 18-802 ncx2inv 18-878
mvncdf 18-806 ncx2pdf 18-879
mvnpdf 18-810 ncx2rnd 18-881
mvnrnd 18-821 ncx2stat 18-883
mvregress 18-812 nearest neighbor linkage 18-699
mvregresslike 18-819 negative binomial distribution
mvtcdf 18-823 confidence intervals 18-848
mvtpdf 18-827 cumulative distribution function (cdf) 18-846
mvtrnd 18-829 definition B-72
inverse cumulative distribution function
(cdf) 18-849
N
mean and variance 18-854
Nakagami distribution B-70 modeling number of auto accidents B-73
nancov 18-836 nbincdf function 18-846
nanmax 18-838 nbininv function 18-849
nanmean 18-839 nbinpdf function 18-850
nanmedian 18-840 parameter estimates 18-848
nanmin 18-841 probability density function (pdf) 18-850
NaNs random matrices 18-852
coding missing values as 3-16 negative binomial fit 18-848
nanstd 18-842 negative log-likelihood
nansum 18-843 functions 5-77 16-26
nanvar 18-844 Newton’s method 18-465
nbincdf 18-846 nlinfit 18-894
nbinfit 18-848 nlintool 18-898
nbininv 18-849 nlmefit 18-900
nbinpdf 18-850 nlparci 18-931
nbinrnd 18-852 nlpredci 18-933
nbinstat 18-854 nnmf 18-936
ncfcdf 18-856 noncentral F distribution B-78
ncfinv 18-858 nonlinear least-squares fit 18-894
ncfpdf 18-860 nonlinear mixed effects 18-900
ncfrnd 18-862 nonlinear regression
ncfstat 18-864 functions 16-35
nctcdf 18-868 nonnegative matrix factorization
nctinv 18-869 dimension-reduction technique 10-29
nctpdf 18-870 functions 16-39
Index-10
Index
nonparametric distributions 5-8 B-82 pcares 18-1034
normal distribution B-83 pdf 18-1037
normal equations 9-6 pdist 18-1048
normal fit 18-961 Pearson system of distributions 6-27 B-87
normal probability plots 4-8 pearsrnd 18-1060
normalizing percentiles
data sets 11-4 computing 3-7
normcdf 18-959 perms 18-1074
normfit 18-961 piecewise distribution fitting
norminv 18-963 functions 16-25
normlike 18-965 piecewise distributions B-88
normpdf 18-966 functions 16-29
normplot 18-967 objects 17-4
normrnd 18-969 Plackett-Burman designs 14-5
normspec 18-971 plsregress 18-1078
normstat 18-973 poisscdf 18-1086
null hypotheses 7-3 poissfit 18-1088
numerical arrays 2-4 poissinv 18-1089
Poisson distribution B-89
Poisson fit 18-1088
O
poisspdf 18-1090
one-sample Kolmogorov-Smirnov test 7-13 poissrnd 18-1091
online updates 18-641 poisstat 18-1093
outliers polyconf 18-1094
measures resistant to 3-3 polynomial regression 9-37
regression 9-11 polytool 18-1100
posterior state probabilities
P estimating 13-11
power
p values 7-3
hypothesis tests 7-4
parallel regression 18-775
prctile 18-1107
parallelcoords 18-1009
principal component analysis (PCA)
pareto 18-1022
component scores 10-36
Pareto distribution B-86
component variances 10-40
partial least-squares regression 9-33
functions 16-39
partial regression
Hotelling’s T-squared 10-42
leverage plots 9-23
overview 10-31
partialcorr 18-1029
principal components 10-36
PCA. See principal component analysis
quality of life example 10-33
pcacov 18-1032
scree plots 10-41
Index-11
Index
principal coordinates analysis 10-4 objects 17-6
princomp 18-1114 quasi-random numbers
probabilities functions 16-28
posterior state, estimating 13-11 generating 6-17
probability density objects 17-3
functions 5-52 16-17 sequences
probability density estimation leaping 6-19
comparing estimates 5-60 point set 6-19
function 18-663 scrambling 6-19
kernel bandwidth 5-57 skipping 6-19
kernel smoothing functions 5-59 streams 6-25
nonparametric estimation 5-55 state 6-25
Probability Distribution Function Tool 5-9
probability distributions
R
disttool 5-9
functions 16-15 randg 18-1167
functions that support 5-3 random 18-1169
mean and variance 5-68 random number generation
objects 17-3 acceptance-rejection methods 6-9
piecewise 5-70 direct methods 6-5
probability mass functions inversion methods 6-7
pmf 5-52 methods 6-5
probplot 18-1139 Random Number Generation Tool 5-49
procrustes 18-1144 random number generators (RNGs) 5-80 6-2
Procrustes analysis 10-14 18-1144 random numbers
functions 16-38 functions 16-26
pseudoinverses 9-6 random samples
pseudorandom numbers inverse Wishart 18-617
generating 6-2 latin hypercube 18-690
latin hypercube with normal
distribution 18-691
Q Wishart 18-1500
qqplot 18-1163 randsample 18-1179
QR decomposition 18-1204 randtool 18-1180
QRNG (quasi-random number generator) 6-17 range 18-1183
quality assurance 18-64 ranksum 18-1184
quantile 18-1165 raylcdf 18-1186
quantile-quantile plots 4-10 Rayleigh distribution B-91
quasi-random designs Rayleigh fit 18-1187
functions 16-49 raylfit 18-1187
Index-12
Index
raylinv 18-1188 relative efficiency 18-599
raylpdf 18-1189 resampling
raylrnd 18-1190 functions 16-9
raylstat 18-1191 statistics 3-9
rcoplot 18-1192 residuals
refcurve 18-1194 linear regression 9-5
refline 18-1198 regression 9-10
regress 18-1200 standardized 18-1204
regression studentized 18-1204
adjusted R-square statistic 18-1204 response surface
ANOVA 9-13 designs
change in covariance 18-1204 functions 16-48
change in fitted values 18-1204 response surfaces
coefficient covariance 18-1204 designs
coefficients 18-1204 Box-Behnken 14-13
delete-1 coefficients 18-1204 central composite 14-9
delete-1 variance 18-1204 overview 14-9
F distribution B-26 linear regression 9-46
F statistic 18-1204 methodology (RSM) 9-46
fitted values 18-1204 resubstitution error 18-940
hat matrix 18-1204 Rician distribution B-93
leverage 18-1204 ridge 18-1218
mean squared error 18-1204 ridge parameters 9-29 18-1219
multivariate 9-4 9-57 ridge regression 9-29 18-1218
partial least squares 9-33 ridge trace 18-1218
projection matrix 18-1204 RNGs. See random number generators
R-square statistic 18-1204 robust linear fit 18-1163
residuals 18-1204 robust linear regression 18-1230
scaled change in coefficients 18-1204 robust regression 9-14
scaled change in fitted values 18-1204 robustdemo 9-16 18-1226
t statistic 18-1204 robustfit 18-1230
regression analysis rotatable designs 14-11
functions 16-33 rotatefactors 18-1237
objects 17-4 rowexch 18-1241
visualization functions 16-13 16-33 RSM. See response surfaces — methodology
regression trees rsmdemo 18-1245
example 9-95 rstool 18-1250
functions 16-35 runs test 7-14
objects 17-4 runstest 18-1254
regstats 18-1204
Index-13
Index
S state sequences
sampsizepwr 18-1257 estimating 13-8
SBS. See sequential backward selection statistical arrays 2-11
scaling arrays statistical data 2-23
classical multidimensional 18-198 statistical functions
scatter operating on numerical data 2-9
visualization functions 16-12 vectorized 2-9
scatter plots statistical process control
functions that produce 4-3 capability studies 15-6
grouped 8-40 control charts 15-3
scatterhist 18-1261 functions 16-51
scree plots 10-41 visualization functions 16-14 16-51
sequential backward selection (SBS) 10-24 statistical visualization
sequential feature selection functions 16-11
criterion 10-23 stepwise 18-1334
sequential forward selection (SFS) 10-24 stepwise regression 9-19
sequentialfs 18-1270 stepwisefit 18-1339
SFS. See sequential forward selection structure arrays
shape storing heterogeneous data in 2-7
functions 16-9 Student’s t distribution B-95
Shepard plots 10-11 noncentral B-80
sign tests 7-14 sum of squares (SS) 18-13
significance levels 7-3 summaries
signrank 18-1286 functions 16-8
signtest 18-1288 supported distribution fitting
silhouette 18-1290 functions 16-24
similarity matrices surfht 18-1368
creating 11-4
single linkage 18-699 T
skewness 18-1302
t location-scale distribution B-97
slicesample 18-1299
t-tests
SPC. See statistical process control
one-sample 7-14
specific variance 10-46
paired-sample 7-14
squareform 18-1315
two-sample 7-14
SS. See sum of squares
tab-delimited data
standard normal 18-966
reading from file 18-1376
standardized data
tabular data
zscore 18-1507
reading from file 18-1370
standardized Euclidean distance 12-14 18-1050
tabulate 18-1369
18-1056
Index-14
Index
tblread 18-1370 unidpdf 18-1450
tblwrite 18-1372 unidrnd 18-1451
tcdf 18-1374 unidstat 18-1452
tdfread 18-1376 unifcdf 18-1453
terms unifinv 18-1454
linear model 9-3 unifit 18-1455
test data 12-2 uniform distribution B-99
test sequences uniformly distributed fit 18-1455
generating, for hidden Markov model 13-8 unifpdf 18-1456
test statistics 7-3 unifrnd 18-1457
tiedrank 18-1387 unifstat 18-1459
tinv 18-1389 utility functions 16-53
tpdf 18-1390
training data 12-2
V
transition matrices
estimating 13-9 variables
treatments container 2-2
experimental 14-3 grouping
treedisp 18-1405 functions for 2-35
treefit 18-1408 use for computing statistics 2-34
treeprune 18-1411 using 2-36
trees 18-1408 variances
See also decision trees of probability distribution 5-68
treetest 18-1416 vartest 18-1473
treeval 18-1420 vartest2 18-1475
trimmean 18-1422 vartestn 18-1477
trnd 18-1426 vectorization
tstat 18-1427 advantages of 2-9
ttest 18-1428
ttest2 18-1432 W
two-level designs 14-4
Wald distribution B-45
two-sample Kolmogorov-Smirnov test 7-13
ward linkage 18-700
two-way ANOVA 8-9
wblcdf 18-1485
type I errors 7-3
wblfit 18-1487
type II errors 7-3
wblinv 18-1489
wbllike 18-1492
U wblpdf 18-1494
unidcdf 18-1448 wblplot 18-1495
unidinv 18-1449 wblrnd 18-1497
Index-15
Index
wblstat 18-1499 feature selection 18-1272
Weibull distribution B-103
Weibull fit 18-1487
X
Weibull, Waloddi B-103
whiskers x2fx 18-1503
on plots 4-7 xptread 18-1502
Wilcoxon rank sum test 7-14
Wilcoxon signed rank tests 7-14 Z
Wishart distribution B-106
z-test, one-sample 7-15
Wishart random matrix 18-1500
zscore 18-1507
inverse 18-617
ztest 18-1509
wishrnd 18-1500
wrapper methods
Index-16
You might also like
- Show Me The Numbers by Stephen Few (2012) PDF80% (10)Show Me The Numbers by Stephen Few (2012) PDF373 pages
- The Effective Change Manager's Handbook - Essential Guidance To The Change Management Body of Knowledge PDF83% (36)The Effective Change Manager's Handbook - Essential Guidance To The Change Management Body of Knowledge PDF633 pages
- Storytelling With Data Cole Nussbaumer Knaflic100% (45)Storytelling With Data Cole Nussbaumer Knaflic291 pages
- The 100+ Business Models by FourWeekMBA - Full Library97% (35)The 100+ Business Models by FourWeekMBA - Full Library780 pages
- Tristan Boutros, Tim Purdie - The Process Improvement Handbook - A Blueprint For Managing Change and Increasing Organizational Performance-McGraw-Hill Education (2014) PDF100% (11)Tristan Boutros, Tim Purdie - The Process Improvement Handbook - A Blueprint For Managing Change and Increasing Organizational Performance-McGraw-Hill Education (2014) PDF333 pages
- Fundamentals of Business Intelligence (2015)89% (19)Fundamentals of Business Intelligence (2015)361 pages
- Max Mckeown The Strategy Book 2016 FT Press80% (10)Max Mckeown The Strategy Book 2016 FT Press264 pages
- Get Elementary Statistics A Step Step Approach 10th Edition Bluman Solutions Manual Free All Chapters Available100% (6)Get Elementary Statistics A Step Step Approach 10th Edition Bluman Solutions Manual Free All Chapters Available49 pages
- Strategic Thinking in Complex Problem Solving100% (16)Strategic Thinking in Complex Problem Solving300 pages
- Toyota Methods and Operating Models Achieve Business Success With The Toyota Way90% (10)Toyota Methods and Operating Models Achieve Business Success With The Toyota Way257 pages
- Berengueres J. Introduction To Data Visualization... Guide... 2020100% (4)Berengueres J. Introduction To Data Visualization... Guide... 2020189 pages
- Solving Problems With Design Thinking Ten Stories of What Works by Jeanne Liedtka, Andrew King, Kevin Bennett100% (7)Solving Problems With Design Thinking Ten Stories of What Works by Jeanne Liedtka, Andrew King, Kevin Bennett193 pages
- Avinash Dixit and Barry Nalebuff - Thinking Strategically100% (7)Avinash Dixit and Barry Nalebuff - Thinking Strategically400 pages
- Statistical Methods For Bioinformatics Lecture 5No ratings yetStatistical Methods For Bioinformatics Lecture 548 pages
- Advances in Industrial and Production Engineering 2021 PDFNo ratings yetAdvances in Industrial and Production Engineering 2021 PDF951 pages
- The Econometric Analysis of Transition Data - Tony LancasterNo ratings yetThe Econometric Analysis of Transition Data - Tony Lancaster374 pages
- Textbook6-Erik Cuevas - Alma Rodriguez - Metaheuristic Computation With MATLAB (R) - CRC Press (2020)No ratings yetTextbook6-Erik Cuevas - Alma Rodriguez - Metaheuristic Computation With MATLAB (R) - CRC Press (2020)281 pages
- 4.EURAMET Guide On Comparisons (Toolbox)No ratings yet4.EURAMET Guide On Comparisons (Toolbox)24 pages
- Gitlow H Melnyck R Levine D A Guide To Six Sigma and Process PDF100% (1)Gitlow H Melnyck R Levine D A Guide To Six Sigma and Process PDF590 pages
- Operational Excellence: Gilad Issar Liat Ramati Navon71% (7)Operational Excellence: Gilad Issar Liat Ramati Navon188 pages
- A First Course in Systems Biology 2nd Edition Eberhard Voit (Author) Ebook All Chapters PDF100% (3)A First Course in Systems Biology 2nd Edition Eberhard Voit (Author) Ebook All Chapters PDF62 pages
- PDF Scheduling in Industry 4 0 and Cloud Manufacturing Boris Sokolov Download100% (6)PDF Scheduling in Industry 4 0 and Cloud Manufacturing Boris Sokolov Download62 pages
- Bioinformatics and Machine Learning For Cancer Biology-5No ratings yetBioinformatics and Machine Learning For Cancer Biology-5198 pages
- Statistical Methods For Bioinformatics Lecture 2No ratings yetStatistical Methods For Bioinformatics Lecture 247 pages
- The Analysis of Time Series: An Introduction with R 7th Edition Chris Chatfield download pdf100% (5)The Analysis of Time Series: An Introduction with R 7th Edition Chris Chatfield download pdf55 pages
- Statistical Methods For Bioinformatics Lecture 3No ratings yetStatistical Methods For Bioinformatics Lecture 333 pages
- Brandes U., Erlebach T. Eds. Network Analysis. Methodological Foundations 2005ã. 482ñ. ISBN ISBN10 3-540-24979-6 PDF100% (1)Brandes U., Erlebach T. Eds. Network Analysis. Methodological Foundations 2005ã. 482ñ. ISBN ISBN10 3-540-24979-6 PDF482 pages
- Recommended Reading For Time Series AnalysisNo ratings yetRecommended Reading For Time Series Analysis2 pages
- Jim Duggan - Exploring Operations Research With R-CRC Pressr (2024)No ratings yetJim Duggan - Exploring Operations Research With R-CRC Pressr (2024)396 pages
- IACT 422 - 03 - Term Project - SUPPLY CHAIN SIMULATION FOR 4th PARTY LOGISTICS100% (1)IACT 422 - 03 - Term Project - SUPPLY CHAIN SIMULATION FOR 4th PARTY LOGISTICS37 pages
- Statistical Methods For Bioinformatics Lecture 4No ratings yetStatistical Methods For Bioinformatics Lecture 429 pages
- Machine Learning, Optimization, and Big Data 2016No ratings yetMachine Learning, Optimization, and Big Data 2016475 pages
- (Chapman & Hall Statistics Textbook Series) Derek Bissell (Auth.) - Statistical Methods For SPC and TQM-Springer US (1994) - 01No ratings yet(Chapman & Hall Statistics Textbook Series) Derek Bissell (Auth.) - Statistical Methods For SPC and TQM-Springer US (1994) - 015 pages
- Download Complete Machine Learning Pocket Reference Working with Structured Data in Python 1st Edition Matt Harrison PDF for All Chapters100% (3)Download Complete Machine Learning Pocket Reference Working with Structured Data in Python 1st Edition Matt Harrison PDF for All Chapters55 pages
- Solutions Manual to Accompany Introduction to Linear Regression Analysis instant downloadNo ratings yetSolutions Manual to Accompany Introduction to Linear Regression Analysis instant download30 pages
- Sequential Analysis Hypothesis Testing and Changepoint Detection ( Etc.) (Z-Library)No ratings yetSequential Analysis Hypothesis Testing and Changepoint Detection ( Etc.) (Z-Library)600 pages
- A Step by Step Backpropagation Example - Matt MazurNo ratings yetA Step by Step Backpropagation Example - Matt Mazur7 pages
- 13 Pag Design and Analysis of Experiments in The Health SciencesNo ratings yet13 Pag Design and Analysis of Experiments in The Health Sciences13 pages
- EEE1007 Neural Network and Fuzzy ControlNo ratings yetEEE1007 Neural Network and Fuzzy Control2 pages
- CONOVER Practical Nonparametric StatisticsNo ratings yetCONOVER Practical Nonparametric Statistics598 pages
- Applications of DOE in Engineering and Science 2019 Revised-LYE Sept 17 PDFNo ratings yetApplications of DOE in Engineering and Science 2019 Revised-LYE Sept 17 PDF214 pages
- Understanding Variation With Shewart ChartsNo ratings yetUnderstanding Variation With Shewart Charts48 pages
- (Image Processing Series) Luciano Da Fona Costa, Roberto Marcond Cesar Jr. - Shape Classification and Analysis - Theory and Practice-CRC Press (2009) PDFNo ratings yet(Image Processing Series) Luciano Da Fona Costa, Roberto Marcond Cesar Jr. - Shape Classification and Analysis - Theory and Practice-CRC Press (2009) PDF674 pages
- A Primer of Multivariate Statistics PDFNo ratings yetA Primer of Multivariate Statistics PDF626 pages
- Sampling Distribution and Simulation in RNo ratings yetSampling Distribution and Simulation in R10 pages
- Tutorial On Agent-Based Modeling and SimulationNo ratings yetTutorial On Agent-Based Modeling and Simulation14 pages
- Statistical Toolbox For Use With MATLABNo ratings yetStatistical Toolbox For Use With MATLAB420 pages
- Storytelling With Data - The New Visualization Data Guide To Reaching Your Business Aim in The Fastest Way100% (6)Storytelling With Data - The New Visualization Data Guide To Reaching Your Business Aim in The Fastest Way100 pages
- PSUnit III Lesson 2 Finding The Mean - and Variance of The Sampling Distribution of Means100% (1)PSUnit III Lesson 2 Finding The Mean - and Variance of The Sampling Distribution of Means21 pages
- Sta301 Finalterm Mcqs With Reference Solved by ArslanNo ratings yetSta301 Finalterm Mcqs With Reference Solved by Arslan45 pages
- Development of A Teacher Grit Scale (TGS) : Predicting The Performance of Educators in The PhilippinesNo ratings yetDevelopment of A Teacher Grit Scale (TGS) : Predicting The Performance of Educators in The Philippines13 pages
- TSBK08 Data Compression Exercises: Informationskodning, ISY, Link Opings Universitet, 2013No ratings yetTSBK08 Data Compression Exercises: Informationskodning, ISY, Link Opings Universitet, 201332 pages
- Session 6 - CSD102 Measures of Divergence From Normality100% (1)Session 6 - CSD102 Measures of Divergence From Normality30 pages
- Regression Analysis and Multiple Regression: Session 7No ratings yetRegression Analysis and Multiple Regression: Session 7100 pages
- Full download Linear and Generalized Linear Mixed Models and Their Applications 2nd Edition Jiming Jiang pdf docx100% (3)Full download Linear and Generalized Linear Mixed Models and Their Applications 2nd Edition Jiming Jiang pdf docx65 pages
- A Critical View On ISO Standard 13528: Wim Coucke EQALM Symposium, Dublin, 20th of October 2017No ratings yetA Critical View On ISO Standard 13528: Wim Coucke EQALM Symposium, Dublin, 20th of October 201722 pages
- A Beginner’s Guide to Variational InferenceNo ratings yetA Beginner’s Guide to Variational Inference48 pages
- Water level prediction using various machine learning algorithms a case study of Durian Tunggal river MalaysiaNo ratings yetWater level prediction using various machine learning algorithms a case study of Durian Tunggal river Malaysia20 pages
- speech processing theory and applicationsNo ratings yetspeech processing theory and applications92 pages