Quantitative Data Analysis

8
Introduction

A small research project

Statistical
significance and Multivariate
Missing data Types of variables Univariate analysis Bivariate analysis
inferential analysis
statistics
Chapter Objectives
By the end of this chapter, students should be able
to:
 understand the importance of anticipating
questions of analysis early in the research process
 distinguish between the different kinds of
variables generated in qualitative research
 understand methods for analyzing a single
variable (univariate analysis)
 understand methods for analyzing relationships
between two variable (bivariate analysis)
 analyze relationships among three or more
This Photo by Unknown Author is licensed under CC BY-SA
variable (multivariate analysis)

Chapter 8
2
 • Data analysis is the final stage of the research
Introduction process.

to Data • It brings the theoretical ideas formulated at the

beginning of the study together with the evidence
Analysis (the data) we collected in the process of research
 The big­gest mistake in quantitative research is to
think that data analysis decisions can wait until
after the data have been collected
 Data analysis is carried out after that stage, but
it’s essential to be fully aware of what techniques
will be used before data collection begins
• Questionnaire, observation schedule, and coding
frame should be designed with the data analysis in
mind
 The statistical techniques that can be used
depend on how a variable is measured
 Inappropriate measurement may make it
impossible to conduct certain types of data

Chapter 8
analysis
 The size and nature of the sample also imposes
limitations on the kinds of techniques that are
suitable for the data set 3
 How Does Analysis Work?
• The analysis of numerical data is built on two basic dimensions:
 The type of variables (levels of measurement):
 Nominal
 Ordinal
 Interval/Ratio
 The number of variables in the analysis = Descriptive Statistics
 Univariate (one variable)
 Bivariate (two variables)
 Multivariate (three or more variables)

Missing Data
• How do we handle it when a respondent does not complete an answer?

• Is it missing because

Chapter 8
 they accidentally skip it?
 they do not want to answer it?
 it does not apply to them?

• Use codes to classify missing data (e.g., 999). 4

 Types of Variables

Values are grouped into Nominal level variables with a also called continuous level
Nominal

Ordinal

Interval/
Ratio
categories that have no meaningful order variables - have the most
meaningful order. The categories of the variable detail associated with them
can be rank ordered Distance or amount of
The only difference that exists For example, educational difference between categories
between participants is being level, age group is uniform (e.g., 0 siblings, 1
in one category or another sibling, 2 siblings, etc.)
Distance between categories
Categories cannot be ordered may not be equal Can do arithmetic and
by rank mathematical operations with
Cannot do arithmetic or the categories (e.g., 1 sibling
Cannot do arithmetic or mathematical operations with + 3 siblings = 4 siblings)
mathematical operations with the categories, except
the categories comparison Ratio variables have a real “0”
start position
e,.g. Frequency or e.g. described with • (e.g., age, salary, weight, height
percentages frequencies percentages and cannot go below 0)
nominal variables can be used non-parametric stats • For ratio variables can compute
meaningful ratios: 20years/10year
to do cross tabulations. means, standard deviations, means that the first person is twice
The chi-square test can be and parametric statistical older than the second one
performed on the cross- tests are NOT appropriate • A ratio variable can be used as a

Chapter 8
dependent variable for most
tabulation parametric statistical tests (t-tests, F-
tests, correlation, and regression)

5
 Measures of
Central Tendency
• Mode
Level of Measures of Central  The score that shows up the most in a
particular category
involvemen Tendency  Can be used with all variable types
t  Most applicable to nominal data

Mean Median Mode • Median

 The middle score when all scores have been
arrayed in order (if even number of scores it is
Nominal Yes the mean of two middle scores)
 Can be used with ordinal or interval/ratio data
Ordinal Yes Yes
• Mean
Interval/Ratio Yes Yes Yes  The sum of all scores, divided by the number of
scores
 Can be used with interval/ratio data
 Vulnerable to outliers (extreme scores)

Chapter 8
6
 Measure of Dispersion
• The amount of variation in a sample/deviation from the mean
• The three most frequently used measures of variability are:

Range

• Highest score minus lowest score

• Shows the influence of outliers
• measures of variability for skewed data

Standard deviation

• The standard deviation is a measure of the average difference of each observation in a

distribution from the average (mean) of the distribution
• Measures the amount of variation around the mean
• Measures of variability for symmetrical data

Variance

Chapter 8
• The variance is an expression of the total amount of variability of the observations for a
variable.
• The value of the variance is obtained by squaring the value of the standard deviation
• measures of variability for symmetrical data

7
Univariate Analysis
• Analysis of one variable at a time
• Often, the first step in the analysis is to create
frequency tables for the variables of interest
 Frequency distribution tables show the
number of times a particular variable shows
up in the distribution, expressed as an actual
number and as percentage of the whole
 When interval/ratio variables are shown in
frequency tables, categories may be
combined as long as they don’t overlap (e.g.
age groups of 20–29, 30–39, …)

• Diagrams can be used to illustrate frequency

distributions
• Use bar charts and pie charts for displaying a
nominal or ordinal variable

Chapter 8
• Use histograms for an interval/ratio variable

8
Univariate Analysis: Frequency Tables

Chapter 8
9
 Bivariate Analysis
Determines whether there is a relationship between two variables –
Establishing the existence of a relationship is not proof of causality

Allow Normally used Shows correlation Shows correlation Shows the Used with an
simultaneous with interval/ratio between pairs of between pairs of strength of the interval/ratio
analysis of two data ordinal variables, ordinal variables relationship variable and a
variables Values from 0 or with one Like Pearson’s r, between two nominal variable
Identify patterns (indicates no ordinal and one values range nominal variables • Nominal variable is
interval/ratio the independent
Pearson’s r

Cramér’s V

and eta
(cross-tabulations)

Spearman’s rho
Contingency tables

Kendall’s tau-b

Comparing means
of association relationship) from 0 to +1 Values range variable
variable from 0 to 1
Can be used for to +1 (indicates Will predict a • Compare means of
any variable type perfect positive Like Pearson’s r, rank position (Nominal interval variable for
relationship) values range from one variable categories cannot
each subgroup of the
Normally used for from 0 to +1 nominal variable
nominal or or -1 (indicates to another be rank ordered) • Determines level of
association between
ordinal data perfect negative Usually reported the two variables
Note: The relationship) with a Values range
independent The relationship contingency table from 0 to 1
variable is between the and a chi-square
test (Nominal
normally variables should categories cannot
displayed as the be approximately be rank ordered)
column variable linear if

Chapter 8
Pearson’s r is to
be used in a
study
• This can be
established using a
scatter plot

10
Bivariate Analysis –
Contingency table
• Contingency tables (cross-
tabulations)
 Allow simultaneous analysis of
two variables
 Identify patterns of association
 Can be used for any variable
type
 Normally used for nominal or
ordinal data

• Note: The independent

Chapter 8
variable is normally
displayed as the column
variable
11
 Measure of Dispersion and
Association
• Amount of Explained Variance
 Squaring eta, Kendall’s tau-b, Spearman’s rho, Pearson’s r: show how
much the variation in one variable will explain variation in the other
variable
 Allows prediction of the second variable based on the score from the first
 R2 shows explained variance in percentages
 For example, years of education explain 25% of variation in income (R2 =0.25)

 Measures of Association
Measure Greek Type of Data High Association
Lambda Nominal 1.0
Gamma Ordinal +1.0, -1.0
Tau (Kendall’s) Ordinal +1.0, -1.0

Chapter 8
Rho Interval/ratio +1.0, -1.0
Chi-square Nominal/ordinal Infinity

12
 Bivariate Analysis, cont’d
• Amount of Explained Variance
 Squaring eta, Kendall’s tau-b, Spearman’s rho,
Pearson’s r: show how much the variation in one
variable will explain variation in the other variable
 Allows prediction of the second variable based on the
score from the first
 R2 shows explained variance in percentages
 For example, years of education explain 25% of variation in
income (R2 =0.25)

Chapter 8
13
 Statistical Significance
• Can a sample finding be used to estimate a characteristic of the
whole population?
• Stated as a probability level
 Significance shows the probability that the results are not due to chance

• A null hypothesis tests the significance of the bivariate

association
 (e.g., State that there is no relationship between two variables, or that two
populations do not differ on some characteristic)

• To test for statistical significance

 Set up a null hypothesis
 Establish an acceptable level of significance (p-value)
 It must be .05 or lower (< .05)

Chapter 8
 If the null is correct there is no relationship
 If the null is rejected and the statistical significance (p) of the findings is at
< .05 level, there is indirect support for the research hypothesis
 It is unlikely that the results occurred by chance 14
 Errors and statistical significance
• Two types of errors
 Type I: rejecting a true null hypothesis
 The results are a chance association
 Type II: not rejecting a false null hypothesis

• The two types of errors act in inverse relationship to each other

and cannot be minimized at the same time
 If one is low the other is high

• Researchers usually choose to minimize the Type I error over

the Type II
True Situation in the World
What the Researcher No Relationship Causal Relationship
Says

Chapter 8
No relationship No error Type II error
Causal relationship Type 1 error No error

15
 Tests and Statistical Significance
• Correlation and statistical significance
 The significance of a Pearson’s r and a Kendall’s tau-b correlation coefficient is determined
by
 the size of the coefficient
 the sample size
 Correlation and statistical significance must be weighed together

• Chi-square (χ2) and statistical significance

 Used with contingency tables
 Measures the likelihood that a relationship between the two variables exists in the
population
 Calculated by comparing the observed frequency in each cell with what would be expected
by chance (if there were no relationship between the variables)
 The chi-square value is affected by the sample size

• Comparing means and statistical significance

 Analysis of variance (F statistic)

Chapter 8
 Total amount of variation in the dependent variable
 Divided into explained variance (variance between groups) and error variance (variance within
groups)
 Establishes whether the difference between groups is significant
 Reported as a statistically significant probability (p) 16
 Multivariate Analysis
• Examines the relationship between three or more variables, also
referred to as “elaboration”
• Can be used to test for spuriousness
 Spuriousness (inaccuracy) exists if two variables are correlated but only
through a third variable
 In a spurious relationship, an antecedent third variable is producing the
variation in the two variables of interest

Chapter 8
17
 Intervening variable and interactions
Intervening Variable
• Multivariate analysis can be used to test for intervening variables
 X  Y (intervening variable?)  Z = (Score-Mean)/Standard Deviation
 e.g., Education  Income  Happiness

• If the possible intervening variable is controlled, and the relationship

between X and Z disappears, then Y is considered an intervening variable
Interactions
• Can be used to test for interactions
• An interaction exists if the effect of one independent variable varies at
different levels to that of a second independent variable
• The two vary independent of each other

Chapter 8
 There is mild connection

• One independent variable moderates (has some impact on) the relationship
between the other independent variable and its dependent variable
 For example, the effect of age on having another source of exercise is different for men 18
Multivariate Analysis:
Linear Regression
• Is used in multiple linear regression
• Multiple linear regression can
determine the following:
 how much of the variation in the
dependent variable is explained
(predicted) by the independent
variables
 which, if any, of the independent
variables is a significant predictor
of the dependent variable

• First consider bivariate regression

 Bivariate regression involves one
independent variable and one

Chapter 8
dependent variable
 This can be calculated using SPSS

19
 Summary of Major Types of
Descriptive Stats
Type of Statistical Technique Purpose Result Display
Technique
Univariate • Frequency Describe 1 variable Frequency table
distribution
• Measure of central
tendency
• Standard Deviation
• Z score
• Charts
Bivariate • Correlation Describe a Cross tabulation
• Percentage table relationship or the
• Chi-square association between 2
• Analysis of Variance variables
(F stat)
Multivariate • Elaboration Describe relations Multi-regression
paradigm among several

Chapter 8
• Multiple regression variables, or see how
several independent
variables have an
effect on a dependent
variable 20
• Data Analysis lab
• Quiz 4 – chapters 7 and 8

This Photo by Unknown Author is licensed under

CC BY-SA

Chapter 8
21