Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Linear Algebra with RcppArmadillo

Advanced Statistical Programming Camp
Jonathan Olmsted (Q-APS)

Day 4: May 30th, 2014

AM Session

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

1 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Outline

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

2 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Why Linear Algebra?

Linear algebra is pervasive in statistical computations:
multiplication
inversion
decompositions

Previous data structures (C++ types and Rcpp classes) do not

provide us with the tools to avoid element-wise implementations of

these algorithms.
Programming these methods by hand would be tedious and

error-prone.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

3 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Linear Algebra by hand

Inner product.
// [[ Rcpp :: export ()]]

double inner1 (NumericVector x,

NumericVector y
) {
int K = x.length() ;
double ip = 0 ;
for (int k = 0 ; k < K ; k++) {
ip += x(k) * y(k) ;
}
return(ip) ;
}

Vectors require 1 loop. Matrix operations would require 2 loops.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

4 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Linear Algebra by Linear Algebra Library

// [[ Rcpp :: export ()]]

double inner2 (arma::vec x,

arma::vec y
) {
arma::mat ip = x.t() * y ;
return(ip(0)) ;
}

Linear algebra structures allow code to resemble standard

mathematical notation.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

5 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

library("Rcpp")
sourceCpp("inner.cpp")
vec <- rnorm(1000)
inner1(vec, vec)
## [1] 983.4
inner2(vec, vec)
## [1] 983.4

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

6 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Optimized Libraries
Consider a simulated dataset from the omitted variable bias example
data generating process.
source("gendata.R")
sourceCpp("lmA.cpp")
dfFake <- genNormData(1000)
m1 <- lm(y ~

x1 + x2 + x3, data = dfFake)

X <- model.matrix(m1)
y <- matrix(dfFake$y)

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

7 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Optimized Libraries
library(microbenchmark)
microbenchmark(lm = lm(y ~ x1 + x2 + x3, data = dfFake),
R = solve(t(X) %*% X) %*% t(X) %*% y,
cpp = lmA(X, y),
times = 10
)
## Unit: microseconds
## expr
min
lq median
uq
max neval
##
lm 2239.54 2338.80 2445.78 2500.3 3335.9
10
##
R 184.46 249.52 411.77 432.0 771.4
10
##
cpp
18.27
21.14
44.71
53.2 108.9
10

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

8 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Outline

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

9 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Armadillo
Armadillo (http://arma.sourceforge.net/) is a popular,

optimized linear algebra library for C++.

tuned algorithms
rich functionality
flexible
relatively easy to use

But, hooking in to external C++ libraries isnt always easy.

Primary classes of interest arma::mat and arma::vec (elements

are double).
Additional classes include sparse matrices and

higher-dimensional arrays.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

10 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

RcppArmadillo
RcppArmadillo (http://cran.r-project.org/web/packages/

RcppArmadillo/index.html) provides all the functionality of

Armadillo
Easy to install because it is distributed as an R package
Tightly integrated with Rcpp:
Conversion from Rcpp classes to Armadillo classes (and back).
Conversion from Armadillo classes to R objects (and back).

Previously . . .
# include <Rcpp.h>

Now . . .
# include <RcppArmadillo.h>
// [[ Rcpp :: depends ( RcppArmadillo )]]

The # include <Rcpp.h> is implied.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

11 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Outline

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

12 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

// [[ Rcpp :: export ()]]

arma::mat a1 (arma::mat x) {
return(x) ;
}

Matrix in, matrix out.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

13 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

sourceCpp("arma_functions.cpp")
a1(diag(2))
##
[,1] [,2]
## [1,]
1
0
## [2,]
0
1

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

14 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

// [[ Rcpp :: export ()]]

arma::vec a2 (arma::vec x) {
return(x) ;
}

Vector in, matrix out.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

15 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

a2(1:2)
##
[,1]
## [1,]
1
## [2,]
2

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

16 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

// [[ Rcpp :: export ()]]

arma::mat a3 (NumericMatrix x) {
arma::mat y = as<arma::mat>(x) ;
return(y) ;
}

From Rcpp::NumericMatrix to arma::mat.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

17 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

a3(diag(2))
##
[,1] [,2]
## [1,]
1
0
## [2,]
0
1

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

18 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

// [[ Rcpp :: export ()]]

NumericMatrix a4 (arma::mat x) {
NumericMatrix y = wrap(x) ;
return(y) ;
}

From arma::mat to Rcpp::NumericMatrix.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

19 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Automatic Handling of Data Types

a4(diag(2))
##
[,1] [,2]
## [1,]
1
0
## [2,]
0
1

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

20 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Dimensions in Armadillo

// [[ Rcpp :: export ()]]

arma::mat a5 (arma::mat x) {
int R = x.n_rows ;
int C = x.n_cols ;
Rcout << "Rows: " << R << std::endl ;
Rcout << "Cols: " << C << std::endl ;
return(x) ;
}

Note the absence of parenthesis (unlike Rcpp classes).

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

21 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Dimensions in Armadillo

a5(matrix(0, nrow = 2, ncol = 3))

##
##
##
##
##

Rows: 2
Cols: 3
[,1] [,2] [,3]
[1,]
0
0
0
[2,]
0
0
0

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

22 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Dimensions in Armadillo

// [[ Rcpp :: export ()]]

arma::vec a6 (arma::vec x) {
int R = x.n_rows ;
int C = x.n_cols ;
Rcout << "Rows: " << R << std::endl ;
Rcout << "Cols: " << C << std::endl ;
return(x) ;
}

arma::vec objects are just matrices.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

23 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Dimensions in Armadillo

a6(1:4)
##
##
##
##
##
##
##

Rows: 4
Cols: 1
[,1]
[1,]
1
[2,]
2
[3,]
3
[4,]
4

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

24 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Inspection of Armadillo Objects

// [[ Rcpp :: export ()]]

int a7 (arma::mat x) {
x.print() ;
x.print("Note") ;
return(0) ;
}

More useful than Rcout. No need to loop through elements.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

25 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Inspection of Armadillo Objects

a7(diag(2))
##
1.0000
##
0
## Note
##
1.0000
##
0
## [1] 0

ASPC

0
1.0000
0
1.0000

Linear Algebra with RcppArmadillo

Day 4 AM

26 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Manipulating Armadillo Objects

// [[Rcpp::export()]]
List a8 (int n, int r, double v) {
arma::mat x1 ;
x1.print() ;
x1.reshape(n, r) ;
x1.fill(v) ;
arma::mat x2(n, r) ;
x2.fill(v) ;
arma::mat x3 = x2 ;
x3.reshape(r, n) ;
List ret ;
ret["x1"] =
ret["x2"] =
ret["x3"] =
return(ret)

x1 ;
x2 ;
x3 ;
;

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

27 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Inspection of Armadillo Objects

a8(2, 3, 3/4)
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##

[matrix size: 0x0]

$x1
[,1] [,2] [,3]
[1,] 0.75 0.75 0.75
[2,] 0.75 0.75 0.75
$x2
[,1] [,2] [,3]
[1,] 0.75 0.75 0.75
[2,] 0.75 0.75 0.75
$x3
[,1] [,2]
[1,] 0.75 0.75
[2,] 0.75 0.75
[3,] 0.75 0.75

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

28 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Manipulating Armadillo Objects

// [[ Rcpp :: export ()]]

arma::mat a9 (int n, int r) {

arma::mat x(n, r) ;
x.print("") ;
x.ones() ;
return(x) ;
}

Uninitialized values are unreliable.

Populate with ones or zeros (.zeros).

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

29 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Manipulating Armadillo Objects

a9(2, 3)
##
6.9442e-310 6.9442e-310
##
2.2225e-314 2.2217e-314
##
[,1] [,2] [,3]
## [1,]
1
1
1
## [2,]
1
1
1

ASPC

6.9442e-310
2.2217e-314

Linear Algebra with RcppArmadillo

Day 4 AM

30 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

// [[ Rcpp :: export ()]]

double a10 (arma::mat x, int i, int j) {

return(x(i, j)) ;
}

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

31 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

Z <- matrix(rnorm(6), 2, 3)
Z
##
[,1]
[,2]
[,3]
## [1,] -0.6580 1.98935 -0.02781
## [2,] -0.1626 -0.06105 0.51153
a10(Z, 1, 2)
## [1] 0.5115

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

32 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

// [[ Rcpp :: export ()]]

arma::mat a11 (arma::mat x, int i) {

return(x.row(i)) ;
}

Index entire rows.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

33 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

a11(Z, 1)
##
[,1]
[,2]
[,3]
## [1,] -0.1626 -0.06105 0.5115

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

34 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

// [[ Rcpp :: export ()]]

arma::mat a12 (arma::mat x, int j) {

return(x.col(j)) ;
}

Index entire columns.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

35 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

a12(Z, 2)
##
[,1]
## [1,] -0.02781
## [2,] 0.51153

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

36 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

// [[ Rcpp :: export ()]]

arma::mat a13 (arma::mat x) {

return(x.cols(0, 1)) ;
}

Index multiple rows or columns at once.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

37 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Indexing Armadillo Objects

a13(Z)
##
[,1]
[,2]
## [1,] -0.6580 1.98935
## [2,] -0.1626 -0.06105

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

38 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a14(arma::mat x) {
return(x + x) ;
}

Element-wise addition.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

39 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a14(Z)
##
[,1]
[,2]
[,3]
## [1,] -1.3160 3.9787 -0.05562
## [2,] -0.3252 -0.1221 1.02305

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

40 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a15(arma::mat x) {
return(x - x) ;
}

Element-wise subtraction.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

41 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a15(Z)
##
[,1] [,2] [,3]
## [1,]
0
0
0
## [2,]
0
0
0

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

42 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a16(arma::mat x) {
return(x % x) ;
}

Element-wise multiplication.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

43 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a16(Z)
##
[,1]
[,2]
[,3]
## [1,] 0.43299 3.957510 0.0007734
## [2,] 0.02644 0.003727 0.2616587

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

44 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a17(arma::mat x) {
return( exp(x) ) ;
}

Element-wise applications of functions.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

45 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a17(Z)
##
[,1]
[,2]
[,3]
## [1,] 0.5179 7.3108 0.9726
## [2,] 0.8499 0.9408 1.6678

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

46 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a18(arma::mat x) {
return(x.t()) ;
}

Transpose.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

47 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a18(Z)
##
[,1]
[,2]
## [1,] -0.65802 -0.16262
## [2,] 1.98935 -0.06105
## [3,] -0.02781 0.51153

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

48 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a19(arma::mat x) {
return( x.t() * x ) ;
}

Matrix multiplication.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

49 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a19(Z)
##
[,1]
[,2]
[,3]
## [1,] 0.45943 -1.29910 -0.06488
## [2,] -1.29910 3.96124 -0.08655
## [3,] -0.06488 -0.08655 0.26243

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

50 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a20(arma::mat x) {
return( inv(x * x.t()) ) ;
}

Matrix inversion.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

51 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a20(Z)
##
[,1]
[,2]
## [1,] 0.22787 0.02239
## [2,] 0.02239 3.42885

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

52 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

arma::mat a21(arma::mat x) {
return( chol(x * x.t()) ) ;
}

Decompositions.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

53 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a21(Z)
##
[,1]
[,2]
## [1,] 2.096 -0.01368
## [2,] 0.000 0.54004

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

54 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

// [[ Rcpp :: export ()]]

List a22(arma::mat x) {
arma::mat xtx = x.t() * x ;
arma::mat U ;
arma::vec s ;
arma::mat V ;
svd(U, s, V, xtx) ;
List ret
ret["U"]
ret["s"]
ret["V"]

;
= U ;
= s ;
= V ;

return(ret) ;
}
ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

55 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Operators and Functions

a22(Z)
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##
##

$U
[,1]
[,2]
[,3]
[1,] -0.31345 -0.30964 0.8977
[2,] 0.94949 -0.08729 0.3014
[3,] -0.01498 0.94684 0.3214
$s
[,1]
[1,] 4.391e+00
[2,] 2.916e-01
[3,] 1.082e-16
$V
[,1]
[,2]
[,3]
[1,] -0.31345 -0.30964 0.8977
[2,] 0.94949 -0.08729 0.3014
[3,] -0.01498 0.94684 0.3214

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

56 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Informative Runtime Error Messages

// [[ Rcpp :: export ()]]

double a23 (int i) {

arma::mat A(2, 3) ;
A.randu() ;
return( A(i, i) ) ;
}

This indexing is valid for some i.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

57 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Informative Runtime Error Messages

a23(1)
## [1] 0.8533
a23(4)
##
## error: Mat::operator(): index out of bounds
## Error:

ASPC

Mat::operator():

index out of bounds

Linear Algebra with RcppArmadillo

Day 4 AM

58 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Informative Runtime Error Messages

// [[ Rcpp :: export ()]]

arma::mat a24 (int i) {

arma::mat A(2, 3) ;
arma::mat B(i, 1) ;
A.randu() ;
B.randu() ;
return( A * B ) ;
}

This multiplication is valid for only i = 3.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

59 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Informative Runtime Error Messages

a24(3)
##
[,1]
## [1,] 0.4896
## [2,] 1.1367
a24(2)

##
## error: matrix multiplication: incompatible matrix dimensions
## Error: matrix multiplication:
dimensions: 2x3 and 2x1

ASPC

incompatible matrix

Linear Algebra with RcppArmadillo

Day 4 AM

60 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Outline

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

61 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Big Linear Regressions

dfTrade <- read.csv("trade.csv", nrows = 4e5) # just first 500k

dfTrade2 <- subset(dfTrade, source1 != -9 & source2 != -9) # drop missing
nrow(dfTrade2)
## [1] 311763
dfTrade2$total <- with(dfTrade2, flow1 + flow2)
dfTrade2$dem <- with(dfTrade2, dem1 + dem2)
dfTrade2$maj <- with(dfTrade2, majpow1 + majpow2)
f <- formula(total ~ dem + poly(year, 3) + factor(abbrev1) + factor(abbrev2))

Total trade flows between country pairs from 1950 through 2008. Over
300 fixed-effects.

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

62 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Big Linear Regression

// [[ Rcpp :: export ()]]

arma::mat lmA (arma::mat X,

arma::mat y
) {
arma::mat betahat ;
betahat = (X.t() * X ).i() * X.t() * y ;
return(betahat) ;
}

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

63 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

Big Linear Regressions

system.time({
fit <- lm(f, data = dfTrade2)
})
##
##

user
36.79

system elapsed
0.77
37.57

X <- model.matrix(fit)
system.time({
fit2 <- lmA(model.matrix(fit), matrix(dfTrade2$total))
})
##
##

user
6.786

system elapsed
0.605
2.439

Dont try this at home (or on Adroit)

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

64 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

EM Probit Regression

EM algorithm augments censored data with expectation of the

latent (censored) variable.

Conditional on latent varible: linear regression.

sourceCpp("em_probit.cpp")
library(Zelig)
data(turnout)

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

65 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

EM Probit Regression
arma::mat em_probit (arma::mat y,
arma::mat X,
int maxit = 10
) {
int N = y.n_rows ; int K = X.n_cols ;
arma::mat beta(K, 1) ;
beta.fill(0.0) ; // initialize betas to 0
arma::mat eystar(N, 1) ;
eystar.fill(0) ;
for (int it = 0 ; it < maxit ; it++) {
arma::mat mu = X * beta ;
for (int n = 0 ; n < N ; n++) {
if (y(n, 0) == 1) { // y = 1
eystar(n, 0) = mu(n, 0) + f(mu(n, 0)) ; // defined elsewhere
}
if (y(n, 0) == 0) { // y = 0
eystar(n, 0) = mu(n, 0) - g(mu(n, 0)) ; // defined elsewhere
}
}
beta = (X.t() * X).i() * X.t() * eystar ; // linear regression
}
return(beta) ;
}
ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

66 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

EM Probit Regression
system.time({
fit3 <- glm(vote ~ income + educate + age,
data = turnout,
family = binomial(link = "probit")
)})
##
##

user
0.019

system elapsed
0.003
0.021

system.time({
fit4 <- em_probit(y = matrix(turnout$vote),
X = model.matrix(fit3),
maxit = 25
)})
##
##

user
0.005
ASPC

system elapsed
0.000
0.005
Linear Algebra with RcppArmadillo

Day 4 AM

67 / 68

Motivation

Armadillo and RcppArmadillo

Armadillo Basics

Applications

EM Probit Regression

coef(fit3)
## (Intercept)
##
-1.68241

income
0.09936

educate
0.10667

age
0.01692

fit4
##
##
##
##
##

[,1]
[1,] -1.68214
[2,] 0.09933
[3,] 0.10666
[4,] 0.01691

ASPC

Linear Algebra with RcppArmadillo

Day 4 AM

68 / 68