Practical Guide to Principal Component Methods in R

Alboukadel KASSAMBARA
ii

Copyright ©2017 by Alboukadel Kassambara. All rights reserved.

Published by STHDA (http://www.sthda.com), Alboukadel Kassambara

Contact: Alboukadel Kassambara <[email protected]>

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form
or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, without the prior
written permission of the Publisher. Requests to the Publisher for permission should
be addressed to STHDA (http://www.sthda.com).

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in
preparing this book, they make no representations or warranties with respect to the accuracy or
completeness of the contents of this book and specifically disclaim any implied warranties of
merchantability or fitness for a particular purpose. No warranty may be created or extended by sales
representatives or written sales materials.

Neither the Publisher nor the authors, contributors, or editors,

assume any liability for any injury and/or damage
to persons or property as a matter of products liability,
negligence or otherwise, or from any use or operation of any
methods, products, instructions, or ideas contained in the material herein.

For general information contact Alboukadel Kassambara <[email protected]>.

Contents

0.1 What you will learn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

0.2 Key features of this book . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
0.3 How this book is organized . . . . . . . . . . . . . . . . . . . . . . . . . . vii
0.4 Book website . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
0.5 Executing the R codes from the PDF . . . . . . . . . . . . . . . . . . . . xii
0.6 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
0.7 Colophon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

About the author xiv

I Basics 1
1 Introduction to R 2
1.1 Installing R and RStudio . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Installing and loading R packages . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Getting help with functions in R . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Importing your data into R . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Demo data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.6 Close your R/RStudio session . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Required R packages 6
2.1 FactoMineR & factoextra . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Main R functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

II Classical Methods 11
3 Principal Component Analysis 12
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 Visualization and Interpretation . . . . . . . . . . . . . . . . . . . . . . . 17
3.5 Supplementary elements . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6 Filtering results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.7 Exporting results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.9 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

iii
iv CONTENTS

4 Correspondence Analysis 51
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Visualization and interpretation . . . . . . . . . . . . . . . . . . . . . . . 55
4.4 Supplementary elements . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.5 Filtering results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.6 Outliers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.7 Exporting results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.9 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5 Multiple Correspondence Analysis 83

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3 Visualization and interpretation . . . . . . . . . . . . . . . . . . . . . . . 86
5.4 Supplementary elements . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.5 Filtering results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.6 Exporting results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.8 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

III Advanced Methods 107

6 Factor Analysis of Mixed Data 108
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3 Visualization and interpretation . . . . . . . . . . . . . . . . . . . . . . . 110
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.5 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

7 Multiple Factor Analysis 120

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
7.3 Visualization and interpretation . . . . . . . . . . . . . . . . . . . . . . . 125
7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.5 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

IV Clustering 141
8 HCPC: Hierarchical Clustering on Principal Components 142
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.2 Why HCPC? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.3 Algorithm of the HCPC method . . . . . . . . . . . . . . . . . . . . . . . 143
8.4 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.6 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
Preface

0.1 What you will learn

Large data sets containing multiple samples and variables are collected everyday by re-
searchers in various fields, such as in Bio-medical, marketing, and geo-spatial fields.
Discovering knowledge from these data requires specific techniques for analyzing data
sets containing multiple variables. Multivariate analysis (MVA) refers to a set of
techniques used for analyzing a data set containing more than one variable.
Among these techniques, there are:
• Cluster analysis for identifying groups of observations with similar profile according
to a specific criteria.
• Principal component methods, which consist of summarizing and visualizing the
most important information contained in a multivariate data set.
Previously, we published a book entitled “Practical Guide To Cluster Analysis in
R” (https://goo.gl/DmJ5y5). The aim of the current book is to provide a solid
practical guidance to principal component methods in R. Additionally, we developed
an R package named factoextra to create, easily, a ggplot2-based elegant plots of the
results of principal component method. Factoextra official online documentation:
http://www.sthda.com/english/rpkgs/factoextra

One of the difficulties inherent in multivariate analysis is the problem of visualizing data
that has many variables. In R, there are many functions and packages for displaying
a graph of the relationship between two variables (http://www.sthda.com/english/
wiki/data-visualization). There are also commands for displaying different three-
dimensional views. But when there are more than three variables, it is more difficult to
visualize their relationships.
Fortunately, in data sets with many variables, some variables are often correlated. This
can be explained by the fact that, more than one variable might be measuring the same
driving principle governing the behavior of the system. Correlation indicates that there is
redundancy in the data. When this happens, you can simplify the problem by replacing
a group of correlated variables with a single new variable.
Principal component analysis is a rigorous statistical method used for achieving this sim-
plification. The method creates 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.

v
vi CONTENTS

The type of principal component methods to use depends on variable types contained in
the data set. This practical guide will describe the following methods:
1. Principal Component Analysis (PCA), which is one of the most popular mul-
tivariate analysis method. The goal of PCA is to summarize the information con-
tained in a continuous (i.e, quantitative) multivariate data by reducing the dimen-
sionality of the data without loosing important information.
2. Correspondence Analysis (CA), which is an extension of the principal com-
ponent analysis for analyzing a large contingency table formed by two qualitative
variables (or categorical data).
3. Multiple Correspondence Analysis (MCA), which is an adaptation of CA to
a data table containing more than two categorical variables.
4. Factor Analysis of Mixed Data (FAMD), dedicated to analyze a data set
containing both quantitative and qualitative variables.
5. Multiple Factor Analysis (MFA), dedicated to analyze data sets, in which
variables are organized into groups (qualitative and/or quantitative variables).
Additionally, we’ll discuss the HCPC (Hierarchical Clustering on Principal Com-
ponent) method. It applies agglomerative hierarchical clustering on the results of prin-
cipal component methods (PCA, CA, MCA, FAMD, MFA). It allows us, for example, to
perform clustering analysis on any type of data (quantitative, qualitative or mixed data).
Figure 1 illustrates the type of analysis to be performed depending on the type of variables
contained in the data set.

0.2 Key features of this book

Although there are several good books on principal component methods and related
topics, we felt that many of them are either too theoretical or too advanced.
Our goal was to write a practical guide to multivariate analysis, visualization and inter-
pretation, focusing on principal component methods.
The book presents the basic principles of the different methods and provide many exam-
ples in R. This book offers solid guidance in data mining for students and researchers.
Key features
• Covers principal component methods and implementation in R
• Short, self-contained chapters with tested examples that allow for flexibility in
designing a course and for easy reference
At the end of each chapter, we present R lab sections in which we systematically work
through applications of the various methods discussed in that chapter. Additionally,
we provide links to other resources and to our hand-curated list of videos on principal
component methods for further learning.
0.3. HOW THIS BOOK IS ORGANIZED vii

Figure 1: Principal component methods

0.3 How this book is organized

This book is divided into 4 parts and 6 chapters. Part I provides a quick introduction to R
(chapter 1) and presents required R packages for the analysis and visualization (chapter
2).
In Part II, we describe classical multivariate analysis methods:
• Principal Component Analysis - PCA (chapter 3)
• Correspondence Analysis - CA (chapter 4)
• Multiple Correspondence Analysis - MCA (chapter 5)
In part III, we continue by discussing advanced methods for analyzing a data set contain-
ing a mix of variables (qualitative & quantitative) organized or not into groups:
• Factor Analysis of Mixed Data - FAMD (chapter 6) and,
• Multiple Factor Analysis - MFA (chapter 7).
Finally, we show in Part IV, how to perform hierarchical clustering on principal com-
ponents (HCPC) (chapter 8), which is useful for performing clustering with a data set
viii CONTENTS

containing only qualitative variables or with a mixed data of qualitative and quantitative
variables.
Some examples of plots generated in this book are shown hereafter. You’ll learn how to
create, customize and interpret these plots.
1) Eigenvalues/variances of principal components. Proportion of information
retained by each principal component.
Scree plot
Percentage of explained variances

50

41.2%
40

30

20 18.4%
12.4%
10 8.2% 7%
4.2% 3% 2.7% 1.6% 1.2%
0
1 2 3 4 5 6 7 8 9 10
Dimensions

2) PCA - Graph of variables:

• Control variable colors using their contributions to the principal components.
Variables - PCA
1.0
Pole.vault
X1500m

0.5
Javeline
X400m contrib
Dim2 (18.4%)

Long.jump 12

X110m.hurdle Shot.put 10
0.0 Discus
X100m
8

High.jump 6
-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

Dim1 (41.2%)
0.3. HOW THIS BOOK IS ORGANIZED ix

• Highlight the most contributing variables to each principal dimension:

Contribution to Dim 1 Contribution to Dim 2

15 30
Contributions (%)

Contributions (%)
10 20

5 10

0 0
X1 ng m
m mp

D le
Sh cus

h. m
Ja mp

le e

0m

h. m

X1 p
Sh m

m us
X1 ault

X1 ault

Ja mp

X4 ne
ng 0m

e
X4 ut

X1 D put
Po elin

dl
d
Lo 100

H 00

H 00

00
.p

10 isc
li
ur

ur
10 .ju

ju

50

ju

0
.ju

.
is

ve
.v

.v
ot

ot
5
v
.h

.h
le
X

ig

ig
Po

Lo
3) PCA - Graph of individuals:
• Control automatically the color of individuals using the cos2 (the quality of the
individuals on the factor map)
Individuals - PCA

CLAY
2
BERNARD SEBRLE
BOURGUIGNON Clay
Schoenbeck Sebrle
1 cos2
HERNU Pogorelov
Dim2 (18.4%)

Schwarzl 0.75

Warners
0 0.50
BARRAS YURKOV
0.25
Barras
Zsivoczky Karpov
-1
McMULLEN
MARTINEAU Hernu
NOOL Bernard
-2 ZSIVOCZKY
Macey
-2.5 0.0 2.5
Dim1 (41.2%)
x CONTENTS

• Change the point size according to the cos2 of the corresponding individuals:
Individuals - PCA

CLAY
2
BERNARD SEBRLE
BOURGUIGNON Clay
Schoenbeck Sebrle
1 Pogorelov
HERNU
Dim2 (18.4%)

cos2
Schwarzl 0.25

0
Warners 0.50
BARRAS YURKOV 0.75
Barras Zsivoczky
Karpov
-1 McMULLEN
MARTINEAU
Hernu
NOOL Bernard
ZSIVOCZKY
-2
Macey
-2.5 0.0 2.5
Dim1 (41.2%)

4) PCA - Biplot of individuals and variables

PCA - Biplot

3 Sepal.Width

Clusters
2
a petal
Sepal.Length
Dim2 (22.9%)

1
a sepal

Petal.Width
0 Species
Petal.Length
setosa

-1 versicolor
virginica
-2

-2 0 2
Dim1 (73%)
0.3. HOW THIS BOOK IS ORGANIZED xi

5) Correspondence analysis. Association between categorical variables.

CA - Biplot
Repairs
Husband
Laundry Breakfeast Driving
0.5 Wife
Main_meal
Dinner Alternating
Official
Dim2 (39.9%)

0.0

Tidying Shopping
Insurance
-0.5
Dishes
Finances

-1.0
Jointly

-1.5 Holidays
-1.0 -0.5 0.0 0.5 1.0 1.5
Dim1 (48.7%)

6) FAMD - Analyzing mixed data

FAMD factor map
Label Soil

2
1VAU 1VAU
2BEA 2BEA
1ROC 1DAM 1ROC 1DAM
1 PER1 2DAM Env4 Env2 2DAM
T2Chinon
1TUR 4EL T2 PER1 4EL
Dim2 (16.9%)

Saumur Reference
1BOI
T1 1BOI 1POY T1 1TUR 1POY
0
DOM1 1BEN 2BOU DOM1 1BEN 2BOU
3ELBourgueuil 3EL
-1
2EL 2EL
Env1
-2 1CHA 1ING 1CHA 1ING
1FON 1FON
-3
2ING 2ING

-6 -4 -2 0 2 -6 -4 -2 0 2
Dim1 (43.9%)
xii CONTENTS

7) Clustering on principal components

Cluster Dendrogram
2

1
Height

0
Mississippi
Louisiana

Maryland
Georgia
Alabama
Tennessee
Arizona

Michigan
Colorado
Nevada

Nebraska
New York
Texas
Illinois
Florida

Alaska
California

Vermont
Montana
Idaho
Wisconsin
Minnesota
Maine
Iowa

Hawaii
Utah

Arkansas
Delaware
Kentucky
Wyoming
Virginia
Ohio
Kansas
Indiana
Oklahoma
Missouri
Oregon
New Mexico

New Hampshire
Massachusetts
New Jersey
South Carolina
North Carolina

Rhode Island

Connecticut
Pennsylvania

Washington
West Virginia
South Dakota
North Dakota

-1

0.4 Book website

The website for this book is located at : http://www.sthda.com/english/. It contains
number of resources.

0.5 Executing the R codes from the PDF

For a single line R code, you can just copy the code from the PDF to the R console.
For a multiple-line R codes, an error is generated, sometimes, when you copy and paste
directly the R code from the PDF to the R console. If this happens, a solution is to:
• Paste firstly the code in your R code editor or in your text editor
• Copy the code from your text/code editor to the R console

0.6 Acknowledgment
I sincerely thank all developers for their efforts behind the packages that factoextra
depends on, namely, ggplot2 (Hadley Wickham, Springer-Verlag New York, 2009), Fac-
toMineR (Sebastien Le et al., Journal of Statistical Software, 2008), dendextend (Tal
Galili, Bioinformatics, 2015), cluster (Martin Maechler et al., 2016) and more.
0.7. COLOPHON xiii

0.7 Colophon
This book was built with:
• R 3.3.2
• factoextra 1.0.5
• FactoMineR 1.36
• ggpubr 0.1.5
• dplyr 0.7.2
• bookdown 0.4.3
About the author

Alboukadel Kassambara is a PhD in Bioinformatics and Cancer Biology. He works

since many years on genomic data analysis and visualization (read more: http://www.
alboukadel.com/).
He has work experiences in statistical and computational methods to identify prognostic
and predictive biomarker signatures through integrative analysis of large-scale genomic
and clinical data sets.
He created a bioinformatics web-tool named GenomicScape (www.genomicscape.com)
which is an easy-to-use web tool for gene expression data analysis and visualization.
He developed also a training website on data science, named STHDA (Statistical Tools for
High-throughput Data Analysis, www.sthda.com/english), which contains many tutorials
on data analysis and visualization using R software and packages.
He is the author of many popular R packages for:
• multivariate data analysis (factoextra, http://www.sthda.com/english/rpkgs/
factoextra),
• survival analysis (survminer, http://www.sthda.com/english/rpkgs/
survminer/),
• correlation analysis (ggcorrplot, http://www.sthda.com/english/wiki/
ggcorrplot-visualization-of-a-correlation-matrix-using-ggplot2),
• creating publication ready plots in R (ggpubr, http://www.sthda.com/english/
rpkgs/ggpubr).
Recently, he published three books on data analysis and visualization:
1. Practical Guide to Cluster Analysis in R (https://goo.gl/DmJ5y5)
2. Guide to Create Beautiful Graphics in R (https://goo.gl/vJ0OYb).
3. Complete Guide to 3D Plots in R (https://goo.gl/v5gwl0).

xiv
Part I

Basics

1
Chapter 1

Introduction to R

R is a free and powerful statistical software for analyzing and visualizing data. If you
want to learn easily the essential of R programming, visit our series of tutorials available
on STHDA: http://www.sthda.com/english/wiki/r-basics-quick-and-easy.
In this chapter, we provide a very brief introduction to R, for installing R/RStudio as
well as importing your data into R for computing principal component methods.

1.1 Installing R and RStudio

R and RStudio can be installed on Windows, MAC OSX and Linux platforms. RStudio
is an integrated development environment for R that makes using R easier. It includes a
console, code editor and tools for plotting.
1. R can be downloaded and installed from the Comprehensive R Archive Network
(CRAN) webpage (http://cran.r-project.org/)
2. After installing R software, install also the RStudio software available at: http:
//www.rstudio.com/products/RStudio/.
3. Launch RStudio and start use R inside R studio.

1.2 Installing and loading R packages

An R package is an extension of R containing data sets and specific R functions to solve
specific questions.
For example, in this book, you’ll learn how to compute and visualize principal component
methods using FactoMineR and factoextra R packages.
There are thousands other R packages available for download and installation from
CRAN1 , Bioconductor2 (biology related R packages) and GitHub3 repositories.
1
https://cran.r-project.org/
2
https://www.bioconductor.org/
3
https://github.com/

2
1.3. GETTING HELP WITH FUNCTIONS IN R 3

Figure 1.1: Rstudio interface

1. How to install packages from CRAN? Use the function install.packages():

install.packages("FactoMineR")
install.packages("factoextra")

2. How to install packages from GitHub? You should first install devtools if you don’t
have it already installed on your computer:
For example, the following R code installs the latest developmental version of factoextra
R package developed by A. Kassambara (https://github.com/kassambara/facoextra)
for multivariate data analysis and elegant visualization.
install.packages("devtools")
devtools::install_github("kassambara/factoextra")

Note that, GitHub contains the latest developmental version of R packages.

3. After installation, you must first load the package for using the functions in the
package. The function library() is used for this task.
library("FactoMineR")
library("factoextra")

Now, we can use R functions, such as PCA() [in the FactoMineR package] for performing
principal component analysis.

1.3 Getting help with functions in R

If you want to learn more about a given function, say PCA(), type this in R console:
?PCA
4 CHAPTER 1. INTRODUCTION TO R

1.4 Importing your data into R

1. Prepare your file as follow:
• Use the first row as column names. Generally, columns represent variables
• Use the first column as row names. Generally rows represent observations or
individuals.
• Each row/column name should be unique, so remove duplicated names.
• Avoid names with blank spaces. Good column names: Long_jump or Long.jump.
Bad column name: Long jump.
• Avoid names with special symbols: ?, $, *, +, #, (, ), -, /, }, {, |, >, < etc. Only
underscore can be used.
• Avoid beginning variable names with a number. Use letter instead. Good column
names: sport_100m or x100m. Bad column name: 100m
• R is case sensitive. This means that Name is different from Name or NAME.
• Avoid blank rows in your data.
• Delete any comments in your file.
• Replace missing values by NA (for not available)
• If you have a column containing date, use the four digit format. Good format:
01/01/2016. Bad format: 01/01/16
2. The final file should look like this:

Figure 1.2: General data format for importation into R

3. Save your file

We recommend to save your file into .txt (tab-delimited text file) or .csv (comma sepa-
rated value file) format.
4. Get your data into R:
Use the R code below. You will be asked to choose a file:
# .txt file: Read tab separated values
my_data <- read.delim(file.choose(), row.names = 1)
1.5. DEMO DATA SETS 5

# .csv file: Read comma (",") separated values

my_data <- read.csv(file.choose(), row.names = 1)

# .csv file: Read semicolon (";") separated values

my_data <- read.csv2(file.choose(), row.names = 1)

Using these functions, the imported data will be of class data.frame (R terminology).
You can read more about how to import data into R at this link: http://www.sthda.
com/english/wiki/importing-data-into-r

1.5 Demo data sets

R comes with several built-in data sets, which are generally used as demo data for playing
with R functions. The most used R demo data sets include: USArrests, iris and mtcars.
To load a demo data set, use the function data() as follow:
data("USArrests") # Loading
head(USArrests, 3) # Print the first 3 rows

## Murder Assault UrbanPop Rape

## Alabama 13.2 236 58 21.2
## Alaska 10.0 263 48 44.5
## Arizona 8.1 294 80 31.0
If you want learn more about USArrests data sets, type this:
?USArrests

To select just certain columns from a data frame, you can either refer to the columns by
name or by their location (i.e., column 1, 2, 3, etc.).
# Access the data in 'Murder' column
# dollar sign is used
head(USArrests$Murder)

## [1] 13.2 10.0 8.1 8.8 9.0 7.9

# Or use this
USArrests[, 'Murder']
# Or use this
USArrests[, 1] # column number 1

1.6 Close your R/RStudio session

Each time you close R/RStudio, you will be asked whether you want to save the data
from your R session. If you decide to save, the data will be available in future R sessions.
Chapter 2

Required R packages

2.1 FactoMineR & factoextra

There are a number of R packages implementing principal component methods. These
packages include: FactoMineR, ade4, stats, ca, MASS and ExPosition.
However, the result is presented differently depending on the used package.
To help in the interpretation and in the visualization of multivariate analysis - such as clus-
ter analysis and principal component methods - we developed an easy-to-use R package
named factoextra (official online documentation: http://www.sthda.com/english/
rpkgs/factoextra)(Kassambara and Mundt, 2017).

No matter which package you decide to use for computing principal component meth-
ods, the factoextra R package can help to extract easily, in a human readable data
format, the analysis results from the different packages mentioned above. factoextra
provides also convenient solutions to create ggplot2-based beautiful graphs.

In this book, we’ll use mainly:

• the FactoMineR package (Husson et al., 2017a) to compute principal component
methods;
• and the factoextra package (Kassambara and Mundt, 2017) for extracting, visu-
alizing and interpreting the results.
The other packages - ade4, ExPosition, etc - will be presented briefly.
The Figure 2.1 illustrates the key functionality of FactoMineR and factoextra.
Methods, which outputs can be visualized using the factoextra package are shown on the
Figure 2.2:

2.2 Installation

6
2.2. INSTALLATION 7

Figure 2.1: Key features of FactoMineR and factoextra for multivariate analysis

2.2.1 Installing FactoMineR

The FactoMineR package can be installed and loaded as follow:

# Install
install.packages("FactoMineR")

# Load
library("FactoMineR")

2.2.2 Installing factoextra

• factoextra can be installed from CRAN1 as follow:

install.packages("factoextra")

• Or, install the latest developmental version from Github2

if(!require(devtools)) install.packages("devtools")
devtools::install_github("kassambara/factoextra")

• Load factoextra as follow :

1
https://cran.r-project.org/package=factoextra
2
https://github.com/kassambara/factoextra
8 CHAPTER 2. REQUIRED R PACKAGES

Figure 2.2: Principal component methods and clustering methods supported by the
factoextra R package

library("factoextra")

2.3 Main R functions

2.3.1 Main functions in FactoMineR

Functions for computing principal component methods and clustering:

Functions Description
PCA Principal component analysis.
CA Correspondence analysis.
2.3. MAIN R FUNCTIONS 9

Functions Description
MCA Multiple correspondence analysis.
FAMD Factor analysis of mixed data.
MFA Multiple factor analysis.
HCPC Hierarchical clustering on principal components.
dimdesc Dimension description.

2.3.2 Main functions in factoextra

factoextra functions covered in this book are listed in the table below. See the online doc-
umentation (http://www.sthda.com/english/rpkgs/factoextra) for a complete list.
• Visualizing principal component method outputs

Functions Description
fviz_eig (or fviz_eigenvalue) Visualize eigenvalues.
fviz_pca Graph of PCA results.
fviz_ca Graph of CA results.
fviz_mca Graph of MCA results.
fviz_mfa Graph of MFA results.
fviz_famd Graph of FAMD results.
fviz_hmfa Graph of HMFA results.
fviz_ellipses Plot ellipses around groups.
fviz_cos2 Visualize element cos2. 3
4
fviz_contrib Visualize element contributions.

• Extracting data from principal component method outputs. The following

functions extract all the results (coordinates, squared cosine, contributions) for the
active individuals/variables from the analysis outputs.

Functions Description
get_eigenvalue Access to the dimension eigenvalues.
get_pca Access to PCA outputs.
get_ca Access to CA outputs.
get_mca Access to MCA outputs.
get_mfa Access to MFA outputs.
get_famd Access to MFA outputs.
get_hmfa Access to HMFA outputs.
facto_summarize Summarize the analysis.

• Clustering analysis and visualization

3
Cos2: quality of representation of the row/column variables on the principal component maps.
4
This is the contribution of row/column elements to the definition of the principal components.
10 CHAPTER 2. REQUIRED R PACKAGES

Functions Description
fviz_dend Enhanced Visualization of Dendrogram.
fviz_cluster Visualize Clustering Results.
 Part II

Classical Methods

11
Chapter 3

Principal Component Analysis

3.1 Introduction
Principal component analysis (PCA) allows us to summarize and to visualize the
information in a data set containing individuals/observations described by multiple inter-
correlated quantitative variables. Each variable could be considered as a different dimen-
sion. If you have more than 3 variables in your data sets, it could be very difficult to
visualize a multi-dimensional hyperspace.
Principal component analysis is used to extract the important information from a mul-
tivariate data table and to express this information as a set of few new variables called
principal components. These new variables correspond to a linear combination of the
originals. The number of principal components is less than or equal to the number of
original variables.
The information in a given data set corresponds to the total variation it contains. The
goal of PCA is to identify directions (or principal components) along which the variation
in the data is maximal.
In other words, PCA reduces the dimensionality of a multivariate data to two or three
principal components, that can be visualized graphically, with minimal loss of informa-
tion.
In this chapter, we describe the basic idea of PCA and, demonstrate how to compute
and visualize PCA using R software. Additionally, we’ll show how to reveal the most
important variables that explain the variations in a data set.

3.2 Basics
Understanding the details of PCA requires knowledge of linear algebra. Here, we’ll explain
only the basics with simple graphical representation of the data.
In the Plot 1A below, the data are represented in the X-Y coordinate system. The
dimension reduction is achieved by identifying the principal directions, called principal
components, in which the data varies.

12
3.2. BASICS 13

PCA assumes that the directions with the largest variances are the most “important” (i.e,
the most principal).
In the figure below, the PC1 axis is the first principal direction along which the samples
show the largest variation. The PC2 axis is the second most important direction
and it is orthogonal to the PC1 axis.
The dimensionality of our two-dimensional data can be reduced to a single dimension by
projecting each sample onto the first principal component (Plot 1B)
Plot 1A Plot 1B
80 40

20
40 PC2 PC1
PC2

0
y

0
-20

-40 -40
-10 -5 0 5 10 -80 -40 0 40
x PC1

Technically speaking, the amount of variance retained by each principal component is

measured by the so-called eigenvalue.
Note that, the PCA method is particularly useful when the variables within the data set
are highly correlated. Correlation indicates that there is redundancy in the data. Due to
this redundancy, PCA can be used to reduce the original variables into a smaller number
of new variables ( = principal components) explaining most of the variance in the
original variables.
Low redundancy High redundancy
3

2 4

1
0
y

-1
-4
-2

-3
-2 -1 0 1 2 -2 0 2
x x
Taken together, the main purpose of principal component analysis is to:
• identify hidden pattern in a data set,
• reduce the dimensionnality of the data by removing the noise and redun-
dancy in the data,
• identify correlated variables
14 CHAPTER 3. PRINCIPAL COMPONENT ANALYSIS

3.3 Computation

3.3.1 R packages

Several functions from different packages are available in the R software for computing
PCA:
• prcomp() and princomp() [built-in R stats package],
• PCA() [FactoMineR package],
• dudi.pca() [ade4 package],
• and epPCA() [ExPosition package]
No matter what function you decide to use, you can easily extract and visualize the
results of PCA using R functions provided in the factoextra R package.

Here, we’ll use the two packages FactoMineR (for the analysis) and factoextra (for
ggplot2-based visualization).

Install the two packages as follow:

install.packages(c("FactoMineR", "factoextra"))

Load them in R, by typing this:

library("FactoMineR")
library("factoextra")

3.3.2 Data format

We’ll use the demo data sets decathlon2 from the factoextra package:
data(decathlon2)
# head(decathlon2)

As illustrated in Figure 3.1, the data used here describes athletes’ performance during two
sporting events (Desctar and OlympicG). It contains 27 individuals (athletes) described
by 13 variables.
Note that, only some of these individuals and variables will be used to perform the
principal component analysis. The coordinates of the remaining individuals and
variables on the factor map will be predicted after the PCA.

In PCA terminology, our data contains :

• Active individuals (in light blue, rows 1:23) : Individuals that are used during
the principal component analysis.
• Supplementary individuals (in dark blue, rows 24:27) : The coordinates of
these individuals will be predicted using the PCA information and parameters
obtained with active individuals/variables