ℝ

Series ISSN: 1938-1743

AKINKUNMI
Introduction to
Series Editor: Steven G. Krantz, Washington University in St. Louis

Introduction to Statistics Using R

Statistics Using R
Mustapha Akinkunmi, American University of Nigeria

Introduction to Statistics Using R is organized into 13 major chapters. Each chapter is broken

INTRODUCTION TO STATISTICS USING R

down into many digestible subsections in order to explore the objectives of the book. There are
many real-life practical examples in this book and each of the examples is written in R codes
to acquaint the readers with some statistical methods while simultaneously learning R scripts.

ABOUT THE AUTHOR

Dr. Mustapha Akinkunmi is a Financial Economist and Technology Strategist. He has over 25
years of experience in estimation, planning, and forecasting using statistical and econometric
methods, with particular expertise in risk, expected utility, discounting, binomial-tree valuation
methods, financial econometrics models, Monte Carlo simulations, macroeconomics, and
exchange rate modeling. Dr. Akinkunmi has performed extensive software development for
quantitative analysis of capital markets, revenue and payment gateway, predictive analytics, data
science, and credit risk management.

Mustapha Akinkunmi

About SYNTHESIS
This volume is a printed version of a work that appears in the Synthesis
Digital Library of Engineering and Computer Science. Synthesis

MORGAN & CLAYPOOL

books provide concise, original presentations of important research and
development topics, published quickly, in digital and print formats.

store.morganclaypool.com
Introduction to Statistics Using R
 Synthesis Lectures on
Mathematics and Statistics
Editor
Steven G. Kranz, Washington University, St. Louis

Introduction to Statistics Using R

Mustapha Akinkunmi
2018

PDE Models for Atherosclerosis Computer Implementation in R

William E. Schiesser
2017

An Introduction to Partial Differential Equations

Daniel J. Arrigo
2017

Numerical Integration of Space Fractional Partial Differential Equations: Vol 2 –

Applicatons from Classical Integer PDEs
Younes Salehi and William E. Schiesser
2017

Numerical Integration of Space Fractional Partial Differential Equations: Vol 1 –

Introduction to Algorithms and Computer Coding in R
Younes Salehi and William E. Schiesser
2017

Aspects of Differential Geometry III

Esteban Calviño-Louzao, Eduardo García-Río, Peter Gilkey, JeongHyeong Park, and Ramón
Vázquez-Lorenzo
2017

The Fundamentals of Analysis for Talented Freshmen

Peter M. Luthy, Guido L. Weiss, and Steven S. Xiao
2016
iv

Aspects of Differential Geometry II

Peter Gilkey, JeongHyeong Park, Ramón Vázquez-Lorenzo
2015

Aspects of Differential Geometry I

Peter Gilkey, JeongHyeong Park, Ramón Vázquez-Lorenzo
2015

An Easy Path to Convex Analysis and Applications

Boris S. Mordukhovich and Nguyen Mau Nam
2013

Applications of Affine and Weyl Geometry

Eduardo García-Río, Peter Gilkey, Stana Nikčević, and Ramón Vázquez-Lorenzo
2013

Essentials of Applied Mathematics for Engineers and Scientists, Second Edition

Robert G. Watts
2012

Chaotic Maps: Dynamics, Fractals, and Rapid Fluctuations

Goong Chen and Yu Huang
2011

Matrices in Engineering Problems

Marvin J. Tobias
2011

The Integral: A Crux for Analysis

Steven G. Krantz
2011

Statistics is Easy! Second Edition

Dennis Shasha and Manda Wilson
2010

Lectures on Financial Mathematics: Discrete Asset Pricing

Greg Anderson and Alec N. Kercheval
2010

Jordan Canonical Form: Theory and Practice

Steven H. Weintraub
2009
 v
The Geometry of Walker Manifolds
Miguel Brozos-Vázquez, Eduardo García-Río, Peter Gilkey, Stana Nikčević, and Ramón
Vázquez-Lorenzo
2009

An Introduction to Multivariable Mathematics

Leon Simon
2008

Jordan Canonical Form: Application to Differential Equations

Steven H. Weintraub
2008

Statistics is Easy!
Dennis Shasha and Manda Wilson
2008

A Gyrovector Space Approach to Hyperbolic Geometry

Abraham Albert Ungar
2008
Copyright © 2019 by Morgan & Claypool

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in
any form or by any means—electronic, mechanical, photocopy, recording, or any other except for brief quotations
in printed reviews, without the prior permission of the publisher.

Introduction to Statistics Using R

Mustapha Akinkunmi
www.morganclaypool.com

ISBN: 9781681735092 paperback

ISBN: 9781681735108 ebook
ISBN: 9781681735115 hardcover

DOI 10.2200/S00899ED1V01Y201902MAS024

A Publication in the Morgan & Claypool Publishers series

SYNTHESIS LECTURES ON MATHEMATICS AND STATISTICS

Lecture #24
Series Editor: Steven G. Kranz, Washington University, St. Louis
Series ISSN
Print 1938-1743 Electronic 1938-1751
Introduction to Statistics Using R

Mustapha Akinkunmi
American University of Nigeria

SYNTHESIS LECTURES ON MATHEMATICS AND STATISTICS #24

M
&C Morgan & cLaypool publishers
ABSTRACT
Introduction to Statistics Using R is organized into 13 major chapters. Each chapter is broken
down into many digestible subsections in order to explore the objectives of the book. There are
many real-life practical examples in this book and each of the examples is written in R codes to
acquaint the readers with some statistical methods while simultaneously learning R scripts.

KEYWORDS
descriptive statistics, probability distributions, sampling distribution, hypothesis
testing, regression analysis, correlation analysis, confidence interval
 ix

To my son,
Omar Olanrewaju Akinkunmi
 xi

Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii

1 Introduction to Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Scales of Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Data, Data Collection, and Data Presentation . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Data Groupings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Categories of Data Groupings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Calculation of Class Interval for Ungrouped Data . . . . . . . . . . . . . . . . . 5
1.3.3 Class Limits and Class Boundaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Methods of Visualizing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Exploring Data Analysis (EDA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5.1 Stem-and-Leaf Displays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5.2 Box Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Introduction to R Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 How to Download and Install R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Using R for Descriptive Statistical and Plots . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Basics of R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.1 R is Vectorized . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.2 R Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.3 Missing Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3.4 Data Creation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.5 Data Type Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3.6 Variable Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4 Basic Operations in R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.1 Subsetting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4.2 Control Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.4.3 Built-In Functions in R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.4 User-Written Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.4.5 Importing, Reporting, and Writing Data . . . . . . . . . . . . . . . . . . . . . . . . 46
xii
2.5 Data Exploration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.5.1 Data Exploration through Visualization . . . . . . . . . . . . . . . . . . . . . . . . 51
2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3 Descriptive Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1 Central Tendency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2 Measure of Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3 Shapes of the Distribution—Symmetric and Asymmetric . . . . . . . . . . . . . . . . 60
3.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4 Basic Probability Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.1 Experiment, Outcome, and Sample Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.1 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.2 Outcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.3 Sample Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2 Elementary Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3 Complementary Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 Mutually Exclusive Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.5 Mutually Inclusive Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.6 Venn Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.7 Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.7.1 Simulation of a Random Sample in R . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.8 Axioms of Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.9 Basic Properties of Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.10 Independent Events and Dependent Events . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.11 Multiplication Rule of Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.12 Conditional Probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.13 Computation of Probability in R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.14 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5 Discrete Probability Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

5.1 Probability Mass Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2 Expected Value and Variance of a Discrete Random Variable . . . . . . . . . . . . . 78
5.3 Binomial Probability Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.4 Expected Value and Variance of a Binomial Distribution . . . . . . . . . . . . . . . . . 82
5.4.1 Expected Value (Mean) of a Binomial Distribution . . . . . . . . . . . . . . . 82
 xiii
5.4.2 Variance of a Binomial Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.5 Solve Problems Involving Binomial Distribution Using R . . . . . . . . . . . . . . . . 87
5.5.1 Generating a Random Sample from a Binomial Distribution . . . . . . . 88
5.5.2 Calculate Probability from a Binomial Distribution . . . . . . . . . . . . . . . 89
5.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

6 Continuous Probability Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

6.1 Normal Distribution and Standardized Normal Distribution . . . . . . . . . . . . . 94
6.2 Standard Normal Score (Z -Score) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3 Approximate Normal Distribution to the Binomial Distribution . . . . . . . . . . 98
6.4 Use of the Normal Distribution in Business Problem Solving Using R . . . . . . 99
6.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

7 Other Continuous Probability Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . 103

7.1 Student-t Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.2 Chi-Square Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.3 F-Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
7.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

8 Sampling and Sampling Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

8.1 Probability and Non-Probability Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.2 Probability Sampling Techniques—Simple Random, Systematic,
Stratified, and Cluster Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
8.2.1 Simple Random Sampling (SRS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
8.2.2 Systematic Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
8.2.3 Cluster Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
8.2.4 Stratified Random Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
8.3 Non-Probability Sampling Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.3.1 Judgment Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.3.2 Quota Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.3.3 Convenience Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.4 Sampling Distribution of the Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.5 Central Limit Theorem and its Significance . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
xiv
9 Confidence Intervals for Single Population Mean and Proportion . . . . . . . . . 131
9.1 Point Estimates and Interval Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
9.2 Confidence Intervals for Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
9.3 Confidence Intervals for Proportion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
9.4 Calculating the Sample Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
9.5 Factors That Determine Margin of Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
9.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

10 Hypothesis Testing for Single Population Mean and Proportion . . . . . . . . . . 141

10.1 Null and Alternative Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
10.2 Type I and Type II Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
10.3 Acceptance and Rejection Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
10.4 Hypothesis Testing Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
10.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

11 Regression Analysis and Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

11.1 Construction of Line Fit Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
11.2 Types of Regression Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
11.2.1 Uses of Regression Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
11.2.2 Simple Linear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
11.2.3 Assumptions of Simple Linear Regression . . . . . . . . . . . . . . . . . . . . . . 155
11.3 Multiple Linear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
11.3.1 Significance Testing of Each Variable . . . . . . . . . . . . . . . . . . . . . . . . . 164
11.3.2 Interpretation of Regression Coefficients and Other Output . . . . . . . 165
11.4 Pearson Correlation Coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
11.4.1 Assumptions of Correlation Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
11.4.2 Types of Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
11.4.3 Coefficient of Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
11.4.4 Test for the Significance of Correlation Coefficient (r) . . . . . . . . . . . . 168
11.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

12 Poisson Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

12.1 Poisson Distribution and its Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
12.2 Mean and Variance of a Poisson Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 177
12.3 Application of Poisson Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
12.4 Poisson to Approximate the Binomial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
12.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
 xv
13 Uniform Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
13.1 Uniform Distribution and Its Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
13.2 Mean and Variance of a Uniform Distribution . . . . . . . . . . . . . . . . . . . . . . . . 189
13.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

A Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

Author’s Biography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

 xvii

Preface

Chapter 1: Introduction to Statistical Analysis

This chapter helps readers that have litle or no knowledge about different scales of measurement
in statistics. It also helps readers to understand the concept of data, data gathering, and how
the data collected can be presented. In addition, this chapter helps the readers to be able to
understand different categories of data grouping and to explore data analysis and visualization.

Chapter 2: Introduction to R Software

This chapter focuses on how to download and install R packages on a computer system, gives
the reason for using R, and also gives full details of the steps involved in the installation of R.
This chapter introduces readers to basic functions in R with illustrative examples. This chapter
uncovers basic operations in R, built-in functions (numerical, statistical probability), and other
useful functions in R. This chapter gives an overview on how to import packages and how to
change directories for reading and writing local flats files. It helps in explaining how to connect
to a database using R commands. Data exploration through visualization in R is demonstrated
and desired charts are plotted.

Chapter 3: Descriptive Data

This chapter provides different kinds of statistical measures to analyze business-related data. The
chapter focuses on measures of central tendency and dispersion, and gives the characteristics
of the measures of central tendency and dispersion. Practical examples are given for each of
these descriptive statistics mentioned in the chapter along with explanatory short R scripts. This
chapter also explains the shapes of the distribution with illustrative examples.

Chapter 4: Basic Probability Concepts

This part of the book gives meaning to some probability terminologies—experiment, sample
space, elementary event, complementary events, mutually exclusive events, mutually inclusive
events, Venn diagram, independent, and dependent events. Axioms of probability is treated
alongside basic properties of probability. This chapter walks the readers through the additive
and multiplicative rules of probability and conditional probability. The chapter shows how to
use R to generate and compute probabilities.
xviii PREFACE
Chapter 5: Discrete Probability Distribution
This chapter explains the concept of a discrete probability distribution, probability mass function,
the expected value, and the variance of a discrete probability distribution. It gives examples of
discrete probabilities and deals much on the binomial probability distribution. A subsection is
dedicated to showing how to derive the expected value and variance for a binomial distribution.
This chapter teaches readers how to generate a random sample from a binomial distribution. All
examples given in this chapter are also written in R to acquaint readers on the software.

Chapter 6: Continuous Probability Distribution

Apart from the discrete probability distribution explained in the previous chapter, there is a
need for readers to know about the continuous probability distribution. This chapter focuses
on the normal distribution among other continuous probability distributions mentioned in the
chapter. The readers can vividly differentiate between the discrete and continuous probability
distributions. The normal curve is explained with clear examples, and the properties of a normal
distribution are stated. In addition, readers will be able to use standard normal score (z-score) to
solve business problems. This chapter exposes readers to using normal approximation to solve a
binomial problem. Many worked examples, as well as the R codes, are given on normal proba-
bility to sensitize readers to its application in businesses.

Chapter 7: Other Continuous Probability Distributions

This chapter is an extension of Chapter 6. The chapter explains the commonly used distribu-
tion which is important in the testing of an hypothesis. These distributions are the student-t,
Chi-square, and F distributions, and their properties and uses. This chapter helps the readers
to be acquainted to the conditions when each of the aforementioned distributions are used in
the testing of an hypothesis. The main objective of this chapter is for the readers to be able to
know when each of the distributions can be used and to obtain respective table values for the
distributions, even using the R commands.

Chapter 8: Sampling and Sampling Distribution

This chapter gives an overview of the sampling techniques with their merits and demerits. This
chapter classifies sampling into probability and non-probability sampling. The sampling distri-
bution of the mean is established by using simple random sampling with and without replace-
ment on data. This section of the book covers the central limit theorem and its significance with
worked examples.

Chapter 9: Confidence Intervals for Single Population Mean and Proportion

This chapter explains how to construct confidence intervals for mean and proportion with illus-
trative examples. It sheds more light on how to calculate the sample size when confidence level
 PREFACE xix
and margin of error are given. The chapter gives the main factors that determine a margin of
error. Readers are expected to construct confidence limit using appropriate test statistics.

Chapter 10: Hypothesis Testing for Single Population Mean and Proportion
This chapter is all about the decision-making process and it explains how to form and test hy-
potheses. It prepares readers on the steps to conduct hypothesis testing and interpret the results.
This chapter helps the readers to use available data to make a valid conclusion.

Chapter 11: Regression Analysis and Correlation

The existence of a relationship between two or more variables is taught in this chapter. The
readers would understand and differentiate between the regression and correlation analyses. The
chapter explains the uses and assumptions of simple and multiple linear regressions. This chapter
helps the readers to know the interpretation of the regression model and to know how to use
regression analysis in making predictions. The extent of the association between variable is also
discussed with the aid of diagrams; this will give the readers intuition into the regression and
correlation analyses. Many examples are given to broaden the scope of the readers on the topic.

Chapter 12: Poisson distribution

This chapter deals with the count data. It gives statistical definition and a real-life example of a
Poisson distribution. Different Poisson distribution graphs are demonstrated to show its char-
acteristics as its parameter increases. This chapter derives and shows that mean and variance of a
Poisson distribution are the same. The chapter explains the applications of a Poisson distribution
in various aspects of life. Furthermore, this section proves how Poisson can be approximated to
binomial distribution with some worked examples.

Chapter 13: Uniform distribution

The main objectives of this chapter is for readers to be able to identify uniform distribution, its
uses, and applications. This chapter explains a uniform distribution with aid of a diagram and
the condition for using uniform distribution. It shows how to derive the mean and variance of
a uniform distribution.

Mustapha Akinkunmi
February 2019
 1

CHAPTER 1

Introduction to Statistical
Analysis
Statistics is a science that supports us in making better decisions in business, economics, and
other disciplines. In addition, it provides the necessary tools needed to summarize data, analyze
data, and make meaningful conclusions to achieve a better decision. These better decisions assist
us in running a small business, a corporation, or the economy as a whole. This book is a bit
different from most statistics books in that it will focus on using data to help you do business
and so the focus is not on statistics for statistics stake but rather on what you need to know to
really use statistics.
To help in that endeavor, examples will include the use to the R programming language
which was created specifically to give statisticians and technicians a tool to create solutions to
standard techiques or to create their own solutions.
Statistics is a word that originated from the Italian word stato meaning “state” and statista
is an individual saddled with the tasks of the state. Thus, statistics is the collection of useful infor-
mation to the statista. Its application commenced in Italy during the 16th century and diffused
to other countries around the world. At present, statistics covers a wide range of information
in every aspect of human activities. In addition, it is not limited to the collection of numerical
information but includes data summarization, presentation, and analysis in meaningful ways.
Statistical analysis is mainly concerned with how to make generalizations from the data.
Statistics is a science that deals with information. In order for us to perform statistical analysis on
information (data) you have on hand or collect, you may need to transform the data or work with
the data to get it in a form where it can be analyzed using statistical techniques. Information can
be found in qualitative or quantitative form. In order to explain the difference between these two
types of information, let’s consider an example. Suppose an individual intends to start a business
based on the information in Table 1.1. Which of the variables are quantitative and which are
qualitative? The product price is a quantitative variable because it provides information based
on quantity—the product price in dollars. The number of similar businesses and the rent for
business premises are also quantitative variables. The location used in establishing the business
is a qualitative variable since it provides information about a quality (in this case a location, such
as Nigeria or South Korea). The presence of basic infrastructures requires a (Yes or No) response,
these are also qualitative variables.
2 1. INTRODUCTION TO STATISTICAL ANALYSIS
Table 1.1: Business feasibility data

Number of Rent for the Presence of basic

Product price Location
similar business business premise infrastructures
US$150 6 US$2,000 Nigeria No
US$100 18 US$3,000 South Korea Yes

A quantitative variable represents a number for which arithmetic operations such as av-
eraging make sense. A qualitative (or categorical) variable is concerned with quality. In a case
where a number is applied to separate members of different categories of a qualitative variable,
the assigned number is subjective but generally intentional. An aspect of statistics is concerned
with measurements—some quantitative and others qualitative. Measurements provide the real
numerical values of a variable. Qualitative variables can be represented with numbers as well,
but such a representation might be arbitrary but intended to be useful for the purposes at hand.
For instance, you can assign numerics to an instance of a qualitative variable such as Nigeria D
1 and South Korea D 0.

1.1 SCALES OF MEASUREMENTS

In order to use statistics effectively, it is helpful to view data in a slightly different way so that
the data can be analyzed successfully. In performing a statistical test, it is important that all data
is converted to the same scale. For instance, if your quantitative data is in meters and feet, you
need to choose one and convert any data to that scale. However, in statistics there is another
definition of scales of measurement. Scales of measurements are commonly classified into four
categories, namely: nominal scale, ordinal scale, interval scale, and ratio scale.
1. Nominal Scale: In this scale, numbers are simply applied to label groups or classes. For
instance, if a dataset consists of male and female, we may assign a number to them such
as 1 for male and 2 for female. In this situation, the numbers 1 and 2 only denote the
category in which a data point belongs. The nominal scale of measurement is applied to
qualitative data such as male, female, geographic classification, race classification, etc.
2. Ordinal Scale: This scale allows data elements to be ordered based on their relative size or
quality. For instance, buyers can rank three products by assigning them 1, 2, and 3, where
3 is the best and 1 is the worst. The ordinal scale does not provide information on how
much better one product is compared to others, only that it is better. This scaling is used
for many purposes, for instance grading, either stanine (1–9) or A–F (no E) where a 4.0
is all A’s and all F’s are 0.0, or rankings on, for instance, Amazon (1–5 stars so that that
5.0 would be a perfect ranking) or restaurants, hotels, and other data sources which may
be ranked on a four- or five-star basis. It is therefore quite important in using this data to
 1.2. DATA, DATA COLLECTION, AND DATA PRESENTATION 3
know the data range. The availability of this type of data on the Internet makes this a good
category for experimenting with real data.
3. Interval Scale: Interval scales are numerical scales in which intervals have the same in-
terpretation. For instance, in the case of temperature, we have a centigrade scale in which
water freezes at 0ı and boils at 100ı and a Fahrenheit scale in which 32ı is the tempera-
ture at which water freezes, but 212ı is the temperature at which it boils. These arbitrary
decisions can make it difficult to take ratios. For example, it is not correct to say 8 PM
is twice as long as 4 PM. However, it is possible to have ratios of intervals. The interval
between 0:00 PM (noon) and 8:00 PM which represents a duration of 8 h is twice as long
as the interval between 0:00 PM and 4:00 PM, which represents a duration of 4 h. The
time 0:00 PM does not imply absence of any time. It is important that you understand
your data and make sure it is converted appropriately so that in cases like this, interval
scale can be used in statistical tests.
4. Ratio Scale: This scale allows us to take ratios of two measurements if the two measure-
ments are in the ratio scale. The zero in the ratio scale is an absolute zero. For instance,
money is measured in a ratio scale. A sum of US$1,000 is twice as large US$500. Other
examples of the ratio scale include the measurements of weight, volume, area, or length.

1.2 DATA, DATA COLLECTION, AND DATA

PRESENTATION
A dataset is defined as a set of measurements gathered on some variables. For example, dividend
measurements for 10 companies may create a dataset. In this case, the concerned variable is
dividend and the scale of measurement is a ratio scale (a dividend for one company may be twice
another company’s dividend). In the actual observations of the companies’ dividends, the dataset
might record US$4, US$6, US$5, US$3, US$2, US$7, US$8, US$1, US$10, and US$9.
Different approaches are used to collect data. Sometimes, a dataset might include the
whole population of our interest, whereas in other cases, it might constitute a portion from the
larger population. In this situation, the data needs to be carefully collected because we want to
draw some inferences about the larger population based on the sample. The conclusion made
about an entire population by relying on the information in a sample from the population is
referred to as a statistical inference. This book assigns a great importance to the topic of statistical
inference (it is fully explained in Chapter 6). The accuracy of statistical inference depends on how
data is drawn from the population of interest. It is also critically important to ensure that every
segment of the population is sufficiently and equally captured in the sample. Data collection
from surveys or experiments has to be appropriately constructed, as statistical inference is based
on them.
Data can also be obtained from published sources such as statistical abstracts of different
publications. The published inflation rate for a number of months is a typical example. This
4 1. INTRODUCTION TO STATISTICAL ANALYSIS
data is collected as it is given without any control over how it is obtained. Inflation data for
over a certain time is not a random sample of any future inflation rate, thus, making statistical
inferences becomes complex and difficult. If the interest data is on the period where data is
available, this makes the data consist of the whole population. However, it is greatly important
to observe any missing data or incomplete observations.
This section focuses on the processing, summarization, and presentation of data which
represents the first stage of statistical analysis. What are the reasons behind more attention on
inference and population? Is it sufficient to observe data and interpret it? Data inspection is
appropriate when the interest is aimed at particular observations, but meaningful conclusions
require statistical analysis. This implies that any business decision based on a sample requires
meaningful insights that can be drawn from the statistical analysis. Therefore, the role of statis-
tics in any business entity is both a necessary and sufficient condition for effective and efficient
decision-making. For instance, marketing research, that often aims to examine the link between
advertising and sales, might obtain a dataset of randomly chosen sales and advertising figures for
a given firm. However, much more useful information can be concluded if implications about
the fundamental process are considered. Other businesses like banking might have interest in
evaluating the diffusion of a particular model of automatic teller machines; a pharmaceutical
firm that intends to market its new drug might require proving the drug does not cause negative
side effects. These different aims of business entities can be attained through the use of statistical
analysis.
Statistics can also have a dark side. The use of statistics takes a significant role in the
success of many business activities. Therefore, it is in a company’s interest to use statistics to
their benefit. So AIG might say that since 2010 their stock has gone up X %, having been bailed
out the year before. You see mutual funds brag all the time about their record since a certain date.
You can be certain that right before that date something bad happened. Or a company may claim
that their survey showed that 55% prefered their brand of tomato soup. How many surveys did
they conduct? How many different types of soup did they survey where the soup did poorly in
statistical results? Another is that in a test, one aspect of a product can come out well, but other
aspects may get bad results. So the company presents the positive results. Sampling is another
area where it is easy to cheat as can be the integrity of existing data. There are many other ways
to “lie” with statistics and it has become so commonplace that there are books written on that
subject. The bottom line is to have a healthy skepticism when you hear statistics promoted in
the media.

1.3 DATA GROUPINGS

Data is information that captures the qualitative or quantitative features of a variable or a set
of variables. Put differently, data can be defined as any set of information that describes a given
entity.
 1.3. DATA GROUPINGS 5
1.3.1 CATEGORIES OF DATA GROUPINGS
In statistics, data can be categorized into grouped and ungrouped data. Ungrouped data is data
in the raw form. For instance, a list of business firms that you know. Grouped data is data that
has been organized into groups known as classes. This implies that grouped data is no longer in
the raw format.
A class can be defined as a group of data values within specific group boundaries. The
frequency distribution table is the table that shows the different measurement categories and how
often each category occurs. Plotting a frequency distribution of grouped data, such a frequency
plot is called a histogram. A histogram is a chart comprised of bars of different heights. The height
of each bar depicts the frequency of values in the class represented by the bar. Adjacent bars share
sides. It is used only for measured or ordinal data. This is an appropriate way of plotting the
frequencies of grouped data. The area of each bar in a histogram corresponds to the frequency.
Table 1.2 is a frequency table that shows a random sample of 30 small-scale businesses based on
their years of existance.

Table 1.2: Frequency table for grouped data

Years of Establishment Number of Businesses

0-2 12
3-5 6
6-8 8
9 - 11 3
12 -14 1

Figure 1.1 shows the histogram representation for the given data in Table 1.2.
A data class refers to a group of data that are similar based on user-defined property. For
instance, ages of students can be grouped into classes such as those in their teens, 20s, 30s, etc.
Each of these groups is known as a class. Each class has a specific width known as the class
interval or class size. The class interval is very crucial in the construction of histograms and
frequency diagrams. Class size depends on how the data is grouped. The class interval is usually
a whole number.
However, an example of grouped data that have different class interval is presented in
Table 1.4. Tables like Table 1.4 are often called frequency tables as they show the frequency
that data is represented in a specific class interval.

1.3.2 CALCULATION OF CLASS INTERVAL FOR UNGROUPED DATA

In order to obtain a class interval for a given set of raw or ungrouped data, the following steps
have to be taken.
6 1. INTRODUCTION TO STATISTICAL ANALYSIS
Years of Establishment

14
12
12

Number of Businesses
10
8
8
6
6

4 3

2 1

0
0-2 3-5 6-8 9 - 11 12 - 14

Number of Years

Figure 1.1: Histogram.

Table 1.3: Grouped data with the same class interval

Daily Income Class

Frequency
(US$) Interval
0-9 5 10
10 - 19 8 10
20 - 29 12 10
30 - 39 17 10
40 - 49 3 10
50 - 59 4 10
60 - 69 7 10

1. Determine the number of classes you want to have.

2. Subtract the lowest value from the highest value in the dataset.

3. Divide the outcome under Step 2 by the number of classes you have in Step 1.

The mathematical expression of these steps is below:

highest value lowest value

Class interval D :
number of classes
 1.3. DATA GROUPINGS 7
Table 1.4: Grouped data with different class interval

Daily Income Class

Frequency
(US$) Interval
0-9 15 10
10 - 19 18 10
20 - 29 17 10
30 - 49 35 20
50 - 79 20 30

For instance, if a market survey is conducted on market prices for 20 non-durable goods with
the following prices: US$11, US$5, US$3, US$14, US$1, US$16, US$2, US$12, US$2, US$4,
US$3, US$9, US$4, US$8, US$7, US$5, US$8, US$6, US$10, and US$15. The raw data in-
dicate the lowest price is US$1 and the highest price is US$16. In addition, the survey expert
decides to have four classes. The class interval is written as follows:
highest value lowest value
Class Interval D
number of classes
16 1 15
D D
4 4
Class interval D 3:75:
The value of the class interval is usually a whole number, but in this case its value is a
decimal number. Therefore, the solution to this problem is to round-off to the nearest whole
number which is 4. This implies that the raw data can be grouped into 4, as presented in the
Table 1.5.
Table 1.5: Class interval generated from ungrouped data

Number Frequency
1-4 7
5-8 6
9 - 12 4
13 - 16 3

1.3.3 CLASS LIMITS AND CLASS BOUNDARIES

Class limits are the actual values in the above-mentioned tables. Taking an example of Table 1.5,
1 and 4 are the class limits of the first class. Class limits are divided into two categories: lower
8 1. INTRODUCTION TO STATISTICAL ANALYSIS
class limit and upper class limit. In Table 1.5 for the first class, 1 is the lower class limit while 4 is
the upper class limit. Class limits are used in frequency tables.
The lower class boundary is obtained by adding the lower class limit of upper class and
upper class limit of the lower class and divide the answer by 2. The upper class boundary is also
obtained by adding the upper class limit of lower class and lower class limit of the upper class and
divide the answer by 2. Class boundaries reflect the true values of the class interval. In Table 1.5,
the first line actually consists of values up to 4.5, which is the upper class boundary there. Class
boundaries are also divided into lower and upper class boundaries. Class boundaries are not often
observed in the frequency tables unless the data in the table is all integers. The class boundaries
for the ungrouped data above are shown in Table 1.6 below.

Table 1.6: Class boundaries generated from ungrouped data

Number Frequency Class Boundaries

(Lower – Upper)
1-4 7 0.5 – 4.5
5-8 6 4.5 – 8.5
9 - 12 4 8.5 – 12.5
13 - 16 3 12.5 – 16.5

A class interval is the range of data in that particular class. The relationship between the
class boundaries and the class interval is given as follows:
Class Interval D Upper Class Boundary Lower Class Boundary:
The lower class boundary of one class is the same to the upper class boundary of the previous
class for both integers and non-integers. The importance of class limits and class boundaries is
highly recognized in the diagrammatical representation of statistical data.

1.4 METHODS OF VISUALIZING DATA

1. Time plot. A time plot shows the history of several data items during the latest time
horizon as lines. The time axis is always horizontal and directed to the right. The plot
displays changes in variable of interest over a time period. It is also known as a line chart.
2. Pie charts. A pie chart simply displays data as a percentage of a given total. This is the
most appropriate way of visualizing quantities in relation to percentages of a given total.
The entire area of the pie is 100% of the quantity of the concerned matter and the size of
each slice is the percentage of the total represented by the category the slice denotes.
3. Bar charts. Bar charts are mainly utilized to showcase categorical data without focus on
the percentage. Its scale of measurement can be either nominal or ordinal. Bar charts can
 1.4. METHODS OF VISUALIZING DATA 9
be displayed with the use of horizontal or vertical bars. A horizontal bar chart is more
appropriate if someone intends to write the name of each category inside the rectangle
that represents that category. A vertical bar chart is required to stress the height of different
columns as measures of the quantity of the concerned variable.
4. Frequency polygons. A frequency polygon displays the same as a histogram, except with-
out rectangles in it. It is only a point, in the midpoint of each interval at a height corre-
sponding to the frequency or a relative frequency of the category of the interval.
5. Ogives. An ogive refers to a cumulative frequency or cumulative relative-frequency graph.
It commences at 0 and reaches a peak of 1.00 (for a relative-frequency ogive) or to the
peak cumulative frequency. The point with height assigned to the cumulative frequency is
situated at the right endpoint of each interval.

Inflation
16
15
Rate (%)

14
13
12
11
10
Jan-18

Feb - 18

Mar - 18

Apr - 18

May - 18

Jun - 18

Jul - 18

Aug - 18

Sep - 18

Oct - 18

Nov - 18

Dec - 18
Figure 1.2: Methods of visualizing data: time plot.

0.32%

26.86%
PMS
HHK
AGO
72.82%

Figure 1.3: Methods of visualizing data: pie chart.

10 1. INTRODUCTION TO STATISTICAL ANALYSIS
50

40

30

20

10

0
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Figure 1.4: Methods of visualizing data: bar chart.

14

12 12
Number of Businesses

10

8 8
6 6
4
3
2
1
0
1 4 7 10 13

Years of Establishment

Figure 1.5: Methods of visualizing data: frequency polygon.

Note: Charts can often be deceiving. This indicates the disadvantage of merely descrip-
tive methods of analysis and the need for statistical inference. Exploring statistical tests makes
the analysis more objective than eyeball analysis and less prone to deception if assumptions of
random sampling and others are established.

1.5 EXPLORING DATA ANALYSIS (EDA)

EDA is the name ascribed to a broad body of statistical and graphical methods. These approaches
present patterns of observing data to determine relationships and trends, identify outliers and
influential observations, and quickly describe or summarize datasets. This analysis was initiated
 1.5. EXPLORING DATA ANALYSIS (EDA) 11
100
90

Cumulative Frequency
80
70
60
50
40
30
20
10
0
0.05 10.5 20.5 30.5 40.5 50.5 60.5 70.5 80.5 90.5

Student Marks

Figure 1.6: Methods of visualizing data: ogive.

from the work of John W. Tukey “Exploratory Data Analysis” in 1977 and has made significant
strides in the past five years as software solutions in large data analysis (big data) and in busi-
ness intelligence (the reporting of characteristics of the data in meaningful ways) has improved
dramatically.

1.5.1 STEM-AND-LEAF DISPLAYS

This presents a quick way of observing a dataset. It has some attributes of a histogram but pre-
vents the loss of information in a histogram that emanates from organizing data into intervals.
Its display adopts the tallying principles as well as the use of the decimal base of a number sys-
tem. The stem refers to the number without its rightmost digit (the leaf ). The stem is written to
the left of a vertical line separating the stem from the leaf.
For instance, we have the following numbers: 201, 202, 203, and 205. This can be displayed
as:
20 j 1235:

1.5.2 BOX PLOTS

A box plot is also known as a box-whisker plot which presents a dataset by determining its
central tendency, spread, skewness, and the presence of outliers. The box plot consists of the
five summary measures of the distribution of data: median, lower quartile, upper quartile, the
smallest observation, and the largest observation (Fig. 1.7). It has two fundamental assumptions.
Its first assumption is that the hinges of the box plot are basically the quartiles of the dataset, and
the median is a line inside the box. Its second assumption is that its whiskers are constructed by
extending a line from the upper quartile to the largest observation, and from the lower quartile
12 1. INTRODUCTION TO STATISTICAL ANALYSIS
to the smallest observation under the condition that the largest and smallest observations lie
within a distance of 1.5 times the interquartile range from the appropriate hinge (quartile).
Observations that exceed the distance are regarded as suspected outlier.

(Lower Quartile) (Median) (Upper Quartile)

Q1 Q2 Q3
Lowest Value Highest Value

Figure 1.7: Whisker and box plot.

1.5.2.1 Importance of Box Plots

1. They are used to identify the location of a dataset using the median.
2. They are explored to determine the spread of the data through the length of the box,
interquartile range, and the length of the whiskers.
3. They provide a means of identifying possible skewness of the distribution of the dataset.
4. They detect suspected outliers in the dataset.
5. They are useful for comparing two or more datasets by drawing box plots for each dataset
and displaying the box plots on the same scale.

1.6 EXERCISES
1.1. Explain why statistics are necessary.
1.2. Describe the difference between a quantitative variable and qualitative variable.
1.3. List the four scales of measurements discussed in the chapter and discuss each with
relevant examples.
1.4. The total liters of petroleum products used for both importation and consumption (truck
out) in Nigeria in the second quarter of 2018 is summarized in Table 1.7.
(a) Represent the liters of petroleum products used for both importation and con-
sumption (truck out) in Nigeria in the second quarter of 2018 in a bar chart.
(b) Represent the total liters of petroleum products used for both importation and
consumption (truck out) in Nigeria in the second quarter of 2018 in a pie chart.
 1.6. EXERCISES 13

Table 1.7: Total liters of petroleum products (Source: NBS, 2018)

Months PMS AGO HHK ATK

April 1,781,949,720 309,988,876 0 14,700,325
May 1,670,390,264 387,434,670 0 64,186,515
June 1,340,777,749 408,153,067 43,790,592 121,501,304
Total 4,793,117,733 1,105,576,613 43,790,592 200,388,144
 15

CHAPTER 2

Introduction to R Software
R is a programming language designed for statistical analysis and graphics. It is based on S-plus
which was developed by Ross Ihaka and Robert Gentleman from the University of Auckland,
New Zealand, and R can be used to open multiple datasets. R is an open-source software which
can be downloaded at http://cran.r-project.org/. Other statistical packages are SPSS,
SAS, and Stata but they are not open source. Apart from this, there are large R group users
online that can provide real-time answers to questions, and that also contribute to add packages
to R. Packages increase the functions that are available for use, thus expanding the users’ abilities.
The R Development Core Team is responsible for maintaining the source code of R.

Why turn to R?
R software provides the following advantages.
1. R is free (meaning open-source software).
2. Any type of data analysis can be executed in R.
3. R includes advanced statistical procedures not yet present in other packages.
4. The most comprehensive and powerful feature for visualizing complex data is available in
R.
5. Importing data from a wide variety of sources can be easily done with R.
6. R is able to access data directly from web pages, social media sites, and a wide range of
online data services.
7. R software generates an easy and straightforward platform for programming new statistical
techniques.
8. It is simple to integrate applications written in other programming languages (such as
C++, Java, Python, PHP, Pentaho, SAS, and SPSS) into R.
9. R can operate on any operating system such as Windows, Unix, and Mac OSX. It can
also be installed on an iPhone. It is also possible to use R on an Android phone (see
https://selbydavid.com/2017/12/29/r-android/).
10. R offers a variety of graphic user interfaces (GUIs) if you are not interested in learning a
new language.
16 2. INTRODUCTION TO R SOFTWARE
Packages in R
Packages refer to the collections of R functions, data, and compiled code in a well-defined for-
mat.

2.1 HOW TO DOWNLOAD AND INSTALL R

Many operating systems such as Windows, Macintosh, and Linux can connect with R, since R
is a free download; and accommodates many different servers around the world. The software
can be downloaded from any of them. All available download links can be found on http:
//www.r-project.org/index.html by viewing the “Get Started” section on the front page
and click on “download R.” These mirrors are linked with the word “CRAN” which stands for
the Comprehensive R Archive Network. The CRAN provides the most recent version of R.
1. By choosing the mirror at the top of your computer screen, a list of the versions of R for
each operating system is displayed.
2. Click the R version that is compatible with your operating system.
3. If you decide to have the base version, just click the “download” link that displays on the
screen.

Installation of R
1. Under the “download” link, there is another link that provides instructions on how to
install it. These might be useful if you encounter problems during the installation process.
2. To install R, double-click on the executable file and follow the instruction on the screen.
The default settings are perfect. Figure 2.1 shows the first screen that comes up when
installing R on a Windows system.

2.2 USING R FOR DESCRIPTIVE STATISTICAL AND

PLOTS
R can be utilized to develop any statistical analysis. Different types of graphs can be constructed
using R statistical software. A range of standard statistical plots such as scatterplots, boxplots,
histograms, bar plots, pie charts, and basic 3D plots is provided by R. These basic types of plots
can be generated using a single function call in R. The R graphics are able to add several graphical
elements together to produce the final outcome. Apart from the conventional statistical plots, R
also generates Trellis plots through the package “lattice.” Special-purpose plots like lines, text,
rectangles, and polygons are available in the R software. Owing to its generality and feasibility,
production of graphical images that are above the normal statistical graphics, is possible. This
also makes it possible to generate figures that can visualize important concepts or useful insights.
The structure of R graphics has four distinct levels, namely: graphics packages, graphics systems,
 2.2. USING R FOR DESCRIPTIVE STATISTICAL AND PLOTS 17
a graphics engine, and graphics device packages (the full application will be discussed in the rest
of the book chapters).

STEPS TO INSTALL R

Figure 2.1: Click the appropriate operating system based on your computer.

Figure 2.2: Click on “R Sources” to get information on the latest version of R software.
18 2. INTRODUCTION TO R SOFTWARE

Figure 2.3: Click on “Packages” to get all available packages in R software.

Figure 2.4: Click on “Manuals” to obtain the R user guide.

 2.3. BASICS OF R 19

Figure 2.5: R manuals.

2.3 BASICS OF R
2.3.1 R IS VECTORIZED
Some of the funtions in R are vectorized. This indicates that the functions operate on the ele-
ments of a vector by acting on each element one after the other without necessarily undergoing
the looping process. Being vectorized allows writing codes in R easy to read, efficient, precise,
and concise. The following examples demonstrate vectorization in R.
1. Multiplication of vector by a constant.
R language allows multiplying (or dividing) a vector by a constant value. A simplex example
is:
x<-c(2, 4, 6, 8, 10)
x*3
[1] 6 12 18 24 30

x/3
[1] 0.6666667 1.3333333 2.0000000 2.6666667 3.3333333

2. Addition (or subtraction) of vectors.

x<-c(2, 4, 6, 8, 10)
y<-c(1, 2, 3, 4, 5)
20 2. INTRODUCTION TO R SOFTWARE
x+y
[1] 3 6 9 12 15
x-y
[1] 1 2 3 4 5

3. Multiplication of vectors.
x<-c(2, 4, 6, 8, 10)
y<-c(1, 2, 3, 4, 5)
x*y
[1] 2 8 18 32 50

4. Logical operations.
a<-(x>=4)
a
[1] FALSE TRUE TRUE TRUE TRUE

5. Matrix operations.
a <- matrix(1:4, 2, 2)
b <- matrix(0:3, 2, 2)
ab<-a*b
ab
[,1] [,2]
[1,] 0 6
[2,] 2 12

2.3.2 R DATA TYPES

There are different types of data in R, these include: scalars, vectors, matrices, data frames, lists,
coersion, etc.

2.3.2.1 Scalar
This is an atomic quantity that can hold only one value at a time. Scalars are the most basic data
types that can be used to construct more complex ones. Scalars in R can be numerical, logical,
and character (string). The following are examples of the different types of scalars in R.
1. Numerical
p<-10
q<-12
class(p)
[1] "numeric"
 2.3. BASICS OF R 21
class(q)
[1] "numeric"
class(p+q)
[1] "numeric"

2. Logical

t<-p<q # Is p less than q

s<-p>q # Is p greater than q
t
[1] TRUE
s
[1] FALSE
class(t)
[1] "logical"
class(s)
[1] "logical"
class(NA)
[1] "logical"

3. Character (string)

m<-"10"
n<-"12"
m
[1] "10"
n
[1] "12"
m+n # this is not the same as p and q in earlier example
Error in m + n : non-numeric argument to binary operator
class(m)
[1] "character"
class(n)
[1] "character"
class(as.numeric(m))
[1] "numeric"
class(as.character(p)) # to coerce this number into number
[1] "character"
class(as.character(p)) # to coerce this number to character
[1] "character"
22 2. INTRODUCTION TO R SOFTWARE
2.3.2.2 Vector
A vector is a sequence of data elements of the same basic type.
d<-c(1, 2, 3, 4 , 5, (6) # Numeric vector
class(d)
[1] "numeric"
e<-c("one", "two", "three", "four", "five", "six") # Character vector
class(e)
[1] "character"

f<-c(TRUE, FALSE, TRUE, TRUE, FALSE, TRUE) # Logical vector

class(f)
[1] "logical"
d;e;f
[1] 1 2 3 4 5 6
[1] "one" "two" "three" "four" "five" "six"
[1] TRUE FALSE TRUE TRUE FALSE TRUE

2.3.2.3 Matrix
A matrix is a collection of data elements arranged in a two-dimensional rectangular form. In
the same manner as a vector, the components in a matrix must be of the same basic type. An
example of a matrix with 2 rows and 4 columns is created below.
# fill the matrix with elements arranged by column in 2 rows
# and 4 columns
mat<-matrix(1:8, nrow=2, ncol=4, byrow= FALSE)
mat
[,1] [,2] [,3] [,4]
[1,] 1 3 5 7
[2,] 2 4 6 8
Alternatively, it is possible to have the elements of the matrix arranged by rows.
# fill the matrix with elements arranged by row in 2 rows
# and 4 columns
mat<-matrix(1:8, nrow=2, ncol=4, byrow= TRUE)
mat
[,1] [,2] [,3] [,4]
[1,] 1 2 3 4
[2,] 5 6 7 8
Braces [ ] can be used to reference elements in a matrix; this is similar to referencing elements
in vectors.
 2.3. BASICS OF R 23
mat[2,3] # refers to the element in the second row
# and third column
[1] 7
mat[,4] # refers to all elements in the fourth column
[1] 4 8
mat[1,] # refers to all elements in the first row
[1] 1 2 3 4

Table 2.1 contains the basic matrix operations and their respective meanings.

Table 2.1: Basic matrix operations

Function Meaning
t(x) Transpose of x
diag (x) Diagonal elements of x
%*% Matrix multiplication
solve (a,b) Solves a %*% x = b for x
rowsum (x) Sum of rows for a matrix-like object;
RowSums(x) is a faster version
rowMeans (x) Fast version of row means
colMeans (x) Fast version of column means

2.3.2.4 Data Frame

A data frame is more general than a matrix, in the sense that different columns can have different
basic data types. Data frame is one of the most common data type presented for data analysis.
Let us consider combining the vectors of numerical, character, and logical vectors to form a
data frame. Data frames have a special attribute called (row.names). Data frames are created
through read.table() or read.csv().

d<-c(1, 2, 3, 4 , 5, 6)
e<-c("one", "two", "three", "four", "five", "six")
f<-c(TRUE, FALSE, TRUE, TRUE, FALSE, TRUE)
data<-data.frame(d, e, f)
names(data)<-c("ID", "words", "state")
data
ID words state
1 1 one TRUE
2 2 two FALSE
24 2. INTRODUCTION TO R SOFTWARE
3 3 three TRUE
4 4 four TRUE
5 5 five FALSE
6 6 six TRUE

In addition, components from data frames can be extracted and this is similar to the extraction
of components in matrices, but after assigning names to each column makes it more flexible.

data$ID
[1] 1 2 3 4 5 6
data[1:2,]
ID words state
1 1 one TRUE
2 2 two FALSE

data[,3]
[1] TRUE FALSE TRUE TRUE FALSE TRUE

2.3.2.5 List
A list is a generic vector containing other objects. There is no restriction on data types or length
of the components. It is easier to work with lists that have named components. List is a special
type of vector. There are two characteristics of this type of vector—it can contain elements of
different classes and each element of a list can have a name.
Consider a list which contains a vector, matrix and data frame.

lst<-list(vector=d, matrix=mat, frame=data, count=10)

lst
$vector
[1] 1 2 3 4 5 6

$matrix
[,1] [,2] [,3] [,4]
[1,] 1 3 5 7
[2,] 2 4 6 8

$frame
ID words state
1 1 one TRUE
2 2 two FALSE
3 3 three TRUE
4 4 four TRUE
 2.3. BASICS OF R 25
5 5 five FALSE
6 6 six TRUE

$count
[1] 10

2.3.2.6 Factor
Factors are used to represent categorical data and can be ordered or unordered. We can regard a
factor as numerical vector when each of the integers contains a label. It is more appropriate using
factors with labels than using integers because factors are self-describing (e.g., variable with the
values of “Male” and “Female” is better than assigning value to them as 1 and 2).

y<-factor(c("yes", "yes", "yes", "no", "no"))

y
[1] yes yes yes no no
Levels: no yes

table(y) # tabulate the outcome

y
no yes
2 3

2.3.2.7 Coersion
This occurs when different objects are mixed in a vector; then every element in the vector has
the same class.

co<-c(50, "big") # character

co
[1] "50" "big"

co<-c(TRUE, (2) # numeric

co
[1] 1 2

co<-c(FALSE, (2) # numeric

co
[1] 0 2
26 2. INTRODUCTION TO R SOFTWARE
2.3.2.8 Date and Time
Date is represented by the date class and time is represented by the POSIXct class, a very large
integer. POSIXlt class is a list which stores a lot of useful metadata. The function as.Date() to
convert strings to dates.
x <- as.Date("2018-03-21")
x
[1] "2018-03-21"

# use as.Date() to convert strings to dates

dates <- as.Date(c("2018-03-21", "2018-10-04"))

# number of days between 03/21/2018 and 10/04/2018

days <- dates[2] - dates[1]
days
Time difference of 197 days
Besides, Sys.Date() returns today’s date, date() returns the current date and time, and strptime()
is used to write times in a different format as characters.
Table 2.2 contains the symbols the can be used with the format() function to print dates.

Table 2.2: Symbols to be used with format()

Symbol Meaning Example

%d Day as a number (0-31) 31-Jan
%a Abbreviated weekday Fri
%A Unabbreviated weekday Friday
%m Month 00-12
%b Abbreviated month Mar
%B Unabbreviated month March
%y Year without century 00-99
%Y Year with century 2018

To print today’s date using Sys.Date() function:

today <- Sys.Date() # print today's date
format(today, format="%B %d %Y")
[1] "October 04 2018"

x <- Sys.time()
 2.3. BASICS OF R 27
x
[1] "2018-10-04 11:35:15 WAT"

p<-as.POSIXlt (x)
names (unclass(p)) # unclass (p) is a list object
[1] "sec" "min" "hour" "mday" "mon" "year" "wday" "yday"
[9] "isdst" "zone" "gmtoff"
p$sec
[1] 15.44309
p$wday
[1] 4
p$mday
[1] 4

timeString<-"October 4, 2018 13:58"

h<-strptime(timeString, "%B %d, %Y %H: %M")
class(h)
[1] "POSIXlt" "POSIXt"
h
[1] "2018-10-04 13:58:00 WAT"

2.3.3 MISSING VALUES

Missing values are represented by NA (not available). Undefined values are represented by NaN
(not a number). NaN is also Na but the converse of this is not true. For example, we have:
mv<-c(1, 2, 3, NA , 5)
is.na(mv) # returns TRUE of mv is missing
[1] FALSE FALSE FALSE TRUE FALSE
In addition, we can exclude missing values from an analysis:
mv<-c(1, 2, 3, NA , 5)
mean(mv) # returns NA
[1] NA
mean(mv, na.rm=T) # returns average of 2.75
# by excluding the missing value
[1] 2.75
The function na.omit() returns the object with listwise deletion of missing values:
# create new dataset without missing data
d<-c(1, 2, 3, NA , 5, 6)
28 2. INTRODUCTION TO R SOFTWARE
e<-c("one", "two", "three", "four", "five", "six")
f<-c(TRUE, FALSE, TRUE, TRUE, FALSE, TRUE)
data1<-data.frame(d, e, f)
data1
d e f
1 1 one TRUE
2 2 two FALSE
3 3 three TRUE
4 NA four TRUE
5 5 five FALSE
6 6 six TRUE
newdata <- na.omit(data1)
newdata # delete all entries for row 4
# since it contains the missing value
d e f
1 1 one TRUE
2 2 two FALSE
3 3 three TRUE
5 5 five FALSE

2.3.4 DATA CREATION

c(…) : generic function to combine arguments with the default forming a vector; with recursive
= TRUE descends through lists combining all elements into one vector.
from:to : generates a sequence; “:” has operator priority.
x<-1:4+1
x
[1] 2 3 4 5

y<-10:14+6
y
[1] 16 17 18 19 20
seq(from, to) : generates a sequence by D specifies increment; length D specifies desired
length.
seq() function generates a sequence of numbers.
seq(from = 1, to = 1, by = ((to - from)/(length.out - 1)), length.out = NULL, along.with =
NULL, ...)
from, to: begin and end number of the sequence
by: step, increment (Default is 1)
 2.3. BASICS OF R 29
length.out: length of the sequence

# generate a sequence from -2 to 10, step 2

s<-seq(-2, 10, 2)
s
[1] -2 0 2 4 6 8 10

# generate a sequence with 15 evenly distributed from -2 to 10

s1<-seq(-2, 10, length.out=15)
s1
[1] -2.0000000 -1.1428571 -0.2857143 0.5714286 1.4285714 2.2857143
[7] 3.1428571 4.0000000 4.8571429 5.7142857 6.5714286 7.4285714
[13] 8.2857143 9.1428571 10.0000000

rep(x, times) : replicate x times; use each D to repeat “each” element of x each times;
rep(c(1, 2, 3, 4, 5), (2) # repeat 1 2 3 4 5 twice
[1] 1 2 3 4 5 1 2 3 4 5
rep(c(1, 2, 3, 4, 5), each = (2) # repeat each of 1 2 3 4 5 twice
[1] 1 1 2 2 3 3 4 4 5 5

replicate(3, c(1, 3, 6)) # repeat vector 1 3 6 thrice

[,1] [,2] [,3]

[1,] 1 1 1
[2,] 3 3 3
[3,] 6 6 6

array(x, dim =) : array with data x; specify dimensions like dim=c(3, 4, 2); elements of x
recycle if c is not long enough. We can give names to the rows, columns, and matrices in the
array by using the dimnames parameter.
# Create two vectors of different lengths.
vector1 <- c(1,2,3)
vector2 <- c(4,5,6,7,8,9)
column.names <- c("Col1","Col2","Col3")
row.names <- c("Row1","Row2","Row3")
matrix.names <- c("Matrix1","Matrix2")
arr <- array(c(vector1,vector2),dim = c(3,3,2),
dimnames = list(row.names,column.names,
matrix.names))
arr
30 2. INTRODUCTION TO R SOFTWARE

, , Matrix1

Col1 Col2 Col3

Row1 1 4 7
Row2 2 5 8
Row3 3 6 9

, , Matrix2

Col1 Col2 Col3

Row1 1 4 7
Row2 2 5 8
Row3 3 6 9

gl(n, k, length=n*k, labels = 1:n) : generates levels (factors) by specifying the pattern of their
levels; k is the number of levels, and n is the number of replications.

expand.grid() : a data frame from all combinations of the factors variables.

expand.grid(height = seq(1.5, 2.2, 0.05),

weight = seq(50, 70, 5),sex = c("Male","Female"))

height weight sex

1 1.50 50 Male
2 1.55 50 Male
3 1.60 50 Male
4 1.65 50 Male
5 1.70 50 Male
6 1.75 50 Male
7 1.80 50 Male
8 1.85 50 Male
9 1.90 50 Male
10 1.95 50 Male
11 2.00 50 Male
12 2.05 50 Male
13 2.10 50 Male
14 2.15 50 Male
15 2.20 50 Male
16 1.50 55 Male
17 1.55 55 Male
 2.3. BASICS OF R 31
18 1.60 55 Male
19 1.65 55 Male
20 1.70 55 Male
21 1.75 55 Male
22 1.80 55 Male
23 1.85 55 Male
24 1.90 55 Male
25 1.95 55 Male
26 2.00 55 Male
27 2.05 55 Male
28 2.10 55 Male
29 2.15 55 Male
30 2.20 55 Male
31 1.50 60 Male
32 1.55 60 Male
33 1.60 60 Male
34 1.65 60 Male
35 1.70 60 Male
36 1.75 60 Male
37 1.80 60 Male
38 1.85 60 Male
39 1.90 60 Male
40 1.95 60 Male
41 2.00 60 Male
42 2.05 60 Male
43 2.10 60 Male
44 2.15 60 Male
45 2.20 60 Male
46 1.50 65 Male
47 1.55 65 Male
48 1.60 65 Male
49 1.65 65 Male
50 1.70 65 Male
51 1.75 65 Male
52 1.80 65 Male
53 1.85 65 Male
54 1.90 65 Male
55 1.95 65 Male
56 2.00 65 Male
32 2. INTRODUCTION TO R SOFTWARE
57 2.05 65 Male
58 2.10 65 Male
59 2.15 65 Male
60 2.20 65 Male
61 1.50 70 Male
62 1.55 70 Male
63 1.60 70 Male
64 1.65 70 Male
65 1.70 70 Male
66 1.75 70 Male
67 1.80 70 Male
68 1.85 70 Male
69 1.90 70 Male
70 1.95 70 Male
71 2.00 70 Male
72 2.05 70 Male
73 2.10 70 Male
74 2.15 70 Male
75 2.20 70 Male
76 1.50 50 Female
77 1.55 50 Female
78 1.60 50 Female
79 1.65 50 Female
80 1.70 50 Female
81 1.75 50 Female
82 1.80 50 Female
83 1.85 50 Female
84 1.90 50 Female
85 1.95 50 Female
86 2.00 50 Female
87 2.05 50 Female
88 2.10 50 Female
89 2.15 50 Female
90 2.20 50 Female
91 1.50 55 Female
92 1.55 55 Female
93 1.60 55 Female
94 1.65 55 Female
95 1.70 55 Female
 2.3. BASICS OF R 33
96 1.75 55 Female
97 1.80 55 Female
98 1.85 55 Female
99 1.90 55 Female
100 1.95 55 Female
101 2.00 55 Female
102 2.05 55 Female
103 2.10 55 Female
104 2.15 55 Female
105 2.20 55 Female
106 1.50 60 Female
107 1.55 60 Female
108 1.60 60 Female
109 1.65 60 Female
110 1.70 60 Female
111 1.75 60 Female
112 1.80 60 Female
113 1.85 60 Female
114 1.90 60 Female
115 1.95 60 Female
116 2.00 60 Female
117 2.05 60 Female
118 2.10 60 Female
119 2.15 60 Female
120 2.20 60 Female
121 1.50 65 Female
122 1.55 65 Female
123 1.60 65 Female
124 1.65 65 Female
125 1.70 65 Female
126 1.75 65 Female
127 1.80 65 Female
128 1.85 65 Female
129 1.90 65 Female
130 1.95 65 Female
131 2.00 65 Female
132 2.05 65 Female
133 2.10 65 Female
134 2.15 65 Female
34 2. INTRODUCTION TO R SOFTWARE
135 2.20 65 Female
136 1.50 70 Female
137 1.55 70 Female
138 1.60 70 Female
139 1.65 70 Female
140 1.70 70 Female
141 1.75 70 Female
142 1.80 70 Female
143 1.85 70 Female
144 1.90 70 Female
145 1.95 70 Female
146 2.00 70 Female
147 2.05 70 Female
148 2.10 70 Female
149 2.15 70 Female
150 2.20 70 Female
rbind : combines arguments by rows for matrices, data frames, and others.
r1<-c(1:10)
r2<-c(1, 3, 5, 7, 9, 11,13,15, 17, 19)
rbind(r1,r2)
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
r1 1 2 3 4 5 6 7 8 9 10
r2 1 3 5 7 9 11 13 15 17 19
cbind : combines arguments by rows for matrices, data frames, and others.
c1<-c(1:10)
c2<-c(1,3,5,7,9, 11,13,15, 17, 19)
cbind(c1,c2)
c1 c2
[1,] 1 1
[2,] 2 3
[3,] 3 5
[4,] 4 7
[5,] 5 9
[6,] 6 11
[7,] 7 13
[8,] 8 15
[9,] 9 17
[10,] 10 19
 2.4. BASIC OPERATIONS IN R 35
2.3.5 DATA TYPE CONVERSION
Data type can be changed in R. Adding a character string to a numeric vector converts all the
elements in the vector to character. These functions can be used in the conversion of a data type;
see Table 2.3.
is.numeric(), is.character(), is.vector(), is.matrix(), is.data.frame(),
as.numeric(), as.character(), as.vector(), as.matrix(), as.data.frame().

Table 2.3: Functions

One Long Vector Matrix

Vector c(x,y) cbind(x,y); rbind(x,y)
Matrix as.vector (mymatrix)
Data frame as.matrix(myframe)

2.3.6 VARIABLE INFORMATION

1. is.na(x), is.null(x), is.array(x), is.data.frame(x), is.numeric(x),
is.complex(x), is.character(x) : test for types; or a complete list, use method
(is).
2. length(x) : number of elements in x.
3. dim(x) : retrieve for set the dimension of an object; dim<-c(3,2).
4. dimnames(x) : retrieve or set the dimension of an object.
5. nrow(x) : number of rows; NROW(x) is the same but treats a vector as a one-row matrix.
6. class(x) : get or set the class attribute of x; class(x)<-"myclass".
7. (unclass(x)) : remove the class attribute of x.
8. attr(x,which) : get or set the attribute which of x.
9. attribute(obj) : get or set the list of attributes obj.

2.4 BASIC OPERATIONS IN R

For a list of numbers in memory, basic operations can be performed. These operations can be
used to quickly perform a large number of calculations with a single command. It is obvious
that if an operation is performed on more than one vector it is often necessary that the vectors
all contain the same number of entries. In this section, we will focus on the subsetting, control
36 2. INTRODUCTION TO R SOFTWARE
structures, build-in functions, user written function, packages and the working directory, and R
script.

2.4.1 SUBSETTING
This is an operator that can be used to extract subsets of R objects, e.g., [ always returns an
object of the same class as the original object. With one exception, it can be used to select more
than one element. [[ is used to extract elements of a list or a data frame. $ is used to extract
elements of a list or data frame by name. The semantics of $ are similar to [[.
In addition, an element of a vector v is assigned an index by its position in the sequence,
starting with 1. The basic function for subsetting is [ ]. v[1] is the first element; v[length(v)]
is the last. The subsetting function takes input in many forms.
# Example 1

sub<-c("a", "b", "c", "d", "e", "f")

sub[4]
[1] "d"

sub[2:5]
[1] "b" "c" "d" "e"
sub>"c"
[1] FALSE FALSE FALSE TRUE TRUE TRUE

sub[sub>"c"]
[1] "d" "e" "f"

# Example 2

sub1<-matrix(1:9,3,3)
sub1
[,1] [,2] [,3]
[1,] 1 4 7
[2,] 2 5 8
[3,] 3 6 9
sub1[1,3]
[1] 7
sub1[2,] # entire elements in the second row
[1] 2 5 8
sub1[,1] # entire elements in the first column
[1] 1 2 3
 2.4. BASIC OPERATIONS IN R 37
# Example 3
sub<-c("a", "b", "c", "d", "e", "f")
v <- c(1, 3, 6)
sub[v]
[1] "a" "c" "e"
v[1:3]
[1] "a" "b" "c"

sub<-c("a", "b", "c", "d", "e", "f")

v <- c(1, 3, 6)
sub[v]
[1] "a" "c" "f"
sub[1:3]
[1] "a" "b" "c"

# Example 4
sub1<-matrix(1:9,3,3)
sub1[1,3]
[1] 7
sub1[1,2, drop = FALSE] # return as a matrix of 1 by 1
[,1]
[1,] 4
sub1[1,2, drop = TRUE] # return as a single element
[1] 4
sub1[1,, drop = TRUE] # return as a vector
[1] 1 4 7

# Example 5
y<-list(w = 1:3, x = 0.5, z ="gender")
y[1]

$w
[1] 1 2 3

y$w
[1] 1 2 3

y$x
38 2. INTRODUCTION TO R SOFTWARE
[1] 0.5

y$z
[1] "gender"

2.4.2 CONTROL STRUCTURES

Control structures allow one to control the flow of execution of a script typically inside of a
function. These control structures cannot be used while working with R interactively. We shall
consider some common control structures in this sections. These include: conditional (if-else),
for-loop, repeat-loop, and while-loop.

2.4.2.1 Conditional
if (condition) {
# do something
} else {
# do something else
}

# Example 6

x <- 1:20
if (sample(x, (1) <= 10) {
print("x is less than 10")
} else {
print("x is greater than 10")
}
[1] "x is greater than 10"

2.4.2.2 For-Loop
Loops are used in programming to repeat a specific block of code. A loop works on an iterable
variable and assigns successive values until the end of a sequence.
# Example 7

for (i in 1:10) {
print(i*2)
}

[1] 2
[1] 4
[1] 6
 2.4. BASIC OPERATIONS IN R 39
[1] 8
[1] 10
[1] 12
[1] 14
[1] 16
[1] 18
[1] 20

# Example 8

x <- c("A", "B", "C", "D")

for (i in seq(x)) {
print(x[i])
}
[1] "A"
[1] "B"
[1] "C"
[1] "D"

Alternatively,

for (i in 1:4) print(x[i])

[1] "A"
[1] "B"
[1] "C"
[1] "D"

2.4.2.3 Repeat-Loop
A repeat-loop is used to iterate over a block of code multiple number of times. There is no
condition check in a repeat-loop to exit the loop. There should be an explicit condition within
the body of the loop and then use a break statement to terminate the loop.
# Example 9

r <- 1
repeat {
print(r)
r = r+1
if (r == 10){
break
}
40 2. INTRODUCTION TO R SOFTWARE
}
[1] 1
[1] 2
[1] 3
[1] 4
[1] 5
[1] 6
[1] 7
[1] 8
[1] 9

2.4.2.4 While-Loop
While loops are used to loop until a specific condition is met.
# Example 10
w <- 1
while (w < 10) {
print(w)
w = w+1
}

[1] 1
[1] 2
[1] 3
[1] 4
[1] 5
[1] 6
[1] 7
[1] 8
[1] 9

2.4.3 BUILT-IN FUNCTIONS IN R

There are a lot of functions in R. We will concentrate on the numeric, character, and statisti-
cal functions that are commonly used in creating or recoding variables. The tables below show
various categories of built-in functions and their descriptions.
 2.4. BASIC OPERATIONS IN R 41
2.4.3.1 Numerical Functions
See Table 2.4.
Table 2.4: Numerical functions

Function Description Example

abs(x) Absolute value abs(-0.127)
[1] 0.127
sqrt(x) Square root sqrt(200)
[1] 14.14214
ceiling(x) Numbers are rounded up to ceiling(100.125)
the nearest integer [1] 101
floor(x) Numbers are rounded down floor(100.125)
to the nearest integer [1] 100
trunc(x) Truncation trunc(100.125)
[1] 100
round(x, digits=n) Rounding up round(100.125, digits=1)
[1] 100.1
signif(x, digits=n) Significant figure signif(100.125, digits=1)
[1] 100
cos(x), sin(x), tan(x) Trigonometric cos(90)
[1] -0.4480736
log(x) Natural logarithm log(90)
[1] 4.49981
log10(x) Common logarithm log10(90)
[1] 1.954243
exp(x) Exponential exp(0.125)
[1] 1.133148
42 2. INTRODUCTION TO R SOFTWARE
2.4.3.2 Character Functions
See Table 2.5.
Table 2.5: Character functions

Function Description Example

substr(x, start=n1, stop=n2) Extract or replace substrings x <- "peter" > substr(x, 2, 4)
in a character vector. [1] "ete"
grep(pattern, x, ignore.case = Search for the pattern in x. grep("Z", c("X","Y","Z"),
FALSE, fixed = FALSE) If fixed = FALSE then the fixed=TRUE)
pattern is a regular expression. [1] 3
If fixed = TRUE then the pat-
tern is a text string. Returns
matching indices.
sub(pattern, replacement, x, Find the pattern in x and re- sub("\\s",".","Lord God")
ignore.case = FALSE, fixed = place with replacement text. [1] "Lord.God"
FALSE) If fixed = FALSE then the
pattern is a regular expression.
If fixed = T then the pattern is
a text string.
strsplit(x, split) Split the elements of character strsplit("faith", "")
vector x at split. [[1]]
[1] "f " "a" "i" "t" "h"
paste(..., sep="") Concatenate strings after paste("child",1:3,sep=" ")
using sep string to seperate [1] "child 1" "child 2" "child
them. 3"
 2.4. BASIC OPERATIONS IN R 43
2.4.3.3 Statistical Probability Functions
See Table 2.6.
Table 2.6: Statistical probability functions

Function Description Example

dnorm(x) Normal density function (by dnorm(0.5)
default m=0 sd=1) [1] 0.3520653
pnorm(q) Cumulative normal probabil- pnorm(1.96)
ity for q (area under the nor- [1] 0.9750021
mal curve to the left of q)
qnorm(p) Normal quantile. Value at the qnorm(.75) # 75th percentile
p percentile of normal distri- [1] 0.6744898
bution
rnorm(n, m=0,sd=1) n random normal deviates #50 random normal variates
with mean m and standard with mean=2.5, sd=0.05
deviation sd x <- rnorm(50, m=2.5,
sd=0.05)
dbinom(x, size, prob) Binomial distribution where # prob of 0 to 5 heads of fair
pbinom(q, size, prob) size is the sample size and coin out of 10 flips
qbinom(p, size, prob) prob is the probability of a dbinom(0:5,10,0.5)
rbinom(n, size, prob) head (pi)
# prob of 5 or less heads of
fair coin out of 10 flips
pbinom(5, 10, .5)
dpois(x, lamda) Poisson distribution with #probability of 0, 1, or 2
ppois(q, lamda) m=std=lamda events with lamda=4
qpois(p, lamda) dpois(0:2, 4)
rpois(n, lamda) [1] 0.01831564 0.07326256
0.14652511

# probability of at least 3
events with lamda=4
1-ppois(2,4)
[1] 0.7618967
dunif(x, min=0, max=1) Uniform distribution, follows #5 uniform random variates
punif(q, min=0, max=1) the same pattern as the nor- y <- runif(5)
qunif(p, min=0, max=1) mal distribution above y
runif(n, min=0, max=1) [1] 0.14192169 0.13701585
0.06418781 0.58657717
0.20230663
44 2. INTRODUCTION TO R SOFTWARE
2.4.3.4 Other Statistical Functions
See Table 2.7.
Table 2.7: Other statistical functions

Function Description Example

mean(x, trim=0, na.rm = Mean of object x # trimmed mean, removing any #
FALSE) missing values and 5% #of highest
and lowest scores
mx <- mean(x,trim=.05,na.
rm=TRUE)
sd(x) Standard deviation of ob- x<-1:10
var(x) ject(x), var (x) for variance, sd(x)
mad(x) and mad(x) for median abso- [1] 3.02765
lute deviation. var(x)
[1] 9.166667
mad(x)
[1] 3.7065
median(x) Median x<-1:10
median(x)
[1] 5.5
quantile(x, probs) Quantiles where x is the nu- # 10th and 75th percentiles of x
meric vector whose quantiles y <- quantile(x, c(0.1,0.75))
are desired and probs is a nu- y
meric vector with probabilities 10% 75%
in [0,1] 1.90 7.75
range(x) Range range(x)
[1] 1 10
sum(x) Sum sum(x)
[1] 55
diff(x, lag=1) Lagged differences, with lag diff(x, lag=1)
indicating which lag to use [1] 1 1 1 1 1 1 1 1 1
min(x) Minimum min(x)
[1] 1
max(x) Maximum max(x)
[1] 10
scale(x, center=TRUE, Column center or standardize
scale=TRUE) a mat
 2.4. BASIC OPERATIONS IN R 45
2.4.3.5 Other Useful Functions
See Table 2.8.
Table 2.8: Other useful functions

Function Description Example

seq(from, to, by) Generate a sequence gen <- seq(1,10,2)
gen
[1] 1 3 5 7 9
rep(x, ntimes) Repeat x n times r <- rep(1:3, 2)
r
[1] 1 2 3 1 2 3
cut(x, n) Divide continuous variable in y <- cut(x, 5)
factor with n levels

2.4.4 USER-WRITTEN FUNCTIONS

R is so flexible to accomodate the user to add functions and some functions in R are functions
of functions.
The structure of a function is given below:
my.function <- function(arg1, arg2, ... ){
statements
return(object)
}
# Example 11
my.result <- function(x,npar=TRUE,print=TRUE){
if (!npar) {
center <- mean(x); spread <- sd(x)
} else {
center <- median(x); spread <- mad(x)
}
if (print & !npar) {
cat("Mean=", center, "\n", "SD=", spread, "\n")
} else if (print & npar) {
cat("Median=", center, "\n", "MAD=", spread, "\n")
}
result <- list(center=center,spread=spread)
46 2. INTRODUCTION TO R SOFTWARE
return(result)
}

# actualizing the function

set.seed(100)
rv1 <- rbinom(10, 3, 0.5)
rv2 <- my.result(rv1)
Median= 1
MAD= 0

2.4.5 IMPORTING, REPORTING, AND WRITING DATA

We will discuss how to import packages and changing directories for reading and writing local
flat files, reading and writing local Excel files, connection interfaces, reading XML/HTML files,
reading and writing to JSON, connecting to a database, and the textual format.

2.4.5.1 Packages
Anything you may think of doing with R, there is a tendency that a package has been written
to execute it. The list of packages can be found in the official repository CRAN: http://cran
.fhcrc.org/web/packages/. The installation of any package of your choice is very easy in R,
you may use the command: install.packages("packagename").

2.4.5.2 Working Directory and R Script

Set the working directory from the menu if using the R-gui (Change dir…) or from the R
command line:

setwd("C:\\MyWorkingDirectory ")
setwd("C:/MyWorkingDirectory ") # can use forward slash
setwd(choose.dir()) # open a file browser
getwd() # returns a string with the current working directory

Besides, in order to have access and see a list of the files in the current directory, you can use the
R command line:

dir () # returns a list of strings of file names

dir (pattern = ".R$ ") # list of files ending in ".R "
dir ("C:\\Users") # show files in directory C:\Users

Run a script

source("helloworld.R")
 2.4. BASIC OPERATIONS IN R 47
2.4.5.3 Reading and Writing Local Flat Files
Local flat files can be read by using the following: read.table, read.csv, readLines. The acronomy
CSV stands for “comma separated values.”
For the purpose of illustration, we will be working with the Nigerian exchange rate dataset
from the Central Bank of Nigeria website. To procure this dataset, please visit the following link:
https://www.cbn.gov.ng/Functions/export.asp?tablename=exchange.
Right click and select “save as” to save it to your local directory. The code samples in the
remainder of this section assumes that you have set your working directory to the location of the
exchrt.csv file.
Reading and writing local flat files:

exch.rt<-read.csv("G:/BRA Research/BUSINESS STATISTICS IN R/exchrt.csv")

head(exch.rt, 7)
Rate.Date Currency Rate.Year Rate.Month Buying.Rate Central.Rate Selling.Rate
1 10/4/2018 US DOLLAR 2018 October 305.4000 305.9000 306.4000
2 10/4/2018 POUNDS STERLING 2018 October 396.6230 397.2723 392.9217
3 10/4/2018 EURO 2018 October 351.1795 351.7544 352.3294
4 10/4/2018 SWISS FRANC 2018 October 307.6458 308.1495 308.6532
5 10/4/2018 YEN 2018 October 2.6766 2.6810 2.6854
6 10/4/2018 CFA 2018 October 0.5178 0.5278 0.5378
7 10/4/2018 WAUA 2018 October 424.2396 424.9341 425.6287

col.head<-readLines("G:/BRA Research/BUSINESS STATISTICS IN R/exchrt.csv", 1)

col.head
[1] "Rate Date,Currency,Rate Year,Rate Month,Buying Rate,Central Rate,Selling Rate"

To write local flat files, you can used the functions such as write.table, write.csv, and
writeLines. Meanwhile, it is advisable for someone to take note of the parameters before reading
or writing a file.
write.table(exch.rt, "new_exchrt.csv")
Conversely, the read.table is the one of the commonest functions for reading a data. Some
of the arguments of read.table() are described below:
function (file, header = FALSE, sep = "", quote = "\"", dec = ".",
numerals = c("allow.loss", "warn.loss", "no.loss"), row.names,
col.names, as.is = !stringsAsFactors, na.strings = "NA",
colClasses = NA, nrows = -1, skip = 0, check.names = TRUE,
fill = !blank.lines.skip, strip.white = FALSE,
blank.lines.skip = TRUE, comment.char = "#", allowEscapes = FALSE,
flush = FALSE, stringsAsFactors = default.stringsAsFactors(),
fileEncoding = "", encoding = "unknown", text, skipNul = FALSE)
48 2. INTRODUCTION TO R SOFTWARE
1. file is the name of a file, or a connection.
2. header a logical value indicating the variables of the first line.
3. sep is a string indicating how the columns are separated.
4. colClasses is a character vector indicating the class of each column in the dataset.
5. nrows are the number of rows in the dataset.
6. comment.char is character string indicating the comment character.
7. skip is the number of lines to skip from the beginning.
8. stringsAsFactors determines whether character variables be coded as factors.

2.4.5.4 Reading and Writing Excel Files

The functions read.xlsx or read.xlsx2 can be used to read an excel file while write.xlsx or
write.xlsx2 can also be used write. This is relatively fast but unstable. First, make sure that you
install the Excel spreadsheet from the library. We can use the following commands to install
the xlsx library:
# install the xlsx library
install.packages("xlsx")

See Fig. 2.6.

The command in Fig. 2.6 prompted this image and then click OK. Second, we have to load the
Excel spreadsheet into the library with the following command lines:
# load the library
library(xlsx)

Alternatively, you can load library through interface:

Click menu “Packages”
Then select “Load package…”
And then select “xlsx”
Click OK.

2.4.5.5 Connection Interfaces

We may not necessarily need to access connection interface directly; it may be made to files or
to other channels such as the following.
 2.4. BASIC OPERATIONS IN R 49

Figure 2.6: Install packages.

1. file opens a connection to a file.

2. gzfile opens a connection to a file compressed with gzip.

3. bzfile opens a connection to a file compressed with bzip2.

4. url opens a connection to a web page.

2.4.5.6 Connect to a Database

For good data management and efficient warehouse, it is required to store our data in a relational
database. A well-designed database architecture improve the robustness on how data can be
mined to give a rigourous analysis. To perform statistical computing we will need very advanced
and complex Sql queries. R can connect easily to many relational databases like MySql, Oracle,
Sql server, among others. One of the advantages of using a database is that the records from the
relational database is fetched as a data frame. Hence, it makes data crunch more easily to analyze
with the aid of sophisticated packages and functions. R has some packages that connect to R,
and these include: JSONPackages, RmySQL, RpostresSQL, RODBC, RMONGO, etc. In this book, we
will use MySql as our reference database for connecting to R.
50 2. INTRODUCTION TO R SOFTWARE
We can install the "RMySQL" package in the R environment using the following command:
install.packages("RMySQL")
library(RMySQL)
After installation we will create a connection object in R to connect to the database. It requires
the username, password, database name, and host name as input:
# Create a connection Object to MySQL database.
# assume our database named "my.database"
# that comes with MySql installation.
mysql.connect = dbConnect(MySQL(), user = 'root', password = '',
dbname = ' my.database ', host = 'localhost')

# to list the tables available in this database

dbListTables(mysql.connect)

2.5 DATA EXPLORATION

This is the initial step that is compulsory to perform before the commencement of the proper
analysis. This process involves data visualization (identify summaries, structure, relationships,
differences, and abnormalities in the data), descriptive statistics, and inferential statistics.
Oftentimes no elaborate analysis is necessary as all the important conclusions required for
a decision are evident from simple visual examination of the data. Other times, data exploration
will be used to help guide the data cleaning, feature selection, and sampling process.
Look at the first three rows; we have:
head(exch.rt[1:7], 3)
Rate.Date Currency Rate.Year Rate.Month Buying.Rate Central.Rate Selling.Rate
1 10/4/2018 US DOLLAR 2018 October 305.4000 305.9000 306.4000
2 10/4/2018 POUNDS STERLING 2018 October 396.6230 397.2723 392.9217
3 10/4/2018 EURO 2018 October 351.1795 351.7544 352.3294

For a more specific sample of the data, assuming that we want to view the last three columns of
the data frame, and the first-two rows in the columns. The command lines are:
# View the last three columns of the data frame,
# and the first two rows in those columns
head(exch.rt[,5:7],2)
Buying.Rate Central.Rate Selling.Rate
1 305.400 305.9000 306.4000
2 396.623 397.2723 392.9217
The str() function can be used to summarize the data frame. Let us look at the summary of the
exchange rate data.
 2.5. DATA EXPLORATION 51
str(exch.rt)
'data.frame': 38728 obs. of 7 variables:
$ Rate.Date : Factor w/ 4126 levels "1/10/2002","1/10/2003",..:
623 623 623 623 623 623 623 623 623 623 ...
$ Currency : Factor w/ 27 levels "CFA","CFA ","DANISH KRONA",..:
21 12 5 19 25 1 23 27 13 18 ...
$ Rate.Year : int 2018 2018 2018 2018 2018 2018 2018
2018 2018 2018 ...
$ Rate.Month : Factor w/ 15 levels "8","April","August",..:
14 14 14 14 14 14 14 14 14 14 ...
$ Buying.Rate : num 305.4 396.62 351.18 307.65 2.68 ...
$ Central.Rate: num 305.9 397.27 351.75 308.15 2.68 ...
$ Selling.Rate: num 306.4 392.92 352.33 308.65 2.69 ...

2.5.1 DATA EXPLORATION THROUGH VISUALIZATION

In this section, we shall limit our scope to few number of visual data exploration methods. This
includes, a bar chart, pie chart, and box-plot, etc.

2.5.1.1 Bar Chart

A bar chart is good at visualizing a categorical data, and it is made up of rectangular bars with
heights or lengths proportional to the values that they represent (see Fig. 2.7). The bars can be
plotted vertically or horizontally.

# Simple bar chart

age <- c(22, 20, 14, 16, 18)
barplot(age, main = "Distribution of Age of Students", xlab = "Age",
ylab = "Name", names.arg = c("Ade", "Ope", "Olu", "John", "Ayo"),
col = "green", horiz = TRUE)

2.5.1.2 Pie Chart

A pie chart depicts the proportions and percentages between categories, by dividing a circle into
proportional segments. Each arc length represents a proportion of each category. The full circle
represents the total sum of all the data and it is equal to 100%. It gives a reader an insight of
the proportional distribution of a dataset. However, a bar chart or dot plot are preferred to a pie
chart because of an inability to judge length more accurately.
Pie charts are created with the function pie(x, labels=) where x is a non-negative numeric
vector indicating the area of each slice and labels= notes a character vector of names for the
slices (see Fig. 2.8).
52 2. INTRODUCTION TO R SOFTWARE
Distribution of Age of Students

Ayo
John
Name
Olu
Ope
Ade

0 5 10 15 20

Age

Figure 2.7: Bar chart.

# Age distribution of students

age <- c(22, 20, 14, 16, 18)
name <- c("Ade", "Ope", "Olu", "John", "Ayo")
age.pct <- round(age/sum(age)*100)
name <- paste(name, age.pct) # add percents to labels
name <- paste(name,"%",sep="") # ad % to labels
pie(age, labels = name, col=rainbow(length(name)),
main="Age Distribution")

2.5.1.3 Box-Plot Distributions

A box plot is very uselful in showing the distribution or pattern between numerical and cate-
gorical attributes (see Fig. 2.9). Box plots can be created for individual variables or for variables
by group. The format is boxplot(x, data=), where x is a formula and data= denotes the data
frame providing the data. A typical example of a formula is ygroup where a separate box plot
for numeric variable y is generated for each value of group. Add varwidth=TRUE to make box
 2.6. EXERCISES 53
Age Distribution

Ope 22%

Ade 24%
Olu 16%

Ayo 20%
John 18%

Figure 2.8: Pie chart.

plot widths proportional to the square root of the samples sizes. Add horizontal=TRUE to
reverse the axis orientation.
# simple box plot
age <- c(22, 20, 14, 16, 18)
gender= c("Female", "Male", "Male", "Male", "Female")
mydata<-data.frame(age, gender)
boxplot(age~gender,data=mydata, labels=c("Male","Female"),
xlab="Gender", ylab="Age", col=c("red","blue"))

2.6 EXERCISES
2.1. What are the advantages of R software?
2.2. With the aid of examples, explain the term vectorization in R.
2.3. Enumerate different types of data in R.
2.4. What do you understand by control structure in R and list common control structures
in R?
2.5. What are the methods to use in visualizing data?
54 2. INTRODUCTION TO R SOFTWARE

22
20
Age

18
16
14

Female Male

Gender

Figure 2.9: Box plot distributions.

 55

CHAPTER 3

Descriptive Data
This chapter presents different statistical measures that can be employed to provide descriptive
analysis of business-related data. In this chapter we will define the statistical measures of central
tendency and dispersion. Explanations of these measures of central tendency and dispersion
using example data are provided in this chapter.

3.1 CENTRAL TENDENCY

Central tendency concerns the measures that may be used to represent the middle value of a set
of data; that is, a value that represents the center of that data. The three common measures of
central tendency are mean, median, and mode.

The Mean
Mean or arithmetic mean: this is the average of a set of observations. It is the aggregate of all
observations divided by the number of observations in the data set. For example, observations
are denoted by x1 ; x2 ; x3 ; : : : ; xn . The sample mean is represented by x which is expressed as
follows.
Mean of a sample,
Pn
xi x1 C x2 C x3 C


C xn
x D i D1 D ; (3.1)
n n
P
where is the summation symbol. The summation covers all of the data points.
As an example, if you have 9 houses on your block and the number of people living in the
houses is: 2, 1, 6, 4, 1, 1, 5, 1, and 3 and you want to know the average (mean) number of people
living in the house, you simply add the numbers together and get 24 and divide that by 9 to get
2.67 people per house.
If the observation set covers a whole population, the symbol  (the Greek letter mu) is
used to represent the mean of the entire population. In addition, N is used instead of n to denote
the number of elements. The mean of the population is specified as follows.
Mean of a population:
PN
xi
 D i D1 : (3.2)
N
Mean is the most commonly used measure of central tendency. In addition, the mean is
relied on information contained in all the observations in the dataset.
56 3. DESCRIPTIVE DATA
Characteristics of Mean

• It summarizes all information in the dataset.

• It captures the average of all the observations.

• It is a single point viewed as the point where all observations are concentrated.

• All observations in the dataset would be equal to the mean if each observation has the
same size or figure.

• It is sensitive to extreme observations (outliers).

The Median
This is a special point because it is located in the center of the data, implying that half the data
lies below it and half above it. The median is defined as a measure of the location or centrality
of the observations. This is just an observation lying in the middle of the set.
In order to find the median by hand, given a small data set like this one, you need to
reorder the dataset in numerical order:

1; 1; 1; 1; 2; 3; 4; 5; 6:

If you divide the number of data points by two and round up to the next integer, you will find
the median. So for our example:
9=2 D 4:5:

Round up the answer is 5, so the 5th data point is the median, which is 2.
This works nicely if you have an odd number of data points, but if you have an even
number, then there would be two middle numbers. To resolve this dilemma, you take those two
data points and average them. So let’s add a house with 4 people. That means that the middle
two numbers would be 2 and 3. So averaging them gives you 2.5 as the median.

Characteristics of Median

• It is an observation in the center of the dataset.

• One half of the data lie above this observation.

• One half of the data lie below the observation.

• It is resistant to extreme observations.

 3.2. MEASURE OF DISPERSION 57
The Mode
The mode denotes the value that has the highest occurrence. Put differently, the mode is the value
with most frequency in the dataset. In our example, the mode would be 1 with four occurrences.

Example 3.1
Calculate the mean, median, and mode of the following observations: 2, 6, 8, 7, 5, 3, 2, and 2.

Solution
Mean of the observations, x D 2C6C8C7C5C3C2C2
7
D 358
D 4:4.
To solve for the mean of the observations in R, the following commands are expressed.
x <- c (2,6,8,7,5,3,2,2)
mean(x)
[1] 4.4
Median of the observation requires the re-arrangement of the number from smallest to largest:
2, 2, 2, 3, 5, 6, 7, and 8. There are eight observations, so the median is the value in the middle,
that is, in the fourth and fifth position. Those values are 3 and 5; so you add the two numbers
together and divide the outcome by 2 in order to get the median.
Median is 3C5 2
D 4 because 3 and 5 are the center of the dataset. In R, the command is:
median (x)
[1] 4
Mode in the observation set is 2 because it appears three times while other values occur once.
In R, the command is:
mode(x)
[1] 2

3.2 MEASURE OF DISPERSION

There are different measures of variability or dispersion. These measure the degree of data spread-
out from the mean of a dataset. These measures include range, interquartile range, variance, and
standard deviation.
(a) Range: this is the difference between the largest observation and the smallest observation.
The range in Example 3.1 is largest number—smallest number D 8 2 D 6.
(b) Interquartile range: this can be defined as the difference between the upper quartile (Q3 )
and lower quartile (Q1 ). Thus, the interquartile range is a measure of the difference be-
tween the higher half of your data and the lower half. To do this, we order the data and
separate the lower half of the data from the higher half and calculate the medians of the
58 3. DESCRIPTIVE DATA

Figure 3.1: The R foundation for statistical computing.

higher half and lower half and then subtract the lower half median from the higher half
median to get the interquartile range.
Using data from our first example, 1, 1, 1, 1, 2, 3, 4, 5, 6, we first need to split the data
set in two. So we look to the middle and see that the number is odd, so we throw out
the median (3.2). Then we find the median of 3, 4, 5, and 6 which is 4.5 and subtract the
median of 1, 1, 1, 1 which is 1 to give us 3.5 as the interquartile range. So we see that the
spread between the lower half of this data and the higher half is quite significant.

(c) Variance: this is the average squared deviation of the data points from their mean.

(d) Standard deviation: this can be defined as the square root of the variance of the entire
dataset. In financial analysis, the standard deviation is explored to capture the volatility as
well as the risk associated with financial variables.

(e) The variance, is calculated by calculating the difference from the mean for each data point,
then squaring each of them and then adding them all together and then divided by the
number of items in the dataset minus one:
Pn
i D1 .xi x/2
:
n 1
 3.2. MEASURE OF DISPERSION 59
(f ) The standard deviation of a sample is the square root of the sample variance while the
standard deviation of a population is the square root of the variance of the population. The
formulae for these two forms of standard deviation are expressed below:
s
Pn
p
2 i D1 .xi x/2
Sample standard deviation: s D s D
n 1
s
Pn
p
i D1 .xi /2
Population standard deviation:  D  2 D :
n 1

Statisticians prefer working with the variance because its mathematical features make
computations easy, whereas applied statisticians like to work with the standard deviation be-
cause of its easy interpretation.
Let’s calculate the variance and standard deviation of the data in Table 3.1. In order to
compute this, a table is used for simplicity (before exploring R software to do everything within
a minute).

Table 3.1: Observations for X variables

x 𝑥𝑖 – 𝑥̅ (𝑥𝑖 – 𝑥̅ )2
2 -2.4 9
2 -2.4 9
2 -2.4 9
3 -1.4 4
5 0.6 0
6 1.6 1
7 2.6 4
8 3.6 9
Total 0 45

In reference to the above equation, the variance of the sample equals the sum of the third
column in the Table 3.1, 45, divided by n 1: s 2 D 45 7
D 4:44. The standard deviation is the
p
square root of the variance: s D 3:44 D 2:335 or putting in two decimal places, s D 2:34.
There is a shortcut formula for the sample variance that is equivalent to the formula above
for variance: Pn Pn
2
2
2 i D1 xi i D1 xi =n
s D :
n 1
Solving the standard deviation and variance in Table 3.1 with the aid of R, the commands are:
60 3. DESCRIPTIVE DATA
x <- c(2,2,2,3,5,6,7,8)
var(x)
[1]4.44
sd(x)
[1] 2.34

3.3 SHAPES OF THE DISTRIBUTION—SYMMETRIC AND

ASYMMETRIC
(a) Symmetric distribution: a dataset or a population is symmetric if one side of the distribu-
tion of the observation reflects a mirror image of the other; and if the distribution of the
observations has only one mode, then the mode, median, and mean are the same.

(b) Asymmetric distribution: this is the kind of distributional pattern in which one side of the
distribution is not a mirror image of the other. In addition, in its data distribution, the
mean, median, and mode will not all be equal.
Additional attributes of a frequency distribution of a dataset are skewness and kurtosis.
Skewness: measures the degree of asymmetry of a distribution. A right skewness occurs when
the distribution stretches to the right more than it does to the left, while a left-skewed distribu-
tion is one that stretches asymmetrically to the left. Graphs depicting a symmetric distribution,
a right-skewed distribution, a left-skewed distribution, and a symmetrical distribution with two
modes, are presented in Fig. 3.2.
For a right-skewed distribution, the mean is to the right of the median, thus lying to the
right of the mode. The opposite is observed for left-skewed distribution (see Fig. 3.2). The calcu-
lation of skewness is reported by a number that may be positive, negative, or zero. Zero skewness
indicates a symmetric distribution. A positive skewness means a right-skewed distribution while
a negative skewness denotes a left-skewed distribution. Skewness could be different in terms of
their shape even if two distributions have the same mean and variance.
Kurtosis: this measures the peak level of distribution. The larger the kurtosis, the more peaked
will be the distribution. Its calculation is reported either as an absolute or a relative value. Ab-
solute kurtosis is usually a positive number. For a normal distribution, the absolute kurtosis is 3.
The value of 3 is used as the data point to compute relative kurtosis. Therefore, relative kurtosis
is the difference between the absolute kurtosis and 3:

Relative kurtosis D absolute kurtosis 3:

The relative kurtosis can be either negative, known as platykurtic indicating a flatter distribution
than the normal distribution or a positive kurtosis known as leptokurtic showing a more peaked
distribution than the normal distribution.
 3.3. SHAPES OF THE DISTRIBUTION—SYMMETRIC AND ASYMMETRIC 61

Mode Mode
Median Median

Mean Mean

Left-Skewed (Negative Skewness) Right-Skewed (Positive Skewness)

(a) Skewness (left and right)

Mean = Median = Mode

(b) Symmetric Distribution

Mode Mean Mode

Median
(c) Asymmetric Distribution

Figure 3.2: (a) Skewness (left and right), (b) symmetric distribution, and (c) asymetric distribu-
tion.
62 3. DESCRIPTIVE DATA
3.4 EXERCISES
3.1. Create X that includes all even numbers between 0 and 100.
3.2. Calculate the mean, mode, and median of X .
3.3. Calculate the dispersion measures (standard deviation and variance) for X .
3.4. Explain the characteristics of central tendency measures (mean, mode, and median).
3.5. Explain the two tools discussed here—kurtosis and skewness—for shapes of distribu-
tion.
 63

CHAPTER 4

Basic Probability Concepts

In this chapter, we shall discuss some basic concepts of probability, and solving the problem of
probability using venn diagram. The axioms and rules of probability shall be discussed. We shall
extend our discussion to conditional probabilities. Practical examples with R codes will be used
for illustration.

4.1 EXPERIMENT, OUTCOME, AND SAMPLE SPACE

As we have mentioned before, it is important to understand the definition of certain terms to
be able to use them successfully. So the first step in gaining an understanding of probability is
to learn the terminology and the rest will be a lot simpler.

4.1.1 EXPERIMENT
This is a measurement process that produces quantifiable results. Some typical examples of an
experiment are tossing of a die, tossing of a coin, playing of cards, measuring weight of students,
and recording growth of plants.

4.1.2 OUTCOME
This is a single result from a measurement. Examples of outcomes are getting a sum of 9 in a
tossing of two dice, turning up of head in a toss of a coin, selecting a spade in a deck of card,
and getting a weight above a threshold (say 50 kg).

4.1.3 SAMPLE SPACE

This is the set of all possible outcomes from an experiment and it is denoted as S. The sample
space of tossing a die is S D f1; 2; 3; 4; 5; 6g; the sample space of tossing a coin is denoted as
S D fH; T g. In addition, the sample space for the tossing of two dice is
8 9
ˆ
ˆ f1; 1g ; f1; 2g ; f1; 3g ; f1; 4g ; f1; 5g ; f1; 6g ; f2; 1g ; f2; 2g ; f2; 3g ; >
>
< =
f2; 4g ; f2; 5g ; f2; 6g ; f3; 1g ; f3; 2g ; f3; 3g ; f3; 4g ; f3; 5g ; f3; 6g ;
SD :
ˆ f4; 1g ; f4; 2g ; f4; 3g ; f4; 4g ; f4; 5g ; f4; 6g ; f5; 1g ; f5; 2g ; f5; 3g ; >
>
:̂ ;
f5; 4g ; f5; 5g ; f5; 6g ; f6; 1g ; f6; 2g ; f6; 3g ; f6; 4g ; f6; 5g ; f6; 6g
64 4. BASIC PROBABILITY CONCEPTS
4.2 ELEMENTARY EVENTS
Any subset of a sample set, empty set, and whole set inclusive is called an event. It is an
event with a single element taken from a sample space. In set theory, the elementary event
is a singleton set (a set that contains only one element) and it is denoted as E . In addi-
tion, elementary event can be written as E  S (where E is a proper subset of the sam-
ple space, but not equal to the sample space). Suppose we have n possible outcomes from
an experiment say E1 ; E2 ; E3 ; : : : ; En , then the space of elementary events which is the
sample space can be written as: S D fE1 [ E2 [ E3 [ : : : [ En g. That is the sample is the
union of all of the events. For instance, when tossing three coins, all possible outcomes are
fHHHg ; fHHTg ; fHTHg ; fTHHg ; fHTTg ; fTHTg ; fTTHg ; fTTTg and when rolling a fair die, all
possible outcomes are getting f1g ; f2g ; f3g ; f4g ; f5g ; f6g. Hence, each of the outcomes are ele-
mentary events.

4.3 COMPLEMENTARY EVENTS

These are events that cannot occur at the same time. Consider an event A, having its complement
denoted by A0 which are mutually exclusive and exhaustive. Hence, the sample space of the
experiment can be written as S D A [ A0 and also  D A \ A0 . That is, the intersection of A
and its complement is the null set. For example, in a tossing of a coin, turning a head and
turning tail are complementary events.

4.4 MUTUALLY EXCLUSIVE EVENTS

Two events are said to be mutually exclusive if they cannot occur at the same time. A good
example of a mutually exclusive event is that in flipping a coin, either a head can appear or a
tail, but not the two can appear simultaneously. Let events A and B be mutually exclusive, then
A \ B D .

4.5 MUTUALLY INCLUSIVE EVENTS

Two events are said to be mutually inclusive, if there are some common outcomes in the two
events. Getting a head and a tail in flipping of two coins is an example of mutually inclusive
events. Another example is getting an odd number or prime number in a throwing of two dice.

4.6 VENN DIAGRAM

Venn diagrams are very useful in visualizing relations between sets. Figures 4.1–4.4 have a dia-
grammatical representation of sets in a rectangular box.
 4.7. PROBABILITY 65


Figure 4.1: Venn diagram, A  S.

Figure 4.2: Two sets A and B , where B  A.

Figure 4.3: Two sets A and B , where A \ B .

4.7 PROBABILITY
The probability of event A could be defined as the number of ways event A can occur divided by
the total number of possible outcomes. It is mathematically defined as:

number of outcomes favorable to A

Probability (A) D :
number of possible outcomes
66 4. BASIC PROBABILITY CONCEPTS


Figure 4.4: Two sets A and B , where A [ B is a shaded area.

n.A/
P .A/ D : (4.1)
n.S/

Example 4.1
A fair die is rolled once, what is the probability of: (i) Rolling an even number? (ii) Rolling an
odd number?

Solution:
number of even number
Probability (even number) D :
number of possible outcomes
Let A be set of even number, then A D f2; 4; 6g and B be a set of odd numbers in a tossing of
a die, the B D f1; 3; 5g and the sample space, S D f1; 2; 3; 4; 5; 6g.
(i) P .A/D 36 .
(ii) P .B/D 36 .

Example 4.2
What is the probability that an applicant resume will be treated within a week of submitting
application if 5,000 graduates applied for a job and the recruitment firm can only treat 1,000 re-
sumes in a week?

Solution:
1;000
Probability (treating a resume) D 5;000
D 0:2.

4.7.1 SIMULATION OF A RANDOM SAMPLE IN R

The R command sample is used to simulate drawing a sample. The function sample contains
three arguments:
 4.7. PROBABILITY 67
x is a vector containing the objects
size the number of samples to be drawn
replace is set TRUE or FALSE depending on whether you want to draw
the sample with replacement or not.
Note: Sampling with replacement (replace D TRUE) indicates replacing the object before per-
forming another sample selection. On the other hand, sampling without replacement (replace
D FALSE) means selection in succession.

Example 4.3
Consider an urn containing one blue ball, one green ball, and one yellow ball. To simulate draw-
ing two balls from the urn with replacement and without replacement, we have:

# Urn contains one blue, one green and one yellow balls
Urn = c('blue', 'green', 'yellow')
# Sampling with replacement
sample(x = Urn, size =2, replace = TRUE)
[1] "yellow" "yellow"

# Sampling without replacement

Urn = c('blue', 'green', 'yellow')
sample(x = Urn, size =2, replace = FALSE)
[1] "yellow" "blue"

For the two cases, there is no unique answer here as long as they satisfied the condition where
they are selected. In the sampling with replacement, our result can be any from the sample space
of:
["yellow" "yellow"], ["yellow" "blue"], ["blue" "yellow"], ["blue" "blue"],
[ "blue" "green"], [ "green" "blue"], [ "green" "yellow"], [ "yellow" "green"],
[ "green" "green"]

However, in the case of sampling without replacement, we can have any of these outcomes:
["yellow" "blue"], ["blue" "yellow"], [ "blue" "green"], [ "green" "blue"],
[ "green" "yellow"], [ "yellow" "green"],

So by putting the yellow ball back in after it was selected the first time, it expands the solution
possibilities significantly. In our case it meant picking the yellow ball twice.

Example 4.4
Draw four random numbers between 1 and 20 with replacement and without replacement.
68 4. BASIC PROBABILITY CONCEPTS
# Sampling with replacement
sample(x = 1:20, size =4, replace = TRUE)
[1] 17 7 7 10

# Sampling without replacement

sample(x = 1:20, size =4, replace = FALSE)
[1] 1 17 18 20

Note: There is no unique answer for both cases in as much the condition of generating the
samples is satisfied. It is generated randomly.
Example 4.5
Simulate rolling a fair die.
# Sample from rolling a fair die
set.seed(25)
sample(x = 1:6, size =1, replace = TRUE)
[1] 3

Note: Replacement option (TRUE or FALSE) does not hinder the outcome here since the
experiment is performed once, i.e., size D 1. A seed is set in the codes in Example 4.5 above;
this is done to make the codes reproducible and to ensure the same random numbers will be
generated each time the script is executed.
Example 4.6
Replicate the experiment in Example 4.5 in 100 times.
# Sample replication
set.seed(25)
replicate(100, sample(x = 1:6, size =1, replace = TRUE) )
[1] 3 5 1 6 1 6 4 3 1 2 2 3 6 4 5 1 4 5 3 5 1 3 1 1 2 2 1 4
4 2 5 3 2 1 5 3 6 3 1 1 5 2 4 5 4 2 6 5 1 5 1 5 5 3 4
[56] 1 1 4 2 1 1 2 2 5 4 6 3 6 4 2 6 1 1 5 2 3 1 6 2 2 4 3
4 1 5 5 2 3 5 4 3 2 4 4 1 3 4 6 2 2

4.8 AXIOMS OF PROBABILITY

The expressions below are the axioms of probability.
• P .A/  0; for all A (nonnegativity)
The probability of A is always greater than or equal to zero (nonegative).
 4.9. BASIC PROPERTIES OF PROBABILITY 69
• P .S/ D 1 (normalization)
The probability of the sample space is 1, meaning that probabilities of a sample will be
from 0 and 1.
• If event A and event B are disjoint, i.e., .A \ B/ D ;, then
P .A [ B/ D P .A/ C P .B/ (additivity)
If event A is disjoint from event B , then the probability of both occurring is the same as
adding the probability that event A will occur and the probability that event B will occur.


• For an infinite number of points, and A1 ; A2 ; : : : are disjoints, then P [1 i D1 Ai D
P1
i D1 P .Ai /.

4.9 BASIC PROPERTIES OF PROBABILITY

The followings are the basic properties of probability and derived from the axioms of probability.
1. P .A/ D 1 P .A0 /.

2. P .A/  1.
3. P .A/ D 0.
4. P .A/  P .B/ if only and if A  B ; note that they can be equal.
5. P .A [ B/ D P .A/ C P .B/ P .A \ B/.

P
6. P .A [ B/  P .A/ C P .B/ or, alternatively, P [niD1 Ai D niD1 P .Ai /.

4.10 INDEPENDENT EVENTS AND DEPENDENT EVENTS

Two events A and B said to be independent if occurrence of event A happening does not af-
fect the occurrence of event B to happen. In other words, two events A and B are statistically
independent if:
P .A and B/ D P .A/  P .B/ :
Also, independent events can be of the form:

P .AjB/ D P .A/ and P .BjA/ D P .B/ :

That is, that the probability of A given B is the same as the probability of A happening, because
A and B are independent of one another. Similarly, the probability of B given A is the same as
the probability of B happening, because they are independent events.
For any finite subset of events Ai1 ; Ai2 ; Ai 3 ; : : : ; Ai n are said to be independent, if:

P .Ai1 ; Ai2 ; Ai3 ; : : : ; Ain / D P .Ai1 / P .Ai 2 / P .Ai3 / : : : P .Ain / :

70 4. BASIC PROBABILITY CONCEPTS
A typical example of independent events is flipping a coin twice, the outcome of head (H)
or tail (T) facing up in the first toss does not affect the outcome in the second toss.
However, two events A and B said to be dependent events if occurrence of event A affects
the occurrence of event B or vice versa. The probability that B will occur given that A has
occurred is referred as to conditional probability of B given A, and it can be written as P .BjA/.

Example 4.7
A fruit basket contained 30 pieces of fruit: 8 oranges, 12 apples, and 10 bananas. If two fruits
are taken at random after one another:

(a) What is the probability that the first fruit is a banana and the second fruit is an orange if
the first fruit is returned in the basket before the second fruit is taken?

(b) What is the probability that the first is an orange and the second is an apple if the fruit is
taken without replacement?

Solution:
Let A represents number of apples.
Let B represents number of bananas.
Let O represents number of oranges.
10 8
(a) P .B and O/ D P .B/:P .O/ D 30
 30
D 0:088.
8 12
(b) P .O and A/ D P .O/:P .AjO/ D 30
 29
D 0:11.

Example 4.8
Table 4.1 below shows the outcome of a survey conducted in Enugu State, Nigeria to look at the
rate small businesses fail despite the programs of government directed at their survival. Calculate
the probability that a restaurant firm is highly prone to chance of occurrence?

Table 4.1: Distribution of firms on their proneness to chance of occurrence

Proneness Apparel/Fashion Restaurant Technical Total

High prone 216 14 20 250
Prone 30 10 5 45
Not prone 10 4 6 20
Total 256 28 31 315
Source: Orga and Ogbo (2012)
 4.11. MULTIPLICATION RULE OF PROBABILITY 71
Solution:
number of high prone restaurants
P .restaurantjHigh prone/ D ;
total number of high prone firms
14
P .restaurantjHigh prone/ D D 0:056:
250

4.11 MULTIPLICATION RULE OF PROBABILITY

The multiplication rule states that: the probability of the occurrence of two events is the probabil-
ity of the intersection of two events and is same as the product (multiplication) of the probability
of the occurrence of one event and the probability of the occurrence of the second event:

P .A and B/ D P .A \ B/ D P .A/  P .B/: (4.2)

If and only if event A and event B are independent:

P .A and B/ D P .A \ B/ D P .A/  P .BjA/:

Thus, P .BjA/ indicates the probability of occurrence of event B given event A has occurred.
For an event A and event B are independent.
In general,

P .A1 \ A2 \ : : : \ An / D P .A1 / P .A2 jA1 / P .A3 jA1 \ A2 / : : :

(4.3)
: : : P .An jA1 \ A2 : : : \ An 1/ :

4.12 CONDITIONAL PROBABILITIES

Let A and B be events in a sample sample, S, where probability of A is not equal to zero, i.e.,
P .A/ ¤ 0, then the condition probability that event B will occur given that event A has occurred
is:
P .A \ B/
P .BjA/ D : (4.4)
P .A/
This is equivalent to:
P .A \ B/ D P .BjA/  P .A/: (4.5)
Conditional probability permits us to calculate probabilities of events based on partial
knowledge of the outcome of a random experiment.
Example 4.9
Consider that 10 non-defective bulbs and 2 defective bulbs are contained in a pack. To find
the defective bulbs the store manager were randomly selecting the bulbs one-by-one without
replacement. What is the probability that he is lucky to find defective bulbs in the first two
tests?
72 4. BASIC PROBABILITY CONCEPTS
Solution:
Let D1 be the event that the first selection is defective.
Let D2 be the event that the second selection is defective:
2 1
P .D1 \ D2 / D P .D1 / P .D2 jD1 / D  D 0:02:
10 9

4.13 COMPUTATION OF PROBABILITY IN R

Example 4.10
Consider rolling two fair dice; calculate the probability of getting a sum of their outcomes when
the experiment is performed in 100 times. Hence, plot the obtained results.
Step 1. Create a function to roll two fair dice and return their sum as follows.
dice.2 <- function () {
dice <- sample (1:6, size = 2, replace =TRUE)
return (sum(dice))
}

Step 2. Replicate the same experiment in 100 times.

# replicate the experiment in 100 times
set.seed (1010)
dice.roll<-replicate(100, dice.2())
dice.roll
[1] 6 7 11 10 8 11 10 2 8 3 8 4 3 4 9 5 8 5 6 4
7 9 5 9 3
[26] 11 4 3 8 2 6 12 9 9 6 10 4 4 6 3 8 9 6 8
5 6 9 8 6 11
[51] 8 4 2 8 7 7 8 10 5 12 9 11 11 5 8 7 6 3 5
6 7 7 5 12 8
[76] 6 9 7 8 12 12 4 12 5 10 7 3 4 6 2 7 9 6 9
3 6 2 7 11 6

Step 3. Tabulate the outcomes of the experiment.

# tabulate the outcomes
table (dice.roll)
dice.roll
2 3 4 5 6 7 8 9 10 11 12
5 8 9 9 15 11 14 11 5 7 6
 4.13. COMPUTATION OF PROBABILITY IN R 73
Step 4. Calculate the probabilities.
# compute the probabilities of the sum of two dice
prob = table(dice.roll)/length(dice.roll)
prob
dice.roll
2 3 4 5 6 7 8 9 10 11 12
0.05 0.08 0.09 0.09 0.15 0.11 0.14 0.11 0.05 0.07 0.06
The probabilities values produced in R is similar to presenting it as:
Sum 2 3 4 5 6 7 8 9 10 11 12
Prob. 0.05 0.08 0.09 0.09 0.15 0.11 0.14 0.11 0.05 0.07 0.06
Step 5. Plot of relative frequency.
plot (table(dice.roll)/length(dice.roll), xlab = 'Sum of outcomes',
ylab ='Relative Frequency', main = 'Graph of 100 Rolls of
two fair Dice')

Graph of 100 Rolls of two fair Dice

0.15
Relative Frequency

0.10
0.05
0.00

2 3 4 5 6 7 8 9 10 11 12

Sum of outcomes

Figure 4.5: Graph of 100 rolls of 2 fair dice.

74 4. BASIC PROBABILITY CONCEPTS
Example 4.11
Consider a bottling company producing three brands of products namely—A, B, and C. The
probability of producing A, B, and C are 0.54, 0.36, and 0.10. The probability of a defective
product on the assembly line is 0.02 and non-defective product is 0.98. If a random sample of
1,000 brands of the products are produced, what is the probability that the defective product is
B?

Solution:
# set seed number to get the same result always since the
# observation is randomly generated
set.seed (1001)
Brands <- sample(c("A","B", "C"), 1000,
prob=c(0.54, 0.36, 0.10), rep=TRUE)
Status <- sample(c("defective","non-defective"), 1000,
prob=c(0.02,0.98), rep=TRUE)
dataset <- data.frame(Status, Brands)
tabular <-with(dataset, table(Status, Brands))
tabular
Brands
Status A B C
defective 11 7 0
non-defective 536 357 89

# compute probability of B given that is defective

probB_D <- tabular[1, 2]/sum(tabular[1, ] )
probB_D
[1] 0.3888889

4.14 EXERCISES
4.1. A fair die is tossed once. Calculate the probability that: (i) exactly 3 comes up, (ii) 3 or
5 come up, and (iii) a prime number comes up.
4.2. Two dice are rolled together once. What is the probability that the sum of the outcome
is: (i) 4, (ii) less than 4, (iii) more than 4, and (iv) between 7 and 12 inclusive.
4.3. A card is drawn from a well shuffled pack of 52 cards. Find the probability of:
(a) a king or a queen,
(b) a black card,
(c) a heart or a red king, and
 4.14. EXERCISES 75
(d) a spade or a jack.
4.4. A research firm has 30 staff, which consists of 2 research managers, 5 research associates,
3 administrative staff, and 20 fieldworkers. The managing director of the firm wants to
set up a committee of five-man. Find the probability that the committee will consist of:
(a) 1 research manager, 1 research associate and 3 fieldworkers;
(b) 1 research manager, 2 research associates and 2 fieldworkers; or
(c) 2 research managers, 3 research associates.
4.5. A firm has 100 employees (65 males and 35 females) and they were asked if they should
adopt the paternal leave or not, their responses are summarized in Table 4.2.

Table 4.2: Employees

Gender Favor Oppose Total

Male 62 3 65
Female 20 15 35
Total 82 18 100

(a) Find the probability that a male opposed to paternal leave policy.
(b) Find the probability that the person is a female, given that the person is in favor of
paternal leave policy.
 77

CHAPTER 5

Discrete Probability
Distributions
Consider a random variable (X ) that takes integers values, X1 ; X2 ; : : : ; Xn with the correspond-
P
ing probabilities of P .X1 /; P .X1 /; : : : ; P .Xn / and the probabilities P .X / such that n1 P .X/ D
1 is called a discrete probability distribution. The type of the random variable determines the
nature of the probability distribution it follows. A discrete random variable usually involves
counting which takes an integer value while the continuous random variable involves measur-
ing and it takes both integer and a fractional part or real number. When the probabilities are
assigned to random variables, then the collection of such probabilities give rise to a probability
distribution. The probability distribution function can be abbreviated as pdf. A discrete proba-
bility distribution satisfies two conditions:

0  P .X /  1 (5.1)

and X
P .X/ D 1: (5.2)

5.1 PROBABILITY MASS DISTRIBUTION

The probability that a discrete random variable X takes on a particular value x , that is,
P .X D x/ D f .x/ is called a probability mass function (pmf ) if it satisfies the following:

• P .X D x/ D f .x/ > 0, if x 2 S .
All probability must be positive for every element x in the sample space S . Hence, if
element x is not in the sample space S , then f .x/ D 0.
P
• x2S f .x/ D 1.
The sum of probabilities for all of the possible x values in the sample space S must equal 1.
P
• P .X 2 x/ D x2A f .x/.
The sum of probabilities of the x values in A is the probability of event A.
78 5. DISCRETE PROBABILITY DISTRIBUTIONS
Example 5.1
Experiment: toss a fair coin two times.
Sample space: S D fHH; HT; TH; TTg.
Random variable X is the number of tosses showing heads.
Thus, X W S ! R

X D .HH/ D 2
X D .HT/ D .TH/ D 1
X D .TT/ D 0
X D f0; 1; 2g:

That is, random variable X takes a range of values 0, 1, and 2. Hence, the pmf is given by:
1 1 1
P .X D 0/ D ; P .X D 1/ D ; and P .X D 2/ D :
4 2 4

5.2 EXPECTED VALUE AND VARIANCE OF A DISCRETE

RANDOM VARIABLE
The expected value is a measure of central tendency in a probability distribution and the variance
is a measure of dispersion of a probability distribution.
The mean or expected value () for a discrete pdf is computed as:
n
X
D Xi P .Xi / : (5.3)
iD1

The variance ( 2 ) of a discrete pdf is computed as:

n
X
2 D .Xi /2 P .Xi / : (5.4)
i D1

Also, the standard deviation ( ) of a discrete pdf is given as:

v
u n
uX
 Dt .Xi /2 P .Xi /: (5.5)
i D1

Examples of a discrete probability distribution are: rolling a die, flipping of coins, counting
of car accidents on highways, producing a defective and non-defective goods, etc.
 5.2. EXPECTED VALUE AND VARIANCE OF A DISCRETE RANDOM VARIABLE 79
The following are the common discrete probability distributions used in statistics:
Bernoulli distribution, binomial distribution, geometric distribution, hypergeometric distribu-
tion, Poisson distribution, negative binomial distribution, and multinomial distribution.

Example 5.2
A single tossing of a fair die.
Roll (X ) 1 2 3 4 5 6
P .X / 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
This has a discrete probability distribution since the random variable (X ) takes the integers
(i.e., 1, 2, 3, 4, 5, and 6) with corresponding probabilities of 0.1667 each. This satisfies that
P
0  P .X/  1 and P .X/ D 1.

Example 5.3
Consider a distribution of the family size.
Assuming an analyst obtained the result in the table below in a survey of 1,000 households
in Nigeria. Let random variable X be the number of households size with probability of outcome.
Household size (X ) 1 2 3 4 5 6C
P .X / 2/69 4/77
3/28 29/81 1/3 3/25
P
This satisfies that 0  P .X /  1 and P .X / D 1.

Example 5.4
The number of defectives items per month and the corresponding probabilities in a manufac-
turing firm are given in the table below.
Defective(X ) 0 1 2 3 4 5
P .X / 1/15 1/6 3/10 1/5 2/15 2/15
P
Let X be number of defective items in a month, then 0  P .X/  1 and P .X / D 1
are satisfied.

Example 5.5
Using the data Example 5.4 above, calculate:

1. expected value of the distribution,

2. standard deviation of distribution, and

3. variance of the distribution.

80 5. DISCRETE PROBABILITY DISTRIBUTIONS
Solution:
Accident (X ) 0 1 2 3 4 5
p.X / 1/15 1/6 3/10 1/5 2/15 2/15
Xp.X / 0 1/6 3/5 3/5 8/15 2/3
X  -2.57 -1.57 -0.57 0.43 1.43 2.43
.X /2 6.59 2.45 0.32 0.19 2.05 5.92
p.X/ .X /2 0.44 0.41 0.10 0.04 0.27 0.79
1.
5
X
D Xi P .Xi / :
i D1

Expected value D 0 C 1=6 C 3=5 C 3=5 C 8=15 C 2=3 D 2:57.

2. Standard deviation
v
u 6
uX
 Dt .Xi /2 P .Xi /
i D1
p
D 0:44 C 0:41 C 0:10 C 0:04 C 0:27 C 0:79 D 1:43:
Pn
3. Variance . 2 / D iD1 .Xi /2 P .Xi /.

 2 D 1:432 D 2:05:

The code below gives how we can use R to get solution to the worked Example 5.5 above.

Using R Code
# to calculate the mean, standard deviation, and variance of the distribution in Example 5.5
set.seed(100)
x <- c(0, 1, 2, 3, 4, 5)
prob <- c(0.067, 0.167, 0.30, 0.20, 0.133, 0.133)
weighted.mean (x, prob)
[1] 2.56
x_miu = x - weighted.mean (x, prob)

# calculate the variance

x_miu2 = x_miu * x_miu
d = prob* x_miu2
var = sum(d)
var
 5.3. BINOMIAL PROBABILITY DISTRIBUTION 81
[1] 2.045904

# calculate the standard deviation

sd = sqrt(var)
[1] 1.430351

5.3 BINOMIAL PROBABILITY DISTRIBUTION

In this section, we are looking at the one of the commonest discrete probability distributions,
its derivations, formula, statistical properties and real life applications using examples. A couple
of useful formulas shall be presented in this section to give more understanding about the bi-
nomial distribution. The derivation of the mean and variance of a binomial distribution shall be
discussed.
Binomial distribution is a discrete probability distribution and it is emanated from the
Bernoulli trial, which based on a trial which there are two possible outcomes, say “success” and
“failure.” The Bernoulli trial can be presented mathematically as:

1 if the outcome is a sucess
XD (5.6)
0 if the outcome is a failure.
Let p represent the probability of success, then 1 p will be probability of failure. The
pmf of X can be defined as:
f .x/ D p x .1 p/1 x
; xD0 or 1: (5.7)
However, the binomial probability function is the extension of the Bernoulli distribution
when the following properties are met.
• Bernoulli trials are performed in n times.
• Two outcomes, success .p/ and failure .q/ are possible in each of the trial.
• The trials are independent.
• The probability of success .p/ remains the same between the trials.
Let X be the number of successes in the n independent trials, then the pmf of X is given
as: !
n x
f .x/ D p .1 p/n x ; x D 0; 1; : : : ; n: (5.8)
x
n


x
D .n nŠ
x/ŠxŠ
reads thus: n combination x and nŠ reads thus: n factorial (e.g., 5Š D 5  4  3 
2  1). For example, !
5 5Š
D D 10;
2 .5 3/Š2Š
82 5. DISCRETE PROBABILITY DISTRIBUTIONS
where n is the number of trials experiment is performed, x is the number of success recorded,
and p is the probability of success.

Example 5.6
A farmer supplies eggs in crates to his customers in the neighboring city by motorcycle. An
egg gets broken in a crate, due to a bad road system, on his way to the city with the probability
of 0.75 if he transports 10 crates of eggs. Assuming that a number of damaged eggs is binomially
distributed, what is the probability that three eggs will break before he reaches his destination?

Solution:
Given that n D 10, x D 3, and p D 0:75. Then, the pmf of a binomial is given by:
!
n x
f .x/ D p .1 p/n x ; x D 0; 1; : : : ; n
x
!
10
f .x D 3/ D .0:75/3 .1 0:75/10 3
3
10Š
f .x D 3/ D .0:75/3 .0:25/7 D 0:0031:
.10 3/Š3Š

5.4 EXPECTED VALUE AND VARIANCE OF A BINOMIAL

DISTRIBUTION
In this section, we shall derive the expected value or mean of binomial distribution and also
show how the variance of binomial distribution is being derived.

5.4.1 EXPECTED VALUE (MEAN) OF A BINOMIAL DISTRIBUTION

From the binomial pmf, we have:
!
n x
f .x/ D p .1 p/n x
; x D 0; 1; : : : ; n:
x

Since the expected value of a discrete distribution is given as:

n
X
E .x/ D x
f .x/:
xD0

Substituting for f .x/, we have:

n
!
X n x
E .x/ D x
p .1 p/n x
:
xD0
x
 5.4. EXPECTED VALUE AND VARIANCE OF A BINOMIAL DISTRIBUTION 83
Let q D 1 p.
n
!
X n x n x
E .x/ D x
p q :
xD0
x

Now, substituting for the binomial coefficient yields

n
X nŠ
E .x/ D x
px qn x

xD0
xŠ .n x/Š
Xn
n.n 1/Š
E .x/ D x
px 1
pq n x
:
xD0
x.x 1/Š.n x/Š

So,
n
X n.n 1/Š
E .x/ D x
px 1
pq n x
:
xD0
x.x 1/Š .n x/Š

Pulling out np yields

n
X .n 1/Š
D np px 1 n x
q
xD1
.x 1/Š .n x/Š

and
n 1
!
X n 1 x 1 .n 1/ .x 1/
D np p q :
x 1D0
x 1

Since sum of probabilities for all of the possible x values in the sample space S is equal 1,
then the result yields
E.x/ D np: (5.9)
Therefore, the expected value (mean) of the binomial probability function is np where n is the
number of trials in the experiment and p is the probability of success.

5.4.2 VARIANCE OF A BINOMIAL DISTRIBUTION

The variance of a distribution can be compute from the expected value of that distribution:


var.x/ D E x 2 .E.x//2 :

Let x 2 D x.x 1/ C x . Then, substitute for the value of x 2

var.x/ D E.x.x 1/ C x/ .E.x//2 :

84 5. DISCRETE PROBABILITY DISTRIBUTIONS
Let’s solve E.x.x 1// the same approach we use to derive the mean (E.x/) of a binomial
distribution. Therefore,
n
X nŠ
E.x.x 1// D x.x 1/
p x .1 p/n x
:
xD0
xŠ.n x/Š

So,
n
X n.n 1/.n 2/Š
E.x.x 1// D x.x 1/
p x .1 p/n x
:
xD0
x.x 1/.x 2/Š.n x/Š

Let p x D p x 2C2 D p 2 p x 2
.
Substituting for p x gives
n
X n .n 1/ .n 2/Š
E .x .x 1// D x .x 1/
p2px 2
.1 p/n x
:
xD0
x .x 1/ .x 2/Š .n x/Š

And letting q D 1 p gives

n
X n .n 1/ .n 2/Š
E .x .x 1// D x .x 1/
p2px 2 n x
q :
xD0
x .x 1/ .x 2/Š .n x/Š

Pulling out the constant terms and then evaluate yields

n
X .n 2/Š
E .x .x 1// D n .n 1/ p 2 px 2 n x
q :
xD0
.x 2/Š .n x/Š
Pn .n 2/Š x 2 n x
Since xD0 .x 2/Š.n x/Š p q D 1. Then, E .x .x 1// D n .n 1/ p 2 :

var .x/ D E .x .x 1// C E.x/ .E .x//2 :

Therefore, we can find the variance by substituting for E .x .x 1//:

var .x/ D n .n 1/ p 2 C np .np/2 ;

var .x/ D n2 p 2 np2 C np n2 p 2 ;
var .x/ D np np2 ;
var .x/ D np.1 p/;

or

var .x/ D npq; (5.10)

 5.4. EXPECTED VALUE AND VARIANCE OF A BINOMIAL DISTRIBUTION 85
where n is the number of trials, p is the probability of success, and q is the probability of failure.
Here, we have shown that the variance of a binomial distribution is the product of number
of trials .n/, the probability of success .p/, and the probability of failure .q/.

Example 5.7
A manufacturing company has 100 employees and an employee has a chance of 5% of being
absent from work at a particular day. It is assumed that absent of a worker from work would
not affect another. The company can continue production at a particular day if no more than
20 worker absent for that day. Calculate the probability that out of the 10 workers randomly
selected, 3 workers will be absent from work. Hence, find the expected number of workers that
will be absent from work at that particular day.

Solution:
1. Number of trials, n D 10 workers.
Possible outcomes: success .p/ is the probability that a worker is absent from work and
failure .q/ is the probability that a worker is not absent from work.
Probability of success: P (worker is absent from work) D 0.05 and it is constant all through
the trials.
The events are independent, i.e., the presence of a worker in the company does not affect
another.
The pmf of binomial is:
!
n x
f .x/ D p .1 p/n x ; x D 0; 1; : : : ; n
x
!
10
f .x D 3/ D .0:05/3 .1 0:05/10 3
3
!
10
f .x D 3/ D .0:05/3 .0:95/7 D 0:0105:
3

2. The expected number of worker to be absent in that day is: np D 10  0:05 D 0:5. This
interpreted that only one worker is expected to absent on that particular day.

Example 5.8
An airline operator has 12 airplanes. On a rainy day, the probability that an airplane will fly is
0.65. What is the probability that:
1. an airplane will fly;
86 5. DISCRETE PROBABILITY DISTRIBUTIONS
2. three airplanes will fly;
3. at most, two airplanes will fly; and
4. at least, two airplanes will fly.
Hence, calculate the number of airplanes that are expected to fly.

Solution:
1. Let x be number of airplane to fly.
N D 10 and p D 0:65
 
10
f .x D 1/ D .0:65/1 .0:35/9 D 0:0005:
1

2.  
10
f .x D 3/ D .0:65/3 .0:35/7 D 0:0212:
3

3.

f .x  2/ D .f .x D 0/ C f .x D 1/ C f .x D 2//
     
10 10 10 9 10
f .x  2/ D 0
.0:65/ .0:35/ C 1
.0:65/ .0:35/ C .0:65/2 .0:35/8
0 1 2
D 0:000028 C 0:0005123017 C 0:004281378 D 0:0048:

4.

f .x  2/ D 1 .f .x D 0/ C f .x D 1/ C f .x D 2//
D1 0:0048 D 0:9952:

5. The expected number airplanes to fly on a rainy day is np D 10  0:65 D 6:5. This indicates
that only seven airplanes will fly on a rainy day.

Example 5.9
A market representative makes a sale on a particular product per day with a probability of 0.25.
If he has 30 products to sell that day, find the probability that:
1. no sales are made;
2. five sales are made; and
 5.5. SOLVE PROBLEMS INVOLVING BINOMIAL DISTRIBUTION USING R 87
3. more than four sales are made.

Solution:

1. Let X be number of sales made on the product

p.x D making sales/ D 0:25

n D 30
!
30
p .x D 0/ D .0:25/0 .0:75/30 D 0:00018:
0

2.
!
30
p .x D 5/ D .0:25/5 .0:75/25 D 0:1047:
5

3.

p .x > 4/ D 1Œp .x D 0/ C p .x D 1/ C p .x D 2/ C p .x D 3/ C p .x D 4/

( ! !
30 30
D1 .0:25/0 .0:75/30 C .0:25/1 .0:75/29
0 1
! !
30 30
C .0:25/2 .0:75/28 C .0:25/3 .0:75/27
2 3
! )
30 4 26
C .0:25/ .0:75/
4
D1 f0:00018 C 0:01786 C 0:008631 C 0:026853 C 0:06042g D 0:90213:

5.5 SOLVE PROBLEMS INVOLVING BINOMIAL

DISTRIBUTION USING R
In this section, we shall demonstrate how we can use R to generate a random sample from
a binomial distribution through simulation and how to calculate probability from a binomial
distribution parameters. We shall use most of the examples illustrated in this chapter to compare
the results with R outputs.
88 5. DISCRETE PROBABILITY DISTRIBUTIONS
5.5.1 GENERATING A RANDOM SAMPLE FROM A BINOMIAL
DISTRIBUTION
We shall use the R function to simulate a binomial random variables with given parameters.
The function rbinom() is used to generate n independent binomial random variables. The
general form is rbinom (n, size, prob).
n number of random sample
size number of trials
prob probability of success.

# To simulate 100 binomial random number with parameters

n = 10 and p = 0.75
set.seed(10)
rv=rbinom (100, 10, 0.75)
rv
[1] 10 8 8 6 9 8 8 8 8 6 7 5 6 8 7 5 7 9 9 8 8
9 10 9 8
[26] 7 8 7 8 7 5 6 7 9 6 7 7 9 10 8 7 7 6 8 6 8
8 10 4 8
[51] 6 6 6 7 9 8 5 9 8 6 6 5 10 8 7 8 6 8 8 6 7
4 6 8 10
[76] 6 7 7 9 4 9 8 8 7 9 10 6 10 8 9 8 9 5 8 6 8
7 6 7 9

# the mean of the random sample

mean(rv)
[1] 7.44

# the standard deviation and variance of the random sample

sd(rv)
[1] 1.465564

var(rv)
[1] 2.147879

# To plot the bar chart of the random samples generated

set.seed(150)
rv=rbinom (100, 10, 0.75)
barplot(table(rv))
 5.5. SOLVE PROBLEMS INVOLVING BINOMIAL DISTRIBUTION USING R 89
barplot(table(rv), ylab="Frequency", xlab="Random Number",
main="X~Binomial (10, 0.75)")

X~Binomial (10, 0.75)

30
25
20
Frequency

15
10
5
0

4 5 6 7 8 9 10

Random Number

Figure 5.1: X -binomial.

5.5.2 CALCULATE PROBABILITY FROM A BINOMIAL DISTRIBUTION

In this section, dbinom—is the function to calculate binomial pdf
(x, size, prob, log = False).
x number of successes
size is the number of independent trials
prob is the probability of success.

Example 5.10
Using the data from Example 5.7 above, use R code to obtain the binomial probability distri-
bution and compare the result.
90 5. DISCRETE PROBABILITY DISTRIBUTIONS
dbinom(3,10,0.05)
[1] 0.01047506

This is exactly what we got in Example 5.7.

Example 5.11
Use the data on airplane Example 5.8 above, and then use R to obtain the probabilities in Ex-
ample 5.8 (1–4). Hence, compare the results:
# an airplane will fly
dbinom(1,10,0.65)
[1] 0.0005123017

# three (3) airplane will fly

dbinom(3,10,0.65)
[1] 0.02120302

# at most two airplanes will fly

no_air=dbinom(0,10,0.65)
air1=dbinom(1,10,0.65)
air2=dbinom(2,10,0.65)
prob_most2= no_air+ air1+ air2
prob_most2
[1] 0.004821265

# at least two airplanes will fly

prob_least2=1-prob_most2
prob_least2
[1] 0.9951787

# Expected number airplane to fly on a raining day

p<-0.65
n<-10
expected_value<-n*p
> expected_value
[1] 6.5

Good! We got the same result.

 5.5. SOLVE PROBLEMS INVOLVING BINOMIAL DISTRIBUTION USING R 91
Example 5.12
Use the data on product sales in Example 5.9 to solve the example in R and then compare the
results:

# no sale of the product

dbinom(0,30,0.25)
[1] 0.0001785821

# 5 sales of the products

dbinom(5,30,0.25)
[1] 0.1047285

# more than 4 sales of the products

p0<-dbinom(0,30,0.25)
p0
[1] 0.0001785821

p1<-dbinom(1,30,0.25)
p1
[1] 0.001785821

p2<-dbinom(2,30,0.25)
p2
[1] 0.008631468

p3<-dbinom(3,30,0.25)
p3
[1] 0.02685346

p4<-dbinom(4,30,0.25)
p4
[1] 0.06042027

prob<-1-(p0+p1+p2+p3+p4)
prob
[1] 0.9021304

The results in Example 5.9 (1–3) are the same with what we got using R.
92 5. DISCRETE PROBABILITY DISTRIBUTIONS
5.6 EXERCISES
5.1. (a) What is a discrete probability distribution and what is it used for?
(b) State the conditions to be satisfied for a discrete probability distribution.
5.2. (a) If X be a discrete random variable with the probability p.X /, what is the expected
value and standard deviation of X ?
(b) The table below shows the cases of malaria recorded in a health center with their
respective probabilities.

Malaria Cases (X ) 0 1 2 3 4 5 6 7 8
P .X/ 0.12 0.20 0.15 0.18 0.12 0.08 0.07 0.03 0.05

(a) Calculate the mean of the distribution of the malaria cases.

(b) Calculate the variance of the distribution.
(c) Calculate the standard deviation of the distribution.
5.3. (a) List the names of the discrete probability distribution you know.
(b) State the binomial probability distribution.
(c) Hence, derive the expected value and the variance of a binomial distribution.
5.4. The probability of having typhoid fever after drinking from a well water in a villager is
0.45. If ten villagers drank out of the well, what is the probability that:
(a) one villager will have typhoid fever;
(b) two villagers will have typhoid fever;
(c) at most two villagers will have typhoid fever;
(d) at least two villagers will have typhoid fever; and
(e) hence, calculate the number of villagers that expected to have typhoid fever.
5.5. Let X be a number of times a head appear in a tossing of a die. If a fair die is tossed
ten times, what is the probability that: (a) P .X D 0/, (b) P .X  1/, (c) P .X > 2/, or
(d) P .X  3/?
 93

CHAPTER 6

Continuous Probability
Distributions
In the previous chapter, we discused discrete probability distributions and their properties. In
the discrete probability distribution, the random variable takes only integer values or countably
infinite number of possible outcomes. However, in the countinous probability distribution, the
random variable takes any real value within a specified range. Typical examples of countinous
random variables are weight, temperature, height, and some economic indicators (prices, costs,
sales, inflation, investments, etc.). A continuous probability distribution demonstrates the com-
plete range of values a continuous random variable can take with their associated probabilities
along the range of values. This distribution is very useful in the prediction of the likelihood of
an event within a specified range of values. In this section we are discussing continous prob-
P
ability distributions and we will observe that the sum symbol, , which is used in derivation
of mean and variance of discrete probability distribution has turned to an integral symbol. The
integral sign indicates sum of continous random variable over an interval of points. Examples of
continous probability distributions are normal distribution, exponential distribution, student-t
distribution, chi-sqaure distribution, etc.
Let X be a continuous random variable that can take any real value within a specified
range, then the probability over a random variable is called a continuous probability distribu-
tion. Consider x to be continous, the probability denisty function denoted by f .x/ such that
the probability of event a  x  b is represented mathematically as:
Z b
P .a  x  b/ D f .x/ dx;
a

where a and b are the limits or points interval.

This implies the probability of a continous variable x within a specified range, say, a and
b , is the same as taking the integral function of the random variable over its bounded range.
However, the probability is zero if the continous random variable x is not bounded an
interval.
94 6. CONTINUOUS PROBABILITY DISTRIBUTIONS
That is, Z a
P .x D a/ D f .x/ dx D 0;
a
since b D a.

Example 6.1  
1
The time required by a driver to drive his boss from home to office is a function of x2
. What
is the probability that the driver will get to the office between 5–10 min?

Solution:
Let x denotes the time required for the driver to move from home to office.
Given that:
1
f .t/ D 2 ; a D 5 and b D 10:
x
Then, Z 10  
1
P .5  x  10/ D 2
dx:
5 x
Integrate the Right-hand side (RHS) over the interval,
 
1 10
P .5  x  10/ D :
x 5
Substitute the value of x and then evaluate the RHS,
   
1 1
P .5  x  10/ D
10 5
P .5  x  10/ : D 0:10

The probability that the driver will reach office from home within 5–10 min is 0.10. This
means that he has a low chance of getting to office under this condition.

6.1 NORMAL DISTRIBUTION AND STANDARDIZED

NORMAL DISTRIBUTION
In this section, we are considering a normal distribution as most important distribution because
if you have many independent variables by non-normal distributions, the aggregation of these
variables will tend to a normal distribution as the number of observations is large (i.e., central
limit theorem). The central limit theorem is discussed in Chapter 7. Normal distribution char-
icterizations makes it the most widely used distribution in statistics and applied mathematics.
It is also useful in measuring moments, kurtosis, skweness, etc. A normal distribution is also
known as Gaussian distribution, named after after the mathematician Karl Friedrich Gauss.
 6.1. NORMAL DISTRIBUTION AND STANDARDIZED NORMAL DISTRIBUTION 95
A normal ditribution is a bell-shaped curve having a mean denoted by  and standrd
deviation denoted by  . The density curve of a normal distribution is symmentrical, centered
about its mean and the spread by its standard deviation. The altitude of a normal density curve
at a point y is given as:
 
2

1 1  y  2
f yI ;  D p exp ; 1  y  1;
 2 2 

where y is a continuous random variable,  is the mean, and  2 is the variance of the distribution.
The standard normal curve is a special case of the normal curve when the mean is 0 and
standard deviation is 1. A normal distribution can be written in the form X  N.;  2 / and
reads, thus X is normally distributed with mean, , and variance,  2 . The total area under the
curve is 100%, i.e., all observations fall under the curve. A dataset is said to be normally dis-
tributed, if 68% of the observations fall within ˙1SD of the mean. Also, about 95% of the
observations will fall within ˙2SD and 99.7% of the observations will fall within ˙3SD , as
shown in Fig. 6.1.

0.13% 2.14% 13.59% 34.13% 34.13% 13.59% 2.14% 0.13%

-3 SD -2 SD -1 SD 0 +1 SD +2 SD +3 SD

Figure 6.1: Normal curve.

Properties of a Normal Distribution

The followings are the properties of a normal distribution.
(a) The mean, median, and mode have the same value.

(b) The curve is symmetric.

(c) The total area under the curve is 1.

(d) The curve is denser in the center and less dense in the tails.

(e) Normal distribution has two parameters: mean ./ and variance . 2 /.
96 6. CONTINUOUS PROBABILITY DISTRIBUTIONS
(f ) About 68% of the area of a normal distribution is within one standard deviation of the
mean.

(g) About 95% of the area of a normal distribution is within two standard deviations of the
mean.

(h) About 99% of the area of a normal distribution is within three standard deviations of the
mean.

6.2 STANDARD NORMAL SCORE (Z -SCORE)

The standard normal score is the standardized value of a normally distributed random variable
and it is usually referred as z -score. Standard normal score is useful in the area of statistics because
it enables us to find the probability of a score occurring within a normal distribution and also
used to make a comparison between two scores that comes from different normal distributions. A
random variable is standardized by substracting the mean of the distribution from standardized
value, and then divides it by the standard deviation of the distribution. The z -score of a random
variable can be written mathematically as:

X 
zD  N.0; 1/:

Thus, z is normally distributed with a mean of 0 and standard deviation of 1.
In the above equation X is the random variable to being standardize,  is the mean of the
distribution and  is the standard deviation of the distribution. Thus, once the random variable
is standardized, it is approximately normal with mean of 0 and standard deviation of 1 which is
same as standard normal.

Example 6.2
A class of 30 students sat for examinations on mathematics, the class mean score is 65 and
the standard deviation is 12.5. Assuming that the scores is normally distributed, what is the
percentage of students scoring above 70 in the mathematics examination?

Solution:
Given that: n D 30,  D 65, and  D 12:5. P .x > 70/ D?
Standardize the students’ scores:
 
x 65 70 65
P .x > 70/ D P > :
12:5 12:5
x 65
Let z D 12:5
:
P .x > 70/ D P .z > 0:4/ :
 6.2. STANDARD NORMAL SCORE (Z -SCORE) 97
Since sum of probability is 1, thus the left side of normal curve can be written as
P .x > 70/ D 1 P .z  0:4/ :
Look at the probability corresponding to z  0:4 from the standard normal table.
That is, ˆ .0:4/ D 0:6554.
P .x > 70/ D 1 ˆ .0:4/ D 1 0:6554 D 0:3446:
Thus, 34% of the students that sat for the mathematics examination scored above 70.
Example 6.3
A plastic production machine produces plastics with the mean of 80 and standard deviation of 5
in a minute. Assuming that the production of plastics followed a normal distribution, calculate:
(a) the probability that the machine will produce between 78 and 85; and
(b) the probability that the machine will produce less than 90.

Solution:
Given that:  D 80 and  D 5.
First, standardize the production.
Let z D X 580 .
78 80 X 80 85 80


(a) P .78  x  85/ D P 5
 5
 5
.
X 80
Subtitute for z D 5
 
78 80 85 80
P .78  x  85/ D P z D P . 0:4  z  1/ :
5 5
Look for the probability corresponding to 0:4  z and z  1 from the standard normal
table and substract small value from high value:
D ˆ .1/ ˆ . 0:4/ D 0:8413 .1 0:6554/ D 0:4967:
This means that the probability that the machine will produce between 78–85 plastics with
a minute is 0.5.
(b) Standardize the production variable:

 
x 80 90 80
P .x < 90/ D P < D ˆ .2/ D 0:9772
5 5
 
90 80
P .x < 90/ D P z< D ˆ .2/ D 0:9772:
5
This indicates that the probability that the machine will produce less than 90 plastics is
0.98; this implies it is almost sure for the machine to produce 90 plastics within a minute.
98 6. CONTINUOUS PROBABILITY DISTRIBUTIONS
Example 6.4
Suppose that a Corporate Affairs Commission official knows that the monthly registration of
new companies followed a normal distributed withaverage of 50 new registrations per month
and variance of 16 new registrations. Find the probability that: (a) new registered companies
will less than 35, (b) new registered companies will be 45, and (c) new registered companies fall
between 35 and 45 new companies.

Solution:
Let x be the number of registered companies.
x is normally distributed with mean (50) and variance (16). That is, x  N.50; 16/.
Since the standard deviation is the square root of variance then,


(a) P .x < 35/ D P x 450 < 35 4 50

P .x < 35/ D P .z < 3:75/ D 1 ˆ .3:75/ D 0:99991:

This means that it is very that less than 35 companies will register within a month period.


(b) P .x D 45/ D P x 450 D 45 4 50 D 1 ˆ .1:25/ D 1 .0:89435/ D 0:1056.
The result shows that there is low possibility of getting exactly 45 companies to register
within a month.


(c) P .35  x  45/ D P 35 4 50  z  45 4 50 D ˆ . 1:25/ ˆ . 3:75/ D 0:10556.
This means there is low chance of getting between 35 and 45 companies to register within
a month period.

6.3 APPROXIMATE NORMAL DISTRIBUTION TO THE

BINOMIAL DISTRIBUTION
In this section, we learn how to approximate normal distribution to the binomial distribution.
The central limit theorem is the tool that allows us to do so. Suppose that we have about 500 pub-
lic servants that are ready to take a promotion test in order to move to the next level of cadre.
The chairman of the civil service commission claimed that 75% of the staff sitting for the test
will pass at first sitting. If we want to find the probability that more than 300 public servants
will pass the promotion test at the first sitting. Then, this is a binomial event with n D 500,
p D 0:75, and P .x > 350/ D.
It will be tedious calculating the required

P .x D 350/ C P .x D 351/ C P .x D 352/ C : : : C P .x D 500/

     
500 150 500 149 500
350
.0:75/ .0:75/ C 351
.0:75/ .0:75/ C ::: C .0:75/500 .0:75/0 :
350 351 500
 6.4. USE OF THE NORMAL DISTRIBUTION 99
Due to computational effort, we can instead use a normal approximation to this problem.
The general rule of thumb for the approximation is that the sample size .n/ is sufficiently large,
i.e., np  5 and n.1 p/  5.
Since np D 500  0:75 D 375  5, then we can apply normal approximation.
Let z D X   .
Recall that the mean of a binomal distribution is pq and the variace is npq .
We can re-define the z -score
X np
zD p :
npq
Thus,  2 D npq D 500  0:75  0:25 D 93:75 and  D 9:68.
Since x > 350 implies that 350 not inclusive, it is possible to write it an approximation
value as x  350:5:
 
x 375 350:5 375
P .x > 350/  P .x  350:5/ D P 
9:68 9:68
P .z  2:53/ D 1 ˆ . 2:53/ D 1 0:00570 D 0:9943:

This is interpreted that more than 350 staff sitting for the test will pass at first sitting.

6.4 USE OF THE NORMAL DISTRIBUTION IN BUSINESS

PROBLEM SOLVING USING R
To generate normal distribution function, we use the function pnorm. The dnorm, qnorm, and
rnorm are used fordensity, quantile function, and random generation for the normal distribution,
respectively:

pnorm(q, mean =0, sd = 1, lower.tail = TRUE, log.p = FALSE)

dnorm(x, mean = 0, sd = 1, log = FALSE)
qnorm(p, mean = 0, sd = 1, lower.tail = TRUE, log.p = FALSE)
rnorm(n, mean = 0, sd = 1)

where
q vector of quantiles
mean vector of means
sd vector of standard deviations
log.p logical; if TRUE, probabilities p are given as log.p/
lower.tail logical; if TRUE (default), probabilities are P .X  x/,
otherwise, P .X > x/.
Note that the lower tail of the normal curve is the direction of the tail to the left of a given point
(value) while the upper tail is direction to the right side of the curve at a given point. Thus, the
100 6. CONTINUOUS PROBABILITY DISTRIBUTIONS
left-tailed test is when the critical region is on the left side of the distribution of the test value.
The right-tailed test is when the critical region is on the right side of the distribution of the test
value.
The following scripts are the solutions to the Examples 6.2, 6.3, and 6.4 above, with the
use of R codes. It can be seen that we got the same results in R with less stress.
# Example 6.2
# To calculate the probability of students scoring above 70
prob70<-pnorm (70, mean=65, sd=12.5, lower.tail=FALSE)
prob70
[1] 0.3445783
percent70<- prob70*100
percent70
[1] 34.45783

# Example 6.3
# (a) To calculate the probability that the machine will produce
# between 78 and 85?
btw78_85 <- pnorm(85, 80, 5) - pnorm(78, 80, 5)
btw78_85
[1] 0.4967665

# (b) To calculate the probability that the machine will produce

# less than 90?
less90 <- pnorm(90, 80, 5)
less90
[1] 0.9772499
 6.5. EXERCISES 101
# Example 6.4

# (a) To calculate the probability of new registration

# will be less than 35
prob35<-pnorm (35, mean=50, sd=4, lower.tail=FALSE)
prob35
[1] 0.9999116

# (b) To calculate the probability of new registration will be 45

prob45<- pnorm (45, mean=50, sd=4, lower.tail=TRUE)
prob45
[1] 0.1056498

# (c) To calculate the probability of new registration will fall

# between 35 and 45 new companies
btw35_45 <- pnorm(45, 50, 4) - pnorm(35, 50, 4)
[1] 0.1055614

6.5 EXERCISES
6.1. Suppose X  Bin.100; 0:8/. Calculate. (i) P .x  70/, (ii) P .x > 75/, and
(iii) P .70  x  90/.
6.2. A machine packs rice nominally in a 50-kg bag. It is observed that there is a variation in
the actual weight that is normally distributed. Records show that the standard deviation
of the distribution is 0.10 kg and the probability that the bag is underweight is 0.05.
Find:

(a) the mean value of the distribution; and

(b) what value standard deviation is needed to ensure that the probability that a bag
is underweight is 0.005. Assuming the mean is constant, and improvement in the
machine relies on reduction in the value of standard deviation.

6.3. The Ozone cinemas revealed that movie customers spent an average of $2,000 on con-
cessions with a standard deviation of $200. If the spending on concessions follows a
normal distribution:

(a) find the percentage of customers that will spend less than $1,800 on concenssions;
(b) find the percentage of customers that will spend more than $1,800 on concenssions;
and
(c) find the percentage of customer that will more than $2,000 on concenssions.
102 6. CONTINUOUS PROBABILITY DISTRIBUTIONS
6.4. Suppose we know that the survival rate of a new business is normally distributed with
mean 60% and standard deviation 8%. (a) What is the probability that more than 25
out of 30 new businesses registered in the week will survive? (b) What is the probability
that between 20 and 25 new businesses will survive?
6.5. The average lifetime of a light bulb is 4500 h with a standard deviation of 500 h. Assume
that the average lifetime of light bulbs is normal. (a) Find the probability that the average
life time of a bulb is between 4200 and 4700. (b) Find the probability that the average
life time of a bulb exceed 4500.
6.6. Suppose the time spent (in min) in opening a new bank account is X with a probability
density function of: 
kx 2 for x > 0
f .x/ D
0 otherwise:
(a) Find the value of k .
(b) Find the probabilities that time to spend to open a new account will less than
10 min.
 103

CHAPTER 7

Other Continuous Probability

Distributions
This chapter will explain some commonly used distributions in the testing of as hypothesis.
We will focus on the formulas and their uses in statistics. These distributions are Student’s t -
distribution, chi-square distribution, and F-distribution. We will see the conditions for using
each of the distributions mentioned.

7.1 STUDENT-T DISTRIBUTION

The t -distribution was first discovered by W. S. Gosset and he published it under the pseudonym
Student, therefore it is often called Student’s t -distribution. The t -distribution is the ratio of
standard normal distribution and square root of chi-square divided by its df. The degrees of
freedom are the number of free choices left after removing a sample statistic.
Suppose Z is normally distributed with mean 0 and standard deviation 1, Z  N.0; 1/,
and U is distributed chi-square with df .v/, U  2 .v/. If Z and U are independent, then the
random variable t D qZU follows a t -distribution with v degrees of freedom.
v
The pdf of t with degrees of freedom v is given by:
€..v C 1/=2/
f .t/ D p ; for 1 < t < 1:
v€.v=2/ .1 C t 2 =v/.vC1/=2
In general, when the sample size is less than 30 and population standard deviation is
unknown, then the random variable x is approximately normally distributed, thus it follows a
Student t -distribution. Furthermore, the difference between t -distribution and normal distribu-
tion is negligible when the sample size is sufficiently large. When population standard deviation
is unknown, estimating it from the sample results to a greater uncertainty and a more spread out
distribution (see Fig. 7.1).
From Fig. 7.1, it can be seen that the graphs of t -distribution is similar to standard normal
distribution in exception that t -distribution is a lower and wider; this attribute is prominent in
the t -distribution with df (1). Also, the graphs show the absolute and relative error for nor-
mal approximation. The graph of t -distribution with df (30) is approximately standard normal
distribution. Therefore, as the df increases, the t -distribution approaches the standard normal
distribution.
104 7. OTHER CONTINUOUS PROBABILITY DISTRIBUTIONS
0.45
0.4
0.35
0.3
t (df = 1)

f(X = x)
0.25
t (df = 5)
0.2
0.15
t (df = 10)
0.1 t (df = 30)
0.05 Std. Normal
0
-6 -4 -2 0 2 4 6
x

Figure 7.1: t -distribution with different degrees of freedom and standard normal distribution.

When the population standard deviation is unknown, the sample size is less than 30, and
the random variable x is approximately normally distributed, it follows a t -distribution. The test
statistics is:
.x /
tD p  t.n 1/;.˛=2/ :
s= n
When you use a t -distribution to estimate a population mean, the df are equal to one less
than the sample size, df D n 1.

Properties of the t -Distribution

The t -distribution has the following properties.
• The t -distribution is bell shaped and symmetric about the mean.
• The t -distribution is a family of curves, each determined by the degrees of freedom.
• The total area under a t -curve is 1 or 100%.
• The mean, median, and mode of the t -distribution are equal to zero.
• The variance of the t -distribution is v=.v 2/, where v is the degree of freedom and v > 2.

• When the degrees of freedom of t -distribution is sufficiently large, the t -distribution ap-
proaches the normal distribution.
 7.1. STUDENT-T DISTRIBUTION 105
Example 7.1
What is the value of t in the following: (i) t0:05 .20/, (ii) t0:01 .20/, (iii) t0:975 .25/, and
(iv) t0:95 .25/?

Solution:
Let’s make use of the t -distribution table (see Appendix A).
(i) t0:05 .20/ D 1:725.
We look at t -distribution table when significance level (˛ ) is 0.05 and df .v/ is 20, then
we get 1.725.
(ii) t0:01 .20/ D 2:528.
We look at t -distribution table when significance level (˛ ) is 0.01 and df .v/ is 20, then
we get 2.528.
(iii) t0:975 D 2:060.
Because of the symmetry nature of t -distribution, we look t -distribution table when sig-
nificance level (˛ ) is 0.025 and df .v/ is 25, then we get 2.060.
(iv) t0:95 .25/ D 1:708.
We use significance level (˛ ) of 0.05 and df .v/ is 25, since t -distribution is symmetric,
then we get 1.708.

R Codes to Obtain t -Values Using Example 7.1

The computation of the t -values in R is of the form:
qt(p, df, ncp, lower.tail = TRUE, log.p = FALSE)
where
qt is the quantile function for the t -distribution
p is the vector of probabilities
ncp is the non-centrality parameter delta;
if omitted, use the central t -distribution
lower.tail is logical; if TRUE (default), probabilities are P ŒX  x,
otherwise, P ŒX > x
log.p is logical; if TRUE, probabilities p are given as log.p/.
For the Example 7.1(i)–(iv), we can simply use the following R codes.
# Example 7.1(i)
alpha <- 0.05
df<-20
106 7. OTHER CONTINUOUS PROBABILITY DISTRIBUTIONS
t.alpha <- qt(alpha, df )
t.alpha
[1] -1.724718
t.abs<-abs(t.alpha) # t distribution is symmetric.
t.abs
[1] 1.724718

# Example 7.1(ii)
alpha<- 0.01
df<-20
t.alpha <- qt(alpha, df, lower.tail =TRUE )
t.alpha
[1] -2.527977
t.abs<-abs(t.alpha) # t distribution is symmetric.
t.abs
[1] 2.527977

# Example 7.1(iii)
alpha<- 0.975
df<-25
t.alpha <- qt(alpha, df, lower.tail =TRUE )
t.alpha
[1] 2.059539

(iv) t0:95 .25/ D 2:485.

# Example 7.1(iv)
alpha<- 0.95
df<-25
t.alpha <- qt(alpha, df, lower.tail =TRUE )
t.alpha
[1] 1.708141

Example 7.2
Suppose a random number X follows a t -distribution with v D 10 df. Calculate the probability
that the absolute value of X is less than 2.228.
 7.2. CHI-SQUARE DISTRIBUTION 107
Solution:

P .jX j < 2:228/ D P . 2:228 < X < 2:228/:

We write the probability in the RHS in terms of cumulative probabilities:

P .jXj < 2:228/ D P .X < 2:228/ P .X < 2:228/ :

V = 10
-2.228 0 2.228

Figure 7.2: t -distribution.

Since t -distribution is symmetric and t -table does not contain have a negative t -values, we can
write the probabilities as:

P .jX j < 2:228/ D P .X < 2:228/ P .X > 2:228/ :

From the t -table, the P .X < 2:228/ D 0:975 and P .X > 2:228/ D 0:025. The required prob-
ability is
P .jXj < 2:228/ D 0:975 0:025 D 0:95:

7.2 CHI-SQUARE DISTRIBUTION

In this section, we shall look at the definition, properties, and uses of chi-square. The chi-square
is denoted by 2 .v/ with the degree of freedom (v ).
Let Y1 ; Y2 ; : : : ; Yn be mutually independent standard normal random variables, and let
X D Y12 C Y12 C : : : C Yn2 has a chi-square distribution. The pdf of the chi-square distribution
is:
e x=2 x v=2
f .x/ D v=2 v
; for x  0;
2 € 2
where v is the shape parameter (df ) and € is the gamma function. For instance, €.x/ D .x 1/.
108 7. OTHER CONTINUOUS PROBABILITY DISTRIBUTIONS
Chi-square distribution being a squared of standard normal random variable takes only
non-negative values and it tends to be right skewed. Its skewness depends on the df or number
of observations. As the number of observations increases, the more symmetrical the chi-square
distribution becomes, and then tends to normal distribution. The higher the degrees of freedom,
the less the skewness of the chi-square (see Fig. 7.3).

0.8
0.7
0.6
0.5
df = 1
f(X = x)
0.4
df = 5
0.3
df = 10
0.2
0.1
0
-0.1
0 5 10 15 20 25
X

Figure 7.3: Chi-square distribution with different df.

The chi-square test statistics is given by:

X .Oi Ei /2
X2 D  2v :
Ei

where Oi is the observed value, Ei is the expected value and v is the df.

Properties of the Chi-Square Distribution

The chi-square distribution has the following properties.

• The chi-square distribution is a continuous probability distribution with the values ranging
from 0 to 1 (nonnegative).

• The mean of a chi-square distribution is the number of df, v .

• The mean of a chi-square distribution is the doubled the number of df, 2v .

• Chi-square has an additive property. The sum of independent 2 is itself a 2 variate.

Suppose two independent 21 and 22 with the df of v1 and v2 , respectively, then 21 C 22
is the same as 2 variate with degree of freedom v1 C v2 .
 7.2. CHI-SQUARE DISTRIBUTION 109
Uses of the Chi-Square Distribution
Chi-square can be use in the following ways.
• Independence of two criteria of the classification of qualitative variables.
• Relationships between categorical variables (contingency tables).
• Sample variance study when the underlying distribution is normal.
• Tests of deviations of differences between expected and observed frequencies (one-way
tables).
• The chi-square test (a goodness-of-fit test). This is to test how well a sample data fits a
certain distribution.
• Confidence interval estimation for a population standard deviation of a normal distribution
from a sample standard deviation.

Example 7.3
What is the value of chi-square in the following: (i) 20:05 .15/, (ii) 20:995 .30/, and (iii) 20:900 .85/.

Solution:
We look at a chi-square table for the significance level (˛ ) and the degree of freedom (v ).
(i) 20:05 .15/ D 7:261.
(ii) 20:995 .30/ D 53:67.
(iii) 20:900 .85/ D 102:07.
We obtained the answer for 20:900 .85/ D 68:7845 from finding the average of
20:900
.80/ D 96:5782 and 20:900 .90/ D 107:565, since we cannot see 20:900 .85/ directly from
the chi-square table. We use the average method to estimate the value.
Note that we use the chi-square table that has the probabilities of P ŒX  x.

R Codes to Obtain Chi-Square Values from the Table Using Example 7.3
The chi-square can be obtain in R using the function:
qchisq(p, df, ncp, lower.tail = TRUE, log.p = FALSE)
110 7. OTHER CONTINUOUS PROBABILITY DISTRIBUTIONS
where
qchisq is the quantile function for the Chi-square distribution
p is the vector of probabilities
ncp is the non-centrality parameter delta;
if omitted, use the central 2 distribution
lower.tail is logical; if TRUE (default), probabilities are P ŒX  x,
otherwise, P ŒX > x
log.p is logical; if TRUE, probabilities p are given as log.p/.
The following are the codes to solve the Example 7.3(i)–(iii):
qchisq(0.05, df=15, lower.tail = TRUE)
[1] 7.260944

qchisq(0.995, df=30, lower.tail = TRUE)

[1] 53.67196

qchisq(0.900, df=85, lower.tail = TRUE)

[1] 102.0789
We easily got the inverse of the p -value for a chi-square distribution in Example 7.3.

7.3 F-DISTRIBUTION
The F-distribution is the ratio of two independent chi-square distributions divided by their
degrees of freedom v1 and v2 . The probability density function of the F-distribution is

v1 Cv2

 v1  v21 v
1
€ 2 v2
x2
f .x/ D  ; x  0;
v1

v2

 v1 x
 v1 Cv2
2
€ 2
€ 2
1C v2

where v1 and v2 are the shape parameter and € is the gamma function.
The F-distribution is mainly spread out when the df are small. Thus, as the df decreases,
the F-distribution is more dispersed. The distribution of F-distribution is asymmetric that has a
minimum value of 0 and maximum of infinity. The curve reaches a peak not far to the right of 0,
and then gradually approaches the highest value of F in the horizontal axis. The F-distribution
approaches the horizontal axis but never touches the horizontal axis (see Fig. 7.4).

Properties of the F-Distribution

The F-distribution has the following properties.
• The F-distribution is asymmetric about the mean but skewed to the right.
 7.3. F-DISTRIBUTION 111

0.8
0.7
0.6
0.5

f(X = x)
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6
x

Figure 7.4: F-distribution with df (v1 D 3 and v2 D 5).

• The F-distribution is a family of curves that determined by two df.

• The F-distribution is a continuous probability distribution with the values ranging from 0
to 1 (nonnegative).
• The curve approaches normal as the df for the numerator and denumerator become larger.
v2
• The mean of the F-distribution is v2 2
.
2v22 .v1 Cv2 2/
• The variance of the F-distribution is .
v1 .v2 2/2 .v2 4/

• Any changes in the first parameter .v1 / does not change the mean of the distribution, but
the density of the distribution is shifted from the tail of the distribution toward the center.

Uses of the F-Distribution

The F-distribution is useful in the following.
• The F-distribution is used to test the validity of a multiple regression model.
• The F-distribution is used to test hypotheses about the equality of two independent pop-
ulation variance.
• It is used to draw an inference about the data that is drawn from a population.
112 7. OTHER CONTINUOUS PROBABILITY DISTRIBUTIONS
Example 7.4
What is the value of F in the following: (i) F.0:05/ .8; 12/, (ii) F.0:95/ .10; 15/, and
(iii) F.0:995/ .20; 50/.

Solution:
Let’s look at F table for the significance level (˛ ) and the degrees of freedom (v1 and v2 ) for the
numerator and denominator, respectively.
(i) F.0:05/ .8; 12/ D 0:3045.
(ii) F.0:95/ .10; 15/ D 2:544.
(iii) F.0:995/ .20; 50/ D 2:470.
Note that we use the F table that has the probabilities of P ŒX  x, the left side of the
curve.

R Codea to Obtain F Values from the F-Distribution Table Using Example 7.4
The F value can be obtained in R using the function:
qf(p, df1,df2, ncp, lower.tail = TRUE, log.p = FALSE)
where
qf is the quantile function for the F-distribution
p is the vector of probabilities
df1 is the degree of freedom for numerator
df2 is the degree of freedom for denominator
ncp is the non-centrality parameter delta;
if omitted, use the central F-distribution
lower.tail is logical; if TRUE (default), probabilities are P ŒX  x,
otherwise, P ŒX > x
log.p is logical; if TRUE, probabilities p are given as log.p/.
With these short R statements we are able to obtain the F values for Example 7.4(i)–(iii).
qf (0.05, 8, 12, lower.tail = TRUE, log.p = FALSE)
[1] 0.3045124

qf (0.95, 10, 15, lower.tail = TRUE, log.p = FALSE)

[1] 2.543719

qf (0.995, 20, 50, lower.tail = TRUE, log.p = FALSE)

[1] 2.470161
 7.4. EXERCISES 113
7.4 EXERCISES
7.1. (a) Mention the continuous probability distributions that you know.
(b) Define student t -distribution and state its properties.
(c) With the aid of charts, explain how t -distribution can be used as an approximation
to the normal distribution.
(d) What is the value of t in the following: (i) t0:95 .18/, (ii) t0:1 .25/, (iii) t0:995 .30/,
and (iv) t0:99 .10/.
7.2. (a) Define student chi-square distribution and state its properties.
(b) What are the uses of chi-square distribution?
(c) What is the value of chi-square in the following: (i) 20:95 .20/, (ii) 20:995 .40/,
(iii) 20:975 .25/, and (iv) 20:99 .50/.
7.3. (a) Define student F-distribution and state its properties.
(b) What are the uses of F-distribution?
(c) What is the value of F in the following: (i) F.0:95/ .10; 12/, (ii) F.0:99/ .5; 25/,
(iii) F.0:05/ .10; 30/, and (iv) F.0:01/ .18; 10/.
 115

CHAPTER 8

Sampling and Sampling

Distribution
Sampling distributions emanated from a group of selected data being calculated using statistics
such as mean, median, mode, range, standard deviation, and variance. The distribution is useful
in testing a hypothesis and making inferences about the population.
The sampling distribution relies on the following:
• an underlying distribution of the population;
• the statistic under investigation;
• sampling techniques; and
• sample size.
In this chapter, we shall consider the probability and non-probability sampling, sampling
techniques, sampling distribution of mean and proportion, and central limit theorem.

8.1 PROBABILITY AND NON-PROBABILITY SAMPLING

A method that uses random sampling techniques to generate a sample, where every sampling
unit in the population has an equal chance of being selected, is called a probability sampling.
Randomization is the key in probability sampling techniques. Suppose a population contains
100 sampling units, and each sampling unit has a chance of 1/100 being selected by probability
sampling techniques. Under probability sampling, the sample selected gives a true representa-
tion of the population that it comes from. The most common probability samplings are simple
random, systematic, stratified, and cluster samples.

Advantages of Probability Sampling

• The outcomes are reliable.
• It increases accuracy of error estimation.
• No sampling bias and systematic error.
• It is good in making inferences about the population.
116 8. SAMPLING AND SAMPLING DISTRIBUTION
Disadvantage of Probability Sampling
• It consumes time.

• More expensive than non-probability sampling.

• More complex than non-probability sampling.

Non-Probability Sampling
It involves a non-random sampling of the sampling units, thus only specific members of the
population have a chance of being selected. The most widely used non-probability methods are
judgment sampling, quota sampling, convenience sampling, and extensive sampling. A non-
probability sampling technique is based on subjective judgment and it is used for exploratory
studies, e.g., pilot survey.

Advantages of Non-Probability Sampling

• It easily gives the description about the sample.

• It saves time.

• Less expensive compared to probability sampling.

• It is effective when unfeasible to conduct probability sampling.

Disadvantages of Non-Probability Sampling

• Lack of representation of the population.

• Difficult to estimate sampling variability.

• Difficult to identify possible bias.

• Lower level of generalization of research findings.

8.2 PROBABILITY SAMPLING TECHNIQUES—SIMPLE

RANDOM, SYSTEMATIC, STRATIFIED, AND
CLUSTER SAMPLES
8.2.1 SIMPLE RANDOM SAMPLING (SRS)
In a simple random sampling (SRS) technique, each unit included in the sample has an equal
chance of inclusion in the sample. In a homogenous population, SRS provides an unbiased
and better estimate of the population parameters. In the process of selection, the probability of
selecting the first sampling unit will not necessarily be the same probability of selecting second
sampling unit.
 8.2. PROBABILITY SAMPLING TECHNIQUES 117
In the event that every selection of the sampling unit remains constant for every selection
with a chance of 1=N , where N is the total number of the elements in the population, then
the procedure is called simple random sampling with replacement (SRSWR). In a SRSWR,
there are N n possible samples with each sample having a N1n chance of selection, where n is the
sample size.
However, if in the first selection, each member of the population has an equal chance of
selection, with probability of 1=N . In the selection of second unit, all the remaining sampling
units have probability of N1 1 of selection. Any successive selection in a SRS is referred as to
a simple random sample
without replacement (SRSWOR). If sampling is done without re-
placement, there are Nn possible samples and each sample has a chance of N1 selection. More
.n/
often, the SRSWOR is regarded as a simple random sample.
The following methods are used to select units of a simple random sample.
1. Random Number Table
In using a random number table, all units of the population are numbered either from 1 to
N or 0 to N 1. Suppose we have a population size of 50 and we want to select a sample
of size 10. We can number sampling units in two digits like 01, 02,…, 49, 50. We shall
read a two-digit random number from the random number table, any two-digit numbers
that are greater than 50 are ignored, and any repetition of a number is also ignored if the
sampling is without replacement (see Table 8.1).
We can read the random number from any columns or row of the random number table.
From Table 8.1, let’s start from first column downward to form our two-digit numbers.
Thus, we have: 11, 21, 10, 36, 49, 17, 24, 16, 29, and 12. Some numbers such as 73, 64,
51, 99, 71, 65, 95, 61, 78, 62, 59, and 96 are discarded in the list because they are not part
of the population. Furthermore, let’s consider a population of 100, and we are asked to
take a random sample of 10 from the population. We shall read the random number table
in 3-digits, that is we will assign a number from 001–100. From Table 8.1, we have: 047,
046, 098, 070, 024, 088, 004, 084, 094, and 097 in that order.
2. Lottery Method
The lottery method is very close to modern methods of sample selection. This process
involves labeling all sampling units from 1–N in the sampling frame. The same numbers
(1–N ) from the sampling frame are written on slips of paper. Next, the slips are mixed
thoroughly, and then selections are made randomly. This method is advisable when the
size of the population is small. It may be difficult if the sample size is large since it is not
easy to mix properly the heap of paper slips.
118 8. SAMPLING AND SAMPLING DISTRIBUTION
Table 8.1: Random digits

3. Computer Generation
Computers have some in-built programs that help to generate random samples at ease.
These are mostly used in selection of winners of applicants for plots of lands, visa lottery,
etc.

Example 8.1
Suppose a population contains the following: 4, 6, 10, 11, 15, 17, and 20 units. Select a sample
size of 2. What are the possible samples with replacement and without replacement?
 8.2. PROBABILITY SAMPLING TECHNIQUES 119
Solution:
Table 8.2: SRSWR and SRSWOR samples

Samples 4 6 10 11 15 17 20
4 (4, 4) (4, 6) (4, 10) (4, 11) (4, 15) (4, 17) (4, 20)
6 (6, 4) (6, 6) (6, 10) (6, 11) (6, 15) (6, 17) (6, 20)
10 (10, 4) (10, 6) (10, 10) (10, 11) (10, 15) (10, 17) (10, 20)
11 (11, 4) (11, 6) (11, 10) (11, 11) (11, 15) (11, 17) (11, 20)
15 (15, 4) (15, 6) (15, 10) (15, 11) (15, 15) (15, 17) (15, 20)
17 (17, 4) (17, 6) (17, 10) (17, 11) (17, 15) (17, 17) (17, 20)
20 (20, 4) (20, 6) (20, 10) (20, 11) (20, 15) (20, 17) (20, 20)
2
(a) SRSWR: Possible sample = 7 = 49
Samples 4 6 10 11 15 17 20
4
6 (6, 4)
10 (10, 4) (10, 6)
11 (11, 4) (11, 6) (11, 10)
15 (15, 4) (15, 6) (15, 10) (15, 11)
17 (17, 4) (17, 6) (17, 10) (17, 11) (17, 15)
20 (20, 4) (20, 6) (20, 10) (20, 11) (20, 15) (20, 17)
7
(b) SRSWOR: Possible sample = (2) = 2

8.2.2 SYSTEMATIC SAMPLING

This sampling procedure is easier done than the simple random sampling, although it uses tech-
nique of SRS in selection of the first sample from the population. Thus, each sampling unit has
an equal chance of selection in the first sample to be taken from the population. The subsequent
selection follows a predefined pattern. Consider a population of N from which a sample size
of n is drawn. Assume that k D N=n. Initially, a random sample shall be taken from 1–k that
will be the first sample, then for every k th unit in the population from a sample systematically.
For example, a recruiting agency received 1,000 qualified applicants, out of which only 50 ap-
plicants are to be considered for jobs. If he decided to interview only 50 applicants by systematic
sampling, then k D 1000 50
D 20. Therefore, a random sampling shall be done between 1–20, say
13th item is selected, so every 20th applicant in the population of 1,000 will form a sample (i.e.,
13th, 33th, 53th, : : :, 953th, 973th, and 993th).
120 8. SAMPLING AND SAMPLING DISTRIBUTION
8.2.2.1 Advantages of Systematic Sampling
• It is simple and convenient to actualize.

• It more easily represents a population without numbering each member of a sample.

• The samples from system sampling are based on precision in the member selection.

• The risk involved in the process of selection is minimal.

• It can be used in a heterogeneous population.

8.2.2.2 Disadvantages of Systematic Sampling

• It cannot be used where the size of the population is not available or being approximated.

• It requires a population to characterize with a natural degree of randomness in order to

mitigate the risk of selecting samples.

• It has a higher risk of manipulation of data thereby increasing the likelihood of achieving
a targeted outcome rather than randomness of a dataset.

8.2.3 CLUSTER SAMPLING

Cluster sampling is a technique of selecting all individuals or group of individuals through a
random selection. Its procedure involves a selection of groups or clusters; then from each group
the individual subjects are selected either by SRS or systematic random sampling. It is desirable
for clusters to be made up of heterogeneous units or elements. The samples can be either entire
cluster or group rather than a subset of the cluster or group. For example, if a survey is to take a
sample of agro-allied companies in the western region of Nigeria. The western region of Nigeria
is made up of six States, and it is fair to have samples representation from each of the state
rather than concentrating on three or four States of the region. The commonly used cluster is
geographical location.
Cluster sampling can be a single-stage or multi-stage cluster sampling. In a single-stage
cluster sampling, clusters are selected by a SRS and then data is collected from every unit of the
sampled clusters. However, a two-stage process involves a simple random sampling of clusters
and then a SRS of the units in each of sampled cluster.

8.2.3.1 Advantages of Systematic Sampling

• It is economical to generate sampling frame.

• It is less time consuming for listing and implementation.

• It works for larger sample with a similar fixed cost.

 8.2. PROBABILITY SAMPLING TECHNIQUES 121
8.2.3.2 Disadvantages of Systematic Sampling
• It gives less information about each observation than SRS.

• It may show variation within the group.

• Elements of the cluster may possess similar attributes.

• High estimates of standard errors when compared with other probability sampling tech-
niques.

8.2.4 STRATIFIED RANDOM SAMPLING

Stratified random sampling is a probability sampling technique in which the population is par-
titioned into relatively homogeneous groups know as strata. In each stratum, a simple random
sample is use for the selection of the sampling units. Thus, the statistic computed from the strata
is aggregated to make inference about the population. Suppose a team of researchers wants to
know the average performance of students in a private university that has five faculties (Engi-
neering, Science, Law, Social Science, and Medicine). Five hundred students study Engineering,
1,200 students study Science, 800 students study Law, 1,550 students study Social Science, and
350 students study Medicine. The distribution of the students population is depicted in the pie
chart in Fig. 8.1.

8%
12%
Engineering
35% Science
Law
27% Social Science
Medicine
18%

Figure 8.1: The distribution of students by faculty.

If a stratified random sample of 1,000 students are be selected in the above example, it
is expected to select students according to the size of the faculties. That is, 120 students from
Engineering, 270 students from Science, 180 students from Law, 350 students from Social
Science, and 80 students from Medicine suppose to be among the selection to achieve a better
precision.
122 8. SAMPLING AND SAMPLING DISTRIBUTION
8.2.4.1 Advantages of Stratified Random Sampling
• It gives a high representation of the population.

• It allows a valid statistical inference from the data collected.

• It improves the potential for units to evenly distribute over the population, hence improve
precision when compared with SRS.

8.2.4.2 Disadvantages of Stratified Random Sampling

• It cannot be practically used if a complete list of the population is not available.

• The list of the population is needed to be clearly delineated into each stratum.

8.3 NON-PROBABILITY SAMPLING TECHNIQUES

8.3.1 JUDGMENT SAMPLING
Judgment sampling is a type of non-probability sampling technique where the researcher uses
his own experience and prior knowledge about the situation before the commencement of se-
lection; the sample is taken without use of any statistical tools. Suppose that a panel decides
to understand the reasons why startup companies failed. The panel should understand that they
should form the samples from the business owners to obtain useful information about the factors
that responsible for failure in business. This process is called judgment sampling.

Advantages of Judgment Sampling

• The method is easily understood.

• It requires no specific knowledge of statistics.

• It consumes less time to actualize.

Disadvantages of Judgment Sampling

• The researcher may be biased in selecting the sample.

• There is no logic in the process of selection of the sample.

• It has no capacity to extrapolate the population as samples selection are not good repre-
sentation of the population.

• In some cases, it gives rise to vague conclusions.

 8.3. NON-PROBABILITY SAMPLING TECHNIQUES 123
8.3.2 QUOTA SAMPLING
This is a non-probability sampling technique, therefore the samples selected have the same pro-
portions of individuals as the entire population with as regard to known characteristics. Its pro-
cedures include the following.
• Dividing the population into exclusive sub-groups.
• Identification of proportions of the subgroups in the population and this proportion has
to be maintained in the sampling process.
• Selection of subjects from subgroups using the same proportion in the previous step.
• Ensure the sample drawn is a representative of the population, thus, it is easy to under-
stand the characteristics noticed in each subgroup. For example, if the population size is
2,000 consists of 1,200 men and 800 women and researcher is interested in the 30% of the
population, it means that he will select a sample of 600 sampling units. The 600 sampling
units is made up of 360 men and 240 women.

Advantages of Quota Sampling

• It allows sampling of a subgroup of interest and investigates the traits and characteristics.
• It helps to establish the relationship in terms of traits and characteristics between sub-
groups.

Disadvantages of Quota Sampling

• Selected characteristics of the population are accounted for while forming the subgroups.
• Its process may result to over-represent in the sample.

8.3.3 CONVENIENCE SAMPLING

This approach is an example of a non-probability sampling where a researcher draws a sample
because of the “convenience” of the data sources. Here, convenience could be the nearest subjects
or availability of resources to be needed in carry out the research project. It is also known as
accidental sampling. It is commonly used in pilot studies before embarking on the main project.
For example, to determine profitability of firms, a researcher may first look at some firms in his
vicinity and sample opinions of the business owners in the area before extending the research
work to a larger communities.

Advantages of Convenient Sampling

• It is time efficient.
• Its outcomes can be used to refine some questions in the questionnaire.
124 8. SAMPLING AND SAMPLING DISTRIBUTION
• It is less expensive since research is able to use available resources.

Disadvantages of Convenient Sampling

• It is not a good representative of the entire population.
• It may lead to a biased outcome and wrong conclusion because of uncontrollable sample
representation.

8.4 SAMPLING DISTRIBUTION OF THE MEAN

1. The mean of the sampling distribution is equal to the mean of the population, that is,
X D . To establish this fact, we can use the data in Example 8.1.
Suppose a sample of size 2 is drawn from a population that contains the following: 4, 6,
10, 11, 15, 17, and 20 units.
4 C 6 C 10 C 11 C 15 C 17 C 20
The population mean D D 11:86:
7

For SRSWR, the distribution of the sample means appears in Table 8.3.

Table 8.3: SRSWR distribution of samples

Samples 4 6 10 11 15 17 20
4 4.0 5.0 7.0 7.5 9.5 10.5 12.0
6 5.0 6.0 8.0 8.5 10.5 11.5 13.0
10 7.0 8.0 10.0 10.5 12.5 13.5 15.0
11 7.5 8.5 10.5 11.0 13.0 14.0 15.5
15 9.5 10.5 12.5 13.0 15.0 16.0 17.5
17 10.5 11.5 13.5 14.0 16.0 17.0 18.5
20 12.0 13.0 15.0 15.5 17.5 18.5 20.0

4:0 C 5:0 C 7:0 C


C 17:5 C 18:5 C 20:0

X D D 11:86:
49
Therefore, the mean of distribution of sample means is same as population mean.
For SRSWOR, the distribution of the sample means appear in Table 8.4.

5:0 C 7:0 C 7:5 C


C 16:0 C 17:5 C 18:5

X D D 11:86:
21
Hence, the mean of distribution of sample means is same as population mean.
 8.4. SAMPLING DISTRIBUTION OF THE MEAN 125
Table 8.4: SRSWOR distribution of samples

Samples 4 6 10 11 15 17 20
4
6 5.0
10 7.0 8.0
11 7.5 8.5 10.5
15 9.5 10.5 12.5 13.0
17 10.5 11.5 13.5 14.0 16.0
20 12.0 13.0 15.0 15.5 17.5 18.5

In summary, irrespective of the sampling processes, the mean of distribution of means still
remain the mean of the population.
2. The standard error of the mean is equal to the ratio of the standard error of the popula-
tion to the square root of the sample size. This can also be demonstrated as follows using
(SRSWR) Table 8.3. The standard deviation of the population 4, 6, 10, 11, 15, 17, and 20
units is:
.4 11:86/2 C .6 11:86/2 C


C .20 11:86/2
D D 5:38:
7
Similarly, the standard deviation of the distribution of sample means is calculated as:

.4 11:86/2 C .5 11:86/2 C


C .18:5 11:86/2 C .20 11:86/2

X D D 3:8065
49
 5:38
X D p D p D 3:8065:
n 2
This result shows the relationship between the standard deviation of distribution of means
and the population standard deviation. That is, the standard deviation of distribution of
means (X ) is the population standard deviation ( ) divided by the square root of sample
size .n/.

Sampling Distribution of the Mean Using R

# To compute the mean of the population
pop=c(4,6,10, 11, 15, 17, 20)
pop_mean<-sum(pop)/7
pop_mean
[1] 11.85714
126 8. SAMPLING AND SAMPLING DISTRIBUTION

# To compute the mean of the distribution of means

# for SRS with replacement
sWR<-49
sampWR<- c(4, 5, 7, 7.5, 9.5, 10.5, 12, 5, 6, 8,
8.5, 10.5, 11.5, 13, 7, 8, 10, 10.5, 12.5, 13.5,
15, 7.5, 8.5, 10.5, 11, 13, 14, 15.5, 9.5, 10.5,
12.5, 13, 15, 16, 17.5, 10.5, 11.5, 13.5, 14, 16,
17, 18.5, 12, 13, 15, 15.5, 17.5, 18.5, 20)
sampWR_mean=sum(sampWR)/sWR
sampWR_mean
[1] 11.85714

# To compute the standard deviation of the distribution

# of means for SRS with replacement
sampWR_std<-sd(sampWR)
sampWR_std
[1] 3.845994

# To compute the mean of the distribution

# of means for SRS without replacement
sWOR<-21
sampWOR<-c(5,7,7.5,9.5,10.5,12,8,8.5,10.5,11.5,13,
10.5,12.5,13.5,15,13,14,15.5,16,17.5,18.5)
sampWOR_mean<-sum(sampWOR)/sWOR
sampWOR_mean
[1] 11.85714

# Population standard deviation

p<-7
pop_std<-sd(pop)*sqrt((p-1)/p)
pop_std
[1] 5.38327

# Sample standard deviation

n<-2 # sample selection of 2
samp_std<- pop_std/sqrt(n)
samp_std
[1] 3.806546
 8.5. CENTRAL LIMIT THEOREM AND ITS SIGNIFICANCE 127
8.5 CENTRAL LIMIT THEOREM AND ITS SIGNIFICANCE
The central limit theorem states that let X1 ; X2 ; : : : ; Xn be a random sample from a distribution
with finite mean () and finite variance ( 2 ). For a sufficiently large sample size n, i.e., .n  30/,
the following hold.

1. Sample mean (X/ is approximately normal.

2. The mean of the sample means is equal to the population mean E X D .

2
3. Variance var.X/ D X2 D n
.

This theorem can be written mathematically as:

 
2
X ! N ; as n ! 1:
n

Alternatively,
X 
zD  N .0; 1/ as n ! 1:

The main significance of the central limit theorem is that it enables us to make probability
statements about the sample mean when compared with the population mean. Let’s distinguish
two separate variables with their corresponding distributions:

1. Let X be a random variable that measures a single element from the population, then the
distribution of X is the same as distribution of the population with mean ./ and standard
deviation ./.

2. Let X be a sample mean from a population; then the distribution of X is its sampling
p
mean with mean .X / and standard deviation .X D = n/, where n is the sample size.

Example 8.2
Let X be the mean of a random sample of 100 selected from a population of mean 30 and
standard deviation of 5. (a) What is the mean and standard deviation of X ? (b) What is the
probability that the value of X falls between 29 and 31? (c) What is the probability that value
of X is more than 31?

Solution:
 p5
(a) X D  D 30; X D p
n
D
100
D 0:5.
128 8. SAMPLING AND SAMPLING DISTRIBUTION
(b)
 
29
X 31 X
P .29 < X < 31/ D P z
0:5 0:5
 
29 30 31 30
DP z
0:5 0:5
D ˆ .2/ ˆ . 2/ D 0:9772 0:0228 D 0:9544:

(c)
 


31 30
P X > 31 D 1 P X  31 D 1 P
0:5
D1 ˆ . 2/ D 1 0:9772 D 0:0228:

R Codes

# Example 8.2a solution

sampl.mean<-30
mu<- sampl.mean
sample.size<-100
sigma<-5
std.dev_mean<-sigma/sqrt(sample.size)
std.dev_mean
[1] 0.5

# Example 8.2b solution

sampl.mean<-mu
lower.mean<-29
z29<-(lower.mean- mu)/ std.dev_mean
p_29<-pnorm(z29, mean =0, sd = 1, lower.tail = TRUE,
log.p = FALSE)
p_29
[1] 0.02275013
upper.mean<-31
z31<-(upper.mean- mu)/ std.dev_mean
p_31<-pnorm(z31, mean =0, sd = 1, lower.tail = TRUE,
log.p = FALSE)
p_31
[1] 0.9772499
 8.5. CENTRAL LIMIT THEOREM AND ITS SIGNIFICANCE 129
btw_p29_p31<- p_31- p_29
btw_p29_p31
[1] 0.9544997

# Example 8.2c solution

p.greater31<-1-p_31
p.greater31
[1] 0.02275013

Example 8.3
A branch manager of a microfinance bank claims that the average number of customers that
deposit cash on monthly basis is 1,250 customers with a standard deviation of 130 customers.
Assume the distribution of cash lodgement is normal. Find:
(a) the probability that less than or equal to 1,220 customers will deposit cash in a month and
(b) the probability that the mean of a random sample of 15 months, less than 1,200 customers
deposit cash.

Solution:
(a)  
1220 1250
P .X < 1220/ D P z  D P .z  0:231/ D 0:4090:
130

(b) !

1200 1250
P X < 1200 D P z 130
D P .z  1:49/ D 0:0681:
p
15

R Codes

# Example 8.3a solution

mean<-1250
std<-130
x<-1220
z<-(x- mean)/ std
p_1220<-pnorm(z, mean =0, sd = 1, lower.tail = TRUE, log.p = FALSE)
p_1220
[1] 0.408747
130 8. SAMPLING AND SAMPLING DISTRIBUTION
# Example 8.3b solution
mean<-1250
std<-130
n<-15
std.err<-std/sqrt(n)
x.bar<-1200
z.xbar<-(x.bar- mean)/ std.err
p.xbar_1200<-pnorm(z.xbar, mean =0, sd = 1,
lower.tail = TRUE, log.p = FALSE)
p.xbar_1200
[1] 0.06816354

8.6 EXERCISES
8.1. (a) Explain what you understand by the sampling distribution.
(b) Differentiate between probability sampling and non-probability sampling.
(c) What are the merits and demerits of probability sampling and non-probability
sampling?
8.2. (a) Mention and describe types of probability sampling techniques you know.
(b) State the advantages and disadvantages of the sampling techniques mentioned
in (a).
8.3. List non-probability sampling techniques and state the advantages and disadvantages
of the techniques.
8.4. With the aid of a demonstrative example, explain the concept of sampling distribution
of means.
8.5. (a) State the central limit theory and its importance.
(b) Assume that the length of time of calls is normal, with average of 60 s and standard
deviation of 10 s. Find the probability that the average time obtained from a sample
of 35 calls is 55 s of the entire population average.
8.6. Suppose a normally distributed population has mean and standard deviation of 75 and
12, respectively.
(a) What is the probability that a random element X selected from the population
falls between 72 and 75?
(b) Calculate the mean and standard deviation of X for a random sample of size 30.
(c) Calculate the probability that the mean of selected sample size of 30 from the
population is between 72 and 75.
 131

CHAPTER 9

Confidence Intervals for Single

Population Mean and
Proportion
When a statistic is computed from a sample of data to estimate the population parameter, due to
the variability in the sample collected, we have to reduce our risk of estimation by constructing
a range in which the exact value of the population parameter lies. This process gives us the
self-assurance at a certain percentage that the estimate is capable of predicting the real value of
the population of interest. This concept describes the confidence interval or confidence limit. In
this chapter, we are focusing on the constructing of confidence interval (CI) for both mean and
proportion.

9.1 POINT ESTIMATES AND INTERVAL ESTIMATES

When a sample is drawn from a population, the fact obtained or quality computed from the
sample is called a statistic. This statistic is closely described the characteristics obtained from
the population, and then the estimate or a single value from the sampled data is referred to as
a point estimate. For example, a human resource person may want to know the productivity
of workers in a firm. He can use the number of cases closed per month as a metric to measure
productivity. He discovered that the average cases closed per month was 22; thus the value (22)
is the point estimate.
On the other hand, in order to know how well the sample statistic accurately describes
the population, we compute a range of values that we are confident enough to say the true value
of population parameter lies. The interval estimate offers a measure of exactness of the point
estimate by giving an interval that contains the plausible values. In most cases, we use a 95%
confident limit for the estimation; this implies that we can boldly say that our estimation about
the population plausible values fall within the specified range of values in 95 out of 100 cases.
This phenomenon is referred to as interval estimation. For instance, a researcher may like to
know the number of cars plying a road on a daily basis. He may conduct the traffic census for
some days and conclude that the average number of car plying a road daily is between 180–
200 cars. This scenario can be represented in normal curve, as shown in Fig. 9.1.
132 9. CONFIDENCE INTERVALS FOR SINGLE POPULATION MEAN AND PROPORTION
Mean = 190

95%

+10
Margin of Error
Lower Bound Upper Bound

2.5% 2.5%

180 Confidence Interval 200

Figure 9.1: Confidence interval.

9.2 CONFIDENCE INTERVALS FOR MEAN

If a sample is drawn many times, there is a certain percentage of boldness that a range of values
computed from the sample will probably contain an unknown population parameter, and this
percentage is the CI. CI is the probability that a value will fall between an upper and a lower
bound of a probability distribution. Suppose an analyst forecasted that the return on DJIA in
the first quarter of the year will fall within 0.5% and C2.5% at 95% CI. This means that the
analyst is 95% sure that the returns on DJIA in the first quarter of the year will lie between
0.5% and C2.5%. CIs can be used to bound numerous statistics such as the mean, proportion,
and regression coefficients (covered in Chapter 11). CI is constructed as follows:

CI D sample statistic ˙ critical value  standard error of estimate:

For example, if the sample size n < 30,  is unknown, and the population is normally distributed,
then we should use the Student t -distribution.
However, if n < 30,  is unknown, and the population is not normally distributed, then we
should use nonparametric statistics (not covered in this book). Specifically, the CI for population
mean is constructed as:
s
CI D x ˙ tn 1;q  p : (9.1)
n
A 100 .1 ˛/% confidence region for  contains:
s s
x tn 1;q  p    x C tn 1;q p ; (9.2)
n n
 9.2. CONFIDENCE INTERVALS FOR MEAN 133
where ˛ represents level of significance, n represents sample size, s is the standard deviation, x
is the mean, t is the critical region from the t -distribution table, and q is the quantile (usually
q D 1 ˛2 for two-tailed test and q D 1 ˛ for one-tailed test).
Suppose we want to construct a 95% CI for an unknown population mean, then a 95%
probability that CI will contain the true population mean could be calculated as follows:
 
s s
P x tn 1;˛=2  p    x C tn 1;˛=2  p D 0:95: (9.3)
n n

Example 9.1
A random sample of 25 customers at a supermarket spent an average of $2,000 with a standard
deviation of $200. Construct a 95% CI estimating the population mean of purchase made at the
supermarket.

Solution:
100.1 ˛/% D 95% ) 100 100˛ D 95 ) ˛ D 0:05

x D 2000; s D 200; n D 25; tn 1;1 ˛=2 D t24;0:975 D 2:064.from t -table/

   
200 200
CI D 2000 2:064  p    2000 C 2:064  p
25 25
CI D 1; 917:44    2; 082:56:

This implies that the average purchase by the customers in the supermarket falls between
$1,917.44 and $2,082.56.

R Codes for the Example 9.1

The R code below shows how to use R to get CIs.

sample.mean <- 2000

sample.std <- 200
n <- 25
std.error <- qt(0.975,df=n-1)*sample.std/sqrt(n)
lower.limit <- sample.mean-std.error
upper.limit <- sample.mean+std.error
conf.interval<-c(lower.limit, upper.limit)
conf.interval
[1] 1917.444 2082.556

The average purchase by the customers in the supermarket lies between $1,917.44 and $2,082.56.
134 9. CONFIDENCE INTERVALS FOR SINGLE POPULATION MEAN AND PROPORTION
In addition, if n  30 and  is known or n < 30,  is known, and the population is nor-
mally distributed, then we use:
 
 
x z1 ˛=2  p    x C z1 ˛=2  p : (9.4)
n n
If n  30 and  is unknown, the standard deviation s of the sample is used to approximate the
population standard deviation  , then we have:
 
s s
x z1 ˛=2  p    x C z1 ˛=2  p : (9.5)
n n

Example 9.2
A sales manager of a company envisage that there dramatic drop in the sale of a particular
product. He took a simple random sample of 50 sales records from the previous days. The sales
(in dollars) was recorded and some summary measures are provided: n D 22, x D 5200 and
s D 400. Assuming that the sales is approximately normal. (a) Construct a 95% CI for the mean
sales of the product. (b) Interpret your result in (a).

Solution:
(a) n D 50, x D 5200 and s D 400 and z0:025 D 1:96.

   
400 400
CI D 5;200 1:96  p    5;200 C 1:96  p
50 50
CI D $5;089:13    $5;310:87 or  D Œ$5;089:13; $5;310:87:

(b) The true sales of the product lies between $5,089.13 and $5,310.87.

R Codes to Calculate the Confidence Interval for the Example 9.2

sample.mean <- 5200
sample.std <- 400
n <- 50
std.error <- qnorm(0.975)*sample.std/sqrt(n)
lower.limit <- sample.mean-std.error
upper.limit <- sample.mean+std.error
conf.interval<-c(lower.limit, upper.limit)
conf.interval
[1] 5089.128 5310.872
This confirms results in the Example 9.2.
 9.3. CONFIDENCE INTERVALS FOR PROPORTION 135
9.3 CONFIDENCE INTERVALS FOR PROPORTION
Suppose we are interested in estimating the proportion of people with a certain ailment in a
population. Let’s consider a diagnosis of such ailment as a “success” and lack of diagnosis of the
ailment as a “failure.” Let X be the number of people with a diagnosis of the ailment, then the
sample proportion is computed by pO D xn , where n is the sample size.
The sampling distribution of pO is approximately normal with pO D p and pO D p.1n p/ .
Therefore, the CI for population proportion is calculated as:
r
p.1
O p/
O
CI D pO ˙ z1 ˛2  : (9.6)
n
The sampling distribution of pO can be approximated by a normal distribution when npO  5.

Example 9.3
In an opinion poll to know whether to establish a National Grazing Reserve bill in Nigeria
or not, a random sample of 8,500 participants were selected, only 6,250 respondents were in
support of the bill while others moved against the bill. Construct 95% CI for the population
proportion.

Solution:
6250
Sample proportion (p/
O D 8500 D 0.74.
Since npO D 8500  0:74 > 5 and npO D 8500  0:36 > 5, then we can use normal distribution
table.
From the normal table, z.1 0:05 / D 1:96.
2

r
0:74.0:26/
CI D 0:74 ˙ 1:96 
r 8500 r
0:74 .0:26/ 0:74.0:26/
CI D 0:74 1:96   p  0:74 C 1:96 
8500 8500
CI D 0:7307  p  0:7493:

Hence, the proportion of respondents that supported the National Grazing Reserve bill in Nige-
ria lies within 0:7307 and 0:7493.

R Codes for the Computation of Confidence Interval for Proportion in Example 9.3
sample.prop <- 0.74
n <- 8500
std.error <- qnorm(0.975)*sqrt(sample.prop*(1-sample.prop)/n)
lower.limit <- sample.prop-std.error
upper.limit <- sample.prop+std.error
136 9. CONFIDENCE INTERVALS FOR SINGLE POPULATION MEAN AND PROPORTION
conf.interval<-c(lower.limit, upper.limit)
conf.interval
[1] 0.7306752 0.7493248

The result shows that between 73% and 75% of the respondents supported the National Grazing
Reserve bill in Nigeria.

Example 9.4
The manager of a commercial bank took a random sample of 120 customers’ account numbers
and found that 15 customers have not had bank verification number (BVN). Compute 90% CI
for the proportion of all the bank customers that are yet to complete the BVN process.

Solution:
15
pO D D 0:125
120 r
0:125.0:875/
CI D 0:125 ˙ 1:645 
r 120 r
0:125 .0:875/ 0:125.0:875/
CI D 0:125 1:645   p  0:125 C 1:645 
120 120
CI D 0:075  p  0:1747:

The percentage of all the bank customers that are yet to complete the BVN process is between
7.5% and 17.5% of the bank customers.

R Codes for the Computation of Confidence Interval for Proportion in Example 9.4
sample.prop <- 0.125
sample.size <- 120
std.error <- qnorm(0.95)
*sqrt(sample.prop*(1-sample.prop)/sample.size)
lower.limit <- sample.prop-std.error
upper.limit <- sample.prop+std.error
conf.interval<-c(lower.limit, upper.limit)
conf.interval
[1] 0.07534126 0.17465874

9.4 CALCULATING THE SAMPLE SIZE

Suppose we have given confidence level .1 ˛/ as well as the margin of error (e ), and we are
asked to calculate the minimum number of sample size. This sample size is computed using the
 9.4. CALCULATING THE SAMPLE SIZE 137
general formula:
 2
z1 ˛
n 2
s2; (9.7)
e
where s is the standard deviation of the sample.

Example 9.5
A researcher claimed that the standard deviation for the monthly utility bill for an individual
household is $50. He wants to estimate the mean of the utility bill in the present month using
95% of confidence level with the margin of error of 12 . How large a sample is required?

Solution:  2
z 2 1:96
0:975 2
n .50/ D .50/2 D 66:69  67 households:
12 12
Therefore, the minimum sample size required is 67 households.

R Codes to Calculate Sample Size Given a Standard Deviation and Margin of Error
pop.std<-50
margin.err<-12
z.normal<-qnorm(0.975)
sample.size<- (pop.std * z.normal/margin.err)**2
sample.size
[1] 66.69199
round(sample.size, digits = 0)
[1] 67

However, we can calculate the sample size for proportion under two conditions.

(i) When pO is known, the sample size is:

 2
z1 ˛
2
n p.1
O p/:
O (9.8)
e
p
Note that p.1
O O is the same as the standard deviation of a proportion .s/.
p/

(ii) When pO is unknown, the sample size is computed as:

 2
z1 ˛
2
n 0:25: (9.9)
e

This assumes that pO D 0:5.

138 9. CONFIDENCE INTERVALS FOR SINGLE POPULATION MEAN AND PROPORTION
Example 9.6
In a survey, 60% of the respondents supported a paternal leave for male workers. Construct a 95%
CI for the population proportion of workers who supported a paternal leave for male workers.
The accuracy of your estimate must fall within 2.5% of the true population proportion. Find the
minimum sample size to achieve this result.

Solution:
60
pO D 100 D 0:6 and qO D 0:4.
To verify the sampling distribution of pO to be approximated by the normal distribution, we have
npO D 100  0:6 > 5 and nqO D 100  0:4 > 5:
 
z1 ˛2 2
n p.1
O p/:
O
e
 
1:96 2
n .0:6/.0:4/
0:025
 
1:96 2
n .0:6/ .0:4/ D 1475:17:
0:025
The minimum sample size should be at least 1,475 respondents.

R Codes to Calculate Sample Size Given a Proportion and Margin of Error

p.estimate<-0.6
q.estimate<-0.4
margin.err<-0.025
alpha<-0.05
qtile<-1-(alpha/2)
z.normal<-qnorm(0.975)
sample.size<- ((z.normal/margin.err)**2)* p.estimate* q.estimate
sample.size
[1] 1475.12
The survey will required a minimum of 1,475 respondents to achieve a margin of error of 2.5%
at 95% CI.

9.5 FACTORS THAT DETERMINE MARGIN OF ERROR

The margin of error is defined as the product of critical value and standard error of estimate.
For instance, a large sample size with unknown population standard deviation has a margin of
error to be z1 ˛=2  psn where z1 ˛=2 is the critical region from a normal table, s is the sample
standard deviation, and n is the sample size. This margin of error is determined by the sample
size, the standard deviation, and the confidence level (1-˛ ).
 9.6. EXERCISES 139
(a) When sample size increases, margin of error decreases and vice versa.
(b) The higher the standard deviation, the greater the uncertainty, thus, margin of error in-
creases.
(c) The higher the confidence level, the greater the margin of error. For instance, at 95%
confidence level, z1 ˛=2 D 1:96 while at 99% confidence level z1 ˛=2 D 2:576. Hence, the
margin of error increases as the confidence level increases.

9.6 EXERCISES
9.1. The scores of students Business Statistics course are normally distributed. If a random
sample of 40 students are selected at random with mean 72 and standard deviation of 14,
compute the 95% CI for the population mean.
9.2. Research department of a telecommunication company wants to know the customers’
usage (in hours) of a new service rendered. Assuming that the usage of the service is
normally distributed, a random sample 3,000 customers under the new service is taken.
They found that the mean usage is 7 h and standard deviation of 1 h 30 min. Construct
a 99% CI for the mean usage of the service.
9.3. During the recession period, the price of a bottle of 60 cl Coke rose. Due to variability
in the price of Coke, retailers sold at diffferent prices. A random sample of 100 retailers
were sampled with mean 150.20 naira and standard deviation of 23.5. Calculate the
95% CI.
9.4. In the process of manufacturing bulbs, the probability that a bulb will be defective is
0.09. A random sample of 200 bulbs is selected, compute 95% confidence limit for the
defective bulbs?
9.5. In order to know the winner of the next presidential election in a country, a survey poll
was conducted to allow the citizens to express their opinions about the contestants. If the
95% CI is not greater than 0.09, what number of random sample size of the respondents
should be taken with the margin of error within 0.15 if the standard deviation is 20?
9.6. A steel rolling company manufanuctures cyclindrical steel with the same length but dif-
ferent diameter (in mm). A random sample of 24 steels is checked and a mean diameter
of 12.5 mm and standard deviation of 3 mm were observed. Compute the 95% CI for
the mean diameter of the steels.
 141

CHAPTER 10

Hypothesis Testing for Single

Population Mean and
Proportion
Hypothesis testing is used to know whether there is enough statistical evidence in favor or against
a belief about a population parameter. In this chapter, we will demonstrate how to use the con-
cept of hypothesis to make a smart decision. We will also dwell on its applications in business.

10.1 NULL AND ALTERNATIVE HYPOTHESES

A hypothesis is a statement speculating upon the result of a research study which can be used
to describe the population parameter. The hypothesis that in favor of the assumption is called
the null hypothesis; this is the hypothesis under investigation we are trying to disprove. It is
denoted by H0 . However, the hypothesis that we are left with should in case the null hypothesis
fails is called alternative hypothesis. An alternative hypothesis is denoted by H1 . For example, if
government wants to know if unemployment rate in the country is different from 5% aclaimed
by the National Bureau of Statistics. The null hypothesis for this scenario is H0 W  D 5% against
the alternative hypothesis H1 W  ¤ 5%.

10.2 TYPE I AND TYPE II ERRORS

A type I error occurs when rejecting the null hypothesis when null hypothesis is true. This error
is also known as a false positive. The probability of rejecting a null hypothesis when it is true is
denoted by ˛ , i.e., P (rejecting H0 jH0 is true) D ˛ . Type 1 error is sometimes called a producer’s
risk or a false alarm. This error is usually set by the researcher, the lower the ˛ value, the lower of
chance of committing type I error. The probability value (p -value) is often set to be 0.05, except
in biomedical research where the p -value is set to be 0.01 because they deal with human life.
The probability of committing a type I error is known as the test’s level of significance.
A type II error occurs when accepting the alternative hypothesis (fails to reject the null
hypothesis) when the alternative hypothesis is false. The probability of accepting an alternative
hypothesis when it is actually false is denoted by ˇ , i.e., P (accepting H1 jH1 is false) D ˇ . The
type II error is also known as consumer’s risk or misdetection. The error is not predetermined by
a researcher; rather, it is derived from the estimation of the distribution based on an alternative
142 10. HYPOTHESIS TESTING FOR SINGLE POPULATION MEAN AND PROPORTION
hypothesis and is usually unknown. The value of 1 ˇ is known as power of a test. The power of
a test gives the probability of rejecting the null hypothesis when the null hypothesis is false. On
the contrary, the level of significance is gives probability of rejecting the null hypothesis when
the null hypothesis is true. The values of both ˛ and ˇ are dependent, as one increases the other
decreases. However, the increase in the sample size, n, causes both to decrease due to reduction
in sampling error. Table 10.1 shows the types of errors and theirs respective probabilities.

Table 10.1: Statistical errors

H0 Rejected Fail to Reject H0

H0 false Correct decision Type II error
P = 1-β P=β
(power of a test)
H0 true Type I error Correct decision
P=α P = 1-α
(level of significance)

10.3 ACCEPTANCE AND REJECTION REGIONS

In the hypothesis testing procedure, the sampling region is divided into two—acceptance and
rejection (critical) regions. The acceptance region contains a set of values for which a test statistic
falls within the specified range. If the test statistic falls within the acceptance region, then the null
hypothesis is accepted. On the other hand, the rejection region is the area of the curve where
the test statistic falls outside the specified range. Thus, if the sample statistic falls within the
rejection region, then the alternative hypothesis is accepted. Figure 10.1 shows the acceptance
and critical regions for a two-tail normal test.

10.4 HYPOTHESIS TESTING PROCEDURES

In carrying out the testing of a hypothesis, the following are various steps to take.
(a) State the Hypotheses:
A hypothesis is stated based on the argument of interest. The hypothesis can be either a
one-tail test or a two-tail test. A one-tail test is a test of a hypothesis where the region
of rejection is on one side of the sampling distribution. Suppose that the null hypothesis
stated that the population mean is greater than zero and the alternative hypothesis stated
that the population mean is less than or equal to zero (i.e., H0 W  < 0 against H1 W   0).
This implies that the rejection region would consist of a range of values located on the left
side of sampling distribution; a set of values less than or equal to zero.
 10.4. HYPOTHESIS TESTING PROCEDURES 143

Acceptance
Region

Critical Region Critical Region

-Zα/2 Z0 Zα/2

Figure 10.1: Two-tail normal test.

Alternatively, a two-tail test is a test hypothesis where the rejection region is on both
sides of the sampling distribution. Assume that the null hypothesis stated that H0 W  D 0
against the alternative hypothesis H1 W  ¤ 0. The non-directional sign would take the
values from both sides of the sampling distribution; thus, the set of values on the right
side of zero and on the left size of zero are the rejection regions. Hypotheses testing can
be of the form:
H0 W  D 0 vs. H1 W  ¤ 0 (two-tail test)
H0 W  D 0 vs. H1 W  < 0 (one-tail test)
H0 W  D 0 vs. H1 W  > 0 (one-tail test)
H0 W  > 0 vs. H1 W   0 (one-tail test)
H0 W   0 vs. H1 W  < 0 (one-tail test)
H0 W  < 0 vs. H1 W   0 (one-tail test)
H0 W  < 0 vs. H1 W   0 (one-tail test)

(b) Specify the Level of Significance:

Prior to the experiment, the researcher would have chosen the level of significance. In most
cases, the level of significance is set to 10%, 5%, and 1% depending on the stringency of the
research investigation. Suppose the level of significance is 5%; this indicates that there is
likelihood that out of 100 cases, only 5 cases might result to rejecting the null hypothesis.
144 10. HYPOTHESIS TESTING FOR SINGLE POPULATION MEAN AND PROPORTION
(c) Compute the Value of the Test Statistic:
The appropriate test statistic is compared with the critical value. The most commonly used
test statistics are Student t -test, normal (z -test), chi-square, and F-test. The type of statis-
tic to be used depends on the properties of the data. The typical example for a test statistic
is:
p
n .x /
tD  tn 1;˛=2
s
or
p
n .x /
zD  z; ˛=2 :
s

(d) Determine the Critical Value and Decision Rule:

The critical region is the cut off point area of the rejection region. As long as the absolute
value of the test statistic is greater than the critical value, the null hypothesis is rejected.
The decision rule states that we should reject the null hypothesis if the absolute value of
test statistic is higher than the critical value.
(e) Draw a Conclusion:
After a decision has been made whether to accept or reject the null hypothesis, the next
process is to draw a conclusion about the parameter under investigation.

Example 10.1
In a pharmaceutical company, the operations manager claimed that the mean of the drugs pro-
duced by the company is 100 mg. If a random sample of 60 drugs is chosen with mean of 98 mg
and standard deviation of 14 mg, test the hypothesis to justify the operations manager’s claim,
use ˛ D 0:05.

Solution:
(a) H0 W  D 100 vs. H1 W  ¤ 100.
(b) ˛ D 0:05.
p
60.98 100/
(c) Test statistic: z D 14
D 1:1066 (we use test statistic z since the sample size is
greater than 30).
(d) z.0:975/ D 1:96 (from the normal distribution table).
The stated hypothesis is a two-tailed test, therefore we use z.1 ˛ / . However, if the stated
2
hypothesis is a one-tailed test and sample size is is equal to or greater than 30, then we
will use z.1 ˛/ .
 10.4. HYPOTHESIS TESTING PROCEDURES 145
Decision rule: reject null hypothesis if j 1:1066j > 1:96, since test statistics is not greater
than critical value, therefore we do not reject H0 .

(e) Conclusion: the data support the claim of the operations manager than the mean of the
drugs produced is 100 mg.

The scripts below show how the question in Example 10.1 can be solved in R. The z -
statistic and critical value are computed in R. This serves as a basis of comparison.

R Codes
# state the given parameters
xbar = 98 # sample mean
mu0 = 100 # hypothesized value
sigma = 14 # sample standard deviation
n = 60 # sample size
# use z-test since sample size is greater than 30
z = (xbar-mu0)/(sigma/sqrt(n))
z # test statistic
[1] -1.106567

This z -statistic is -1.1066.

# compute the critical value at 0.05 significance level

alpha = 0.05
z_alpha = qnorm(1-alpha/2)
z_alpha # critical value
[1] 1.959964

The absolute value of computed z -statistic is 1.1066, which is less than the critical value of 1.96.
Therefore, we do not reject null hypothesis and we conclude that the operations manager is right
in his claim with the given data.

Example 10.2
A stockbroker claimed that weekly average return on a stock is normal with an average of return
of 0.5%. He took the 20 previous weeks return and found that the weekly average returns was
0.48 with standard deviation 0.08. At 5% level of significance, does his claim about the relevant?
If the level of significance is reduce to 1%, compare the result.

Solution:

(a) Hypothesis: H0 W  D 0:5% vs. H1 W  ¤ 0:5%.

146 10. HYPOTHESIS TESTING FOR SINGLE POPULATION MEAN AND PROPORTION
(b) ˛ D 0:05.
p
(c) Test statistic: t D 20.0:48
0:08
0:50/
D 1:118 (we use t -test because the sample size is less
than 30). In this case, our sample size is 20.
(d) tn 1;.1 ˛ / D t19;.0:975/ D 2:093 (critical value for two-tailed test is obtained from student-
2
t table).
Decision rule: reject null hypothesis if j 1:118j > 2:093, since test statistics is not greater
than critical value, therefore we do not reject H0 .
(e) Conclusion: the stockbroker is correct on his claim based on the data. Therefore, the av-
erage weekly returns of the stock is 0.5%.
At ˛ D 0:01, t D 1:118, tn 1;.1 ˛
/ D t19;.0:995/ D 2:861 (obtained from the Student-t
2
table).

Decision rule: Reject null hypothesis if j 1:118j > 2:861. Since the absolute of test statistics is
less than critical value, therefore we do not reject H0 .
Conclusion: We accept the null hypothesis based on the data, and conclude that the average
weekly returns is 0.5%.
Comparison: The stockbroker is claim that the average weekly return is 0.5% is right at both 1%
and 5% level of significance.
Let’s demonstrate this Example 10.2 with R codes.

R Codes
# Given parameters
x.bar = 0.48
mu0 = 0.50
sigma = 0.08
sample.size = 20

# compute the t-statistics

t = (x.bar-mu0)/(sigma/sqrt(sample.size ))
t
[1] -1.118034

# compute the critical value at 0.05 significance level

alpha1 = 0.05
t.alpha1 = qt(1-alpha1/2, sample.size-1 )
 10.4. HYPOTHESIS TESTING PROCEDURES 147
t.alpha1
[1] 2.093024

# At 1% level of significance
alpha2 = 0.01
t.alpha2 = qt(1-alpha2/2, sample.size-1 )
t.alpha2
[1] 2.860935
These results are the same as the outcomes we obtained in Example 10.2.
Example 10.3
The National Bureau of Statistics (NBS) claimed that less than 10% of the graduate youths are
unemployed while the opposition party argued that the percentage of graduate youths is more
than NBS claims. To ascertain the validity of the claim, a random sample of 10,000 graduate
youths are selected in which 1,250 graduate youths are unemployed. Test the hypothesis that
less than 10% of the graduate youths are unemployed. (Hint: use ˛ D 0:05.)

Solution:
(a) Hypothesis: H0 W P < 10% vs. H1 W P  10%.
(b) ˛ D 0:05.
(c) pO D 0:125.
Test statistic: z D q pO p0 .
p0 .1 p0 /
n

Here, we used z -statistics because the sample size is 10,000 which is large enough, that is,
it is greater than 30.
0:125 0:1
zDq D 8:3:
0:1.1 0:1/
10000

(d) Critical value: z.1 ˛/ D z.0:95/ D 1:64 (Our hypothesis is one-tailed, thus z.1 ˛/ is to be
used).
Decision rule: Reject null hypothesis if z > z.0:95/ , since 8:3 > 1:64, then we reject null
hypothesis.
(e) Conclusion: There is no evidence to support the claim of NBS that the percentage of
unemployed graduate youths is less than 10% based on the data.
Note: There is a possibility that another set of dataset might justified the claim of the NBS.
However, based on the information given in this particular question, no justification for
the claim that the percentage of unemployed graduate youths is less than 10%.
148 10. HYPOTHESIS TESTING FOR SINGLE POPULATION MEAN AND PROPORTION
The R codes below describe how the z -statistics for the proportion .p/ and critical value
can be obtained for proper comparison.

R Codes
# compute the z-statistic
p.hat = 0.125
p0 = 0.10
n = 10000
z = (p.hat-p0)/ sqrt (p0*(1-p0)/n)
z
[1] 8.333333

# compute the critical value at 0.05 significance level

alpha = 0.05
z.alpha = qnorm(1-alpha)
z.alpha
[1] 1.644854
The z -statistic and critical value under z are 8.333333 and 1.644854, respectively.
Example 10.4
There is a popular saying that girls perform better than the boys. To justify that, a random
sample of 1,500 students which consists of 950 boys was selected to participate in an entrance
examination into the secondary schools. Out of the boys that participated, only 480 passed and
only 455 girls passed. Test the hypothesis that girls perform better than boys at a 5% level of
significance.

Solution:
(a) Let pO be the proportion of girls that passed the examination and p0 be the proportion of
girls that participated in the examination.
Hypothesis: H0 W p > 0:5% vs. H1 W p  0:5.
(b) ˛ D 0:05.
(c) pO D 0:83 and p0 D 0:5.
Test statistic: z D q pO p0
p0 .1 p0 /
n

The z -statistic is used because the sample size is greater than 30:
0:83 0:50
zDq D 25:56:
0:5.1 0:5/
1500
 10.5. EXERCISES 149
(d) Critical value: z.1 ˛/ D z.0:95/ D 1:64 (from z -table with one tailed test).
Decision rule: Reject null hypothesis if z > z.0:95/ , since 25:56 > 1:64, then we reject null
hypothesis.

(e) Conclusion: The data supported the general claim that girls perform better than the boys.

R Codes for Computing z -test for a Proportion and its Critical Value
# compute the z-test for the proportion
p.cap = 0.83
p0 = 0.5
n = 1500
z = (p.cap-p0)/ sqrt (p0*(1-p0)/n)
z
[1] 25.56169

# compute the critical value at 0.05 significance level

alpha = 0.05
z.alpha = qnorm(1-alpha)
z.alpha
[1] 1.644854

Since both z -test for the proportion and critical value are the same as in Example 10.4. Therefore,
we are arriving at the same conclusion.

10.5 EXERCISES
10.1. (a) Define the following:
(i) null hypothesis and alternative hypothesis; and
(ii) type 1 and type 2 error.
(b) The resident doctor of a hospital stated that the weight of a newborn baby is 3.5 kg
and above because of the kind of foods pregnant women ate during pregnancy. A
random sample of the 20 newborn babies are selected from the records and find
that the mean is 2.95 kg and the standard deviation is 0.85 kg. Test the hypothesis
that the weight of newborn babies is greater than or equal to 3.5 kg, assuming the
weight is normal (use ˛ D 0:01).
10.2. The managing director claims that brewery company notices that the sales of the prod-
ucts decline less than 20% during Ramadan periods and summon the sales manager
to investigate his claim. The sales manager collated the sales of the products during the
150 10. HYPOTHESIS TESTING FOR SINGLE POPULATION MEAN AND PROPORTION
month of Ramadan in the past 30 years. Table 10.2 shows the distribution of percentage
decrease in the sales during Ramadan periods. What would you say about the managing
director’s claim?
Table 10.2: Sales of a brewery company during Ramadan

Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
Decrease in
27.41 9.31 20.36 26.43 22.75 14.48 10.21 12.42 20.83 18.53
sales (%)

Year 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Decrease in
19.98 25.22 12.08 26.75 35.43 19.44 19.62 24.06 25.04 10.61
sales (%)

Year 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Decrease in
31.32 24.13 16.38 20.08 21.59 13.98 19.80 23.97 29.64 21.29
sales (%)

10.3. A quality assurance manager argued that the average lifetime of light bulbs is 520 h. To
ascertain his claim, he took a random sample of 50 bulbs and the lifetime readings (in
hours) are as follows:
427.82, 425.76, 395.28, 444.67, 437.26, 442.67, 424.36, 416.63, 431.49,
401.58, 407.57, 461.93, 436.58, 423.20, 429.79, 447.74, 430.27, 434.41,
414.26, 435.79, 427.24, 401.04, 433.63, 404.31, 400.14, 437.70, 437.36,
424.59, 410.77, 448.75, 421.52, 416.88, 427.79, 425.87, 412.31, 423.88,
397.12, 430.68, 418.87, 411.51, 418.85, 405.59, 416.06, 388.01, 439.83,
419.70, 443.24, 422.75, 419.85, 420.28

Test whether the population mean is 520 h at a significance level of 5%.

10.4. Due to the economic recession in Nigeria, the chairman of National Union Transport
and Road Workers (N.U.T.R.W) wanted to increase their fares and he told to the man-
agement of the University of Lagos that the daily income of the taxi operators within
the campus is less than N2,500 per daily. The institution set of a committee to confirm
his affirmation, then the committee took a random sample of 15 taxi operators with
average income of N2,580 and standard deviation of N200. Assume that the income is
normal distributed, justify the claim of the chairman at level of significance of 5%.
10.5. (a) With the aid of illustration, explain what you understand about the acceptance and
rejection region.
 10.5. EXERCISES 151
(b) Discuss the procedure involved in the testing of hypotheses.
(c) The manager of a grocery store insisted that a customer spent an average of NGN
5,000 or more for food on a daily basis. A random sample of 18 customers are
selected from the sales records and found that the average sales is NGN 4,975 with
variance of NGN 22,500. Test whether the manager’s claim agrees with the sample
collected. Assume that the sales of groceries is normal and use 5% significance level.
10.6. A Coca-Cola production plant produces bottles containing 35 cl of soda, the new man-
ager suspected that the volume of the drink do not follow the specification ascribed on
the bottle. He decided to take a sample of the products and the following measurements
are obtained:
34.73, 36.02, 35.54, 34.72, 35.49, 35.15, 35.99, 35.35,
35.81, 34.98, 34.23, 35.45, 34.99, 34.75, 35.57, 34.40,
35.09, 34.20, 34.42, 36.09, 34.34, 35.81, 34.89

Test the H0 W  D 35 cl against H1 W  ¤ 35 cl (use ˛ D 0:05).

10.7. A rector of a polytechnic institution said that not fewer than 52% of the students that
graduated in the institution fully gained employment immediately after leaving the
school. A random sample of 1,000 graduates of the institution are taken across dif-
ferent years to examine if they actually gained employment immediately after leaving
the school. The table below shows the number of the graduates from the institution and
employment status immediately after graduation.

Employment status Employed Not employed

Number of graduated students 542 458

Test whether the rector is right in his statement and test that H0 W p  0:5 against H1 W
p < 0:5, use ˛ D 0:05.
 153

CHAPTER 11

Regression Analysis and

Correlation
This chapter will focus on how two or more variables are interrelated. We shall understand the
two concepts in statistics—regression analysis and correlational analysis—and the relationship
between the concepts. We shall discuss what to look for in the output of regression analysis
and how the output can be interpreted. We shall elucidate on how to use regression analysis for
forecasting.

11.1 CONSTRUCTION OF LINE FIT PLOTS

In showing the relationship between two variables, we can draw a line across the variables after
plotting the scatter plot and ensuring the line passes through as many points as possible. The
straight line that gives the best approximation in a given set of data is referred to as the line of
best fit. Least squares method is the most accurate method of finding the line of best fit of a
given dataset.
For example, Fig. 11.1 shows sales revenue ($’million) and amount spent on advertisement
($’million) of a production company.

11.2 TYPES OF REGRESSION ANALYSIS

There are many regression analyses which are based on different assumptions. We mention a
few, and we will dwell on the first and second mentioned below:
1. simple linear regression;
2. multiple regression;
3. ridge regression;
4. quantile regression; and
5. Bayesian regression.

11.2.1 USES OF REGRESSION ANALYSIS

Regression analysis can be used for the following.
154 11. REGRESSION ANALYSIS AND CORRELATION

Sales revenue ($’million) 115 118 120 125 126 128 131 132
Advertisement expenses ($’million) 4 7 9 14 15 17 20 21

134
Sales = Advert + 111
132

Sales Revenue ($’ million)

130
128
126
124
122
120
118
116
114
0 5 10 15 20 25
Advertisement Expenses ($’million)

Figure 11.1: Chart of sales revenue on advertisement.

1. Causal analysis—to establish the relationship between two or more variables.

2. Forecasting an effect—it is used to predict a response variable fully knowing the indepen-
dent variables.

3. Forecasting a trend—regression analysis can be used to predict trend in a dataset.

11.2.2 SIMPLE LINEAR REGRESSION

Simple linear regression is a statistical technique to show the relationship between one depen-
dent variable and one independent variable. The dependent variable is denoted by Y while the
independent variable is denoted by X . The variables X and Y are linearly related. Simple linear
regression can be used for the following: (a) description of the linear dependence of one vari-
able on another variable; (b) prediction of one variable from the values of another variable; and
(c) correction for the linear dependence of one variable on another variable.
The simple linear regression model is of the form:

Yi D ˇ0 C ˇ1 Xi C "i ; (11.1)

where ˇ0 and ˇ1 are the intercept and regression coefficient of X and " is the error term.
 11.2. TYPES OF REGRESSION ANALYSIS 155
The solution to the regression coefficients in (11.1) can be derived using Least Square method:

ei D Yi ˇ0 ˇ1 Xi : (11.2)

To find a minimum sum of squares of residuals you set the sum below equal to zero:
n
X n
X
.ei /2 D .Yi ˇ0 ˇ1 Xi /2 D 0: (11.3)
i D1 i D1

Taking partial derivative of (11.3) with respect to ˇ1 :

n n n
!
ı X 2
X X
.Yi ˇ0 ˇ1 Xi / D 2 nˇ0 C ˇ1 Xi Yi D 0: (11.4)
ıˇ0
i D1 i D1 i D1

Divide (11.4) by 2, and solve for ˇ0 .

Then,
ˇ0 D Y ˇ1 X : (11.5)
Now,
n n
ı X X

.Yi ˇ0 ˇ1 Xi /2 D 2 Xi Yi ˇ0 Xi ˇ1 Xi2 D 0: (11.6)
ıˇ1
i D1 i D1

Substitute for the value of ˇ0 into (11.6):

n
X


Xi Yi Y ˇ1 X Xi ˇ1 Xi 2 D 0
i D1
n
X

Xi Yi Xi Y ˇ1 Xi X ˇ1 Xi 2 D 0
i D1
n
X n
X



Xi Yi Xi Y ˇ1 Xi2 Xi X D 0:
i D1 i D1

Therefore,
Pn
Pn
i D1 Xi Yi Xi Y i D1 .Xi Y i / nXY cov.X; Y /
ˇ1 D Pn
D P
2
D : (11.7)
i D1 Xi2 Xi X n
i D1 X 2
nX var.X /
i

11.2.3 ASSUMPTIONS OF SIMPLE LINEAR REGRESSION

1. There is a linear relationship between variable Y and variable X .

2. The variable X is deterministic or non-stochastic.

156 11. REGRESSION ANALYSIS AND CORRELATION
3. The error terms are statistically independent.
4. There is no correlation between X and "i , that is, Cov .X; "i / D 0.
5. The error terms are distributed normally with mean zero and constant variance, i.e., " 
N.0;  2 /.

6. There is no correlation between the error terms, i.e., no serial auto-correlation in the data.
7. The number of sample observations must greater than the number of parameters to be
estimated.
8. For each of value of X, the distribution of residuals has equal variance, i.e., homoscedacity.

Example 11.1
In a business statistics class, the weight and height of 30 students were measured, as shown in
Table 11.1.
(i) Find the regression of the weight on the height of the students.
(ii) Use your answer in (i) to estimate the value of student’s weight when the height is 1.80?

Table 11.1: Weight and height of students

Student 1 2 3 4 5 6 7 8 9
Height (m) 1.43 1.10 2.24 1.36 2.26 1.25 1.74 1.55 1.51
Weight (kg) 92.18 77.76 65.44 114.19 82.81 106.66 94.44 75.32 67.35
Student 10 11 12 13 14 15 16 17 18
Height (m) 1.82 1.57 1.59 2.19 1.54 2.06 1.86 1.76 1.51
Weight (kg) 101.55 76.37 91.66 75.85 88.82 83.02 74.66 97.57 104.56
Student 19 20 21 22 23 24 25 26 27
Height (m) 2.39 1.83 2.02 1.99 1.40 1.54 1.60 1.88 1.52
Weight (kg) 113.36 64.71 103.79 70.02 78.35 80.70 90.54 91.55 82.57
Student 28 29 30
Height (m) 1.41 1.38 1.18
Weight (kg) 82.49 87.98 67.54
 11.2. TYPES OF REGRESSION ANALYSIS 157
Solution:
From the data above, we obtained the following results:
X X X
XY D 4347:56; Y D 2583:81; X D 50:48;

n D 30; X D 1:68; Y D 86:13;

X
X 2 D 88:18:
Substitute for these values in Equation (11.7) above:
Pn
.Xi Yi / nX Y 4347:56 30 .1:68/ .86:13/
ˇ1 D Pi D1
2
D 2
D 0:042:
n 2 88:18 30.1:68/
i D1 X i nX

The slope of the regression is 0:042.

Then, substitute for Y , X , and ˇ1 in Equation (11.5), to get the constant term:

ˇ0 D Y ˇ1 X D 86:13 . 0:042/ .1:68/ D 86:20:

(i) The regression model is Weight D 86:20 0:042  Height.

(ii) Weight D 86:20 0:042  1:80 D 78:64.

To Estimate the Regression Estimates in R

Assign values for height and weight in R as follows.
height<-c(1.43, 1.10, 2.24, 1.36, 2.26, 1.25, 1.74, 1.55, 1.51,
1.82, 1.57, 1.59, 2.19, 1.54, 2.06, 1.86, 1.76, 1.51,
2.39, 1.83, 2.02, 1.99, 1.40, 1.54, 1.60, 1.88, 1.52,
1.41, 1.38, 1.18)
weight<-c(92.18, 77.76, 65.44, 114.19, 82.81, 106.66, 94.44, 75.32,
67.35, 101.55, 76.37, 91.66, 75.85, 88.82, 83.02, 74.66,
97.57, 104.56, 113.36, 64.71, 103.79, 70.02, 78.35, 80.70,
90.54, 91.55, 82.57, 82.49, 87.98, 67.54)
Fit the model by regressing weight on height of the students and the results follow.
model.fit<-lm (weight~height)
summary(model.fit)

Call:
lm(formula = weight ~ height)

Residuals:
158 11. REGRESSION ANALYSIS AND CORRELATION
Min 1Q Median 3Q Max
-21.411 -10.132 -3.192 7.747 28.049

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 86.19791 13.55401 6.360 6.99e-07 ***
height -0.04214 7.90579 -0.005 0.996
---
Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1

Residual standard error: 14.23 on 28 degrees of freedom

Multiple R-squared: 1.015e-06, Adjusted R-squared: -0.03571
F-statistic: 2.841e-05 on 1 and 28 DF, p-value: 0.9958

In the results above, the statistics for residuals, the coefficient estimates (with standard errors
and the associated p-values), and all other statistics (Multiple R-squared, Adjusted R-squared,
F-statistics, etc.) are shown in the output.

Example 11.2
Table 11.2 shows the log of gross domestic products (GDP) and the log of government spending
in Nigeria between 1981–2015. Regress the log of GDP on the log of government spending and
interpret your result.

Solution:
From the table above, we obtained the following:
X X X
XY D 20140:45; Y D 907:92; X D 775:62;

n D 35; X D 22:16; Y D 25:94;

X
X 2 D 17243:21:
Substitute for these values to compute regression coefficients (ˇ0 and ˇ1 )
Pn
.Xi Yi / nXY 20140:45 35 .22:16/ .25:94/
ˇ1 D Pi D1
2
D D 0:382
n
i D1 Xi
2
nX 17243:21 35.22:16/2
ˇ0 D Y ˇ1 X D 25:94 .0:382/ .22:16/ D 17:475:

Thus, the regression model is LNGDP D 17:475 C 0:382  LNGEXP.

Interpretation: This implies that 1% increase in the government spending would lead to
0.4% increase in gross domestic products.
 11.2. TYPES OF REGRESSION ANALYSIS 159
Table 11.2: Gross domestic product

Year 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
LNGDP (Y) 25.545 25.535 25.483 25.463 25.543 25.451 25.337 25.410 25.473 25.593
LNGEXP (X) 21.082 21.105 21.128 21.150 21.171 21.192 21.213 21.233 21.253 21.273

Year 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
LNGDP (Y) 25.587 25.591 25.612 25.621 25.618 25.666 25.694 25.721 25.725 25.777
LNGEXP (X) 21.283 21.312 21.340 21.354 21.354 21.382 21.399 21.416 21.433 21.449

Year 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
LNGDP (Y) 25.820 25.858 25.956 26.247 26.281 26.360 26.426 26.486 26.554 26.629
LNGEXP (X) 21.320 21.377 21.103 22.998 23.098 23.404 23.854 24.069 24.076 24.188

Year 2011 2012 2013 2014 2015

LNGDP (Y) 26.677 26.719 26.771 26.832 26.859
LNGEXP (X) 24.233 24.213 24.105 24.032 24.028

R Codes for Regressing the Log of Gross Domestic Products on the Log of Government
Spending
Assign the values for the log of GDP and the log of government spending (LNGEXP).
LNGDP<-c(25.545, 25.535, 25.483, 25.463, 25.543, 25.451, 25.337, 25.410,
25.473, 25.593, 25.587, 25.591, 25.612, 25.621, 25.618, 25.666,
25.694, 25.721, 25.725, 25.777, 25.820, 25.858, 25.956, 26.247,
26.281, 26.360, 26.426, 26.486, 26.554, 26.629, 26.677, 26.719,
26.771, 26.832, 26.859)
LNGEXP<-c(21.082, 21.105, 21.128, 21.150, 21.171, 21.192, 21.213, 21.233,
21.253, 21.273, 21.283, 21.312, 21.340, 21.354, 21.354, 21.382,
21.399, 21.416, 21.433, 21.449, 21.320, 21.377, 21.103, 22.998,
23.098, 23.404, 23.854, 24.069, 24.076, 24.188, 24.233, 24.213,
24.105, 24.032, 24.028)
data<-data.frame (LNGDP, LNGEXP)
model<-lm(LNGDP~LNGEXP, data)
summary(model)

Call:
lm(formula = LNGDP~LNGEXP)
160 11. REGRESSION ANALYSIS AND CORRELATION

Residuals:
Min 1Q Median 3Q Max
-0.25038 -0.06998 -0.01893 0.04646 0.40962

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 17.68058 0.39751 44.48 <2e-16 ***
LNGEXP 0.37273 0.01791 20.81 <2e-16 ***
---
Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1

Residual standard error: 0.1328 on 33 degrees of freedom

Multiple R-squared: 0.9292, Adjusted R-squared: 0.9271
F-statistic: 433.2 on 1 and 33 DF, p-value: < 2.2e-16

From the output above, the regression coefficients are 17.68 and 0.37 for the intercept and slope,
respectively.

11.3 MULTIPLE LINEAR REGRESSION

In simple linear regression, we considered two variables where one is the response and the other
one is explanatory variable. In the case of multiple linear regression, it is an extension of simple
linear regression whereby we have two or more explanatory variables that account for the vari-
ation in a dependent variable. Each of the explanatory variables Xi is associated with a value of
the response variable Y . The multiple linear regression model is of the form:

Yi D ˇ0 C ˇ1 Xi1 C ˇ2 Xi 2 C ˇ3 Xi3 C


C ˇk Xik C "ik ; for i D 1; 2; : : : ; n; (11.8)

where ˇ0 is a constant term, ˇ1 ; ˇ2 ; : : : ; ˇk are regression coefficients, "i k is error term, and
"ik  N.0;  2 /.
The estimates of the regression coefficients (ˇ0 ; ˇ1 ; ˇ2 ; : : : ; ˇk ) are the values that mini-
mize the sum of squared errors for the residuals.
For two independents variables,

Yi D ˇ0 C ˇ1 Xi1 C ˇ2 Xi2 C "ik : (11.9)

 11.3. MULTIPLE LINEAR REGRESSION 161
The regression coefficients can be estimated as:

P
P P P
x22 . x1 y/ . x1 x2 / . x2 y/
ˇ1 D P 2
P 2
P (11.10)
X1 X2 . X1 X2 /2
P 2
P P P
x1 . x2 y/ . x1 x2 / . x1 y/
ˇ2 D P 2
P 2
P (11.11)
x1 x2 . x1 x2 /2
ˇ0 D Y ˇ1 X 1 ˇ2 X 2 ; (11.12)

where
P P
X X . X1 / . Y /
x1 y D X1 Y (11.13)
P NP
X X . X2 / . Y /
x2 y D X2 Y (11.14)
P N P
X X . X1 / . X2 /
x1 x 2 D X1 X 2 : (11.15)
N

Interpretation: ˇ0 is the constant term or intercept at Y . ˇ1 is the change in Y for each

1 unit change in X1 while X2 is held constant, and ˇ2 is the change in Y for each 1 unit change
in X2 holding X1 constant.

Example 11.3
Table 11.3 shows the level of education .X 1 /, year of experience .X 2 /, and the log of monthly
compesation of a company (Y ). The level of education X1 D 1 for primary education, X1 D 2 for
secondary education, X1 D 3 for polytechnic graduate, X1 D 4 for university Bachelor’s degree,
and X1 D 5 for university Master’s degree holder. Regress log of monthly compensation (Y ) on
level of education .X 1 / and the year of experience in the company .X 2 /.

Solution:
We obtained the following values from the data above:

X X X X
X22 D 921; X12 D 387; X1 Y D 759:61; X2 Y D 1165:55;
X X X X
X1 X2 D 499; n D 35; X1 D 107; X2 D 167; Y D 242:76:
162 11. REGRESSION ANALYSIS AND CORRELATION
Table 11.3: Data for Example 11.3

Level of education (X1) Experience (X 2) Log of Compensation (Y)

1 8 6.43
2 4 6.73
2 6 6.76
4 2 7.00
3 7 7.10
4 5 7.34
5 5 7.53
3 3 6.75
4 2 6.83
2 6 7.00
2 3 6.51
3 5 6.82
3 8 6.94
1 5 6.23
3 5 6.82
5 5 7.53
5 3 7.31
5 4 7.43
4 5 7.34
4 5 7.34
2 3 6.51
4 9 7.61
3 2 6.71
5 4 7.43
3 4 6.79
1 7 6.40
4 5 7.34
2 3 6.51
2 9 7.17
2 5 6.75
5 6 7.55
1 2 5.92
1 5 6.23
3 3 6.75
4 4 7.31
 11.3. MULTIPLE LINEAR REGRESSION 163
Compute the following and substitute the values for obtained the regression coefficients:
X .107/.242:76/
x1 y D 759:61 D 17:458
35
X .167/.242:76/
x2 y D 1165:55 D 7:238
35
X .107/.167/
x1 x2 D 499 D 11:543
35
X .167/.167/
x22 D 921 D 124:171
35
X .107/.107/
x12 D 387 D 59:886
35
P
P P P
x22 . x1 y/ . x1 x2 / . x2 y/
ˇ1 D P 2
P 2
P
x1 x2 . x1 x2 /2
.124:171/ .17:458/ . 11:543/ .7:238/
D D 0:308
.59:886/ .124:171/ . 11:543/2
.59:886/ .7:238/ . 11:543/ .17:458/
ˇ2 D D 0:084
.59:886/ .124:171/ . 11:543/2
ˇ0 D 6:94 0:308 .3:06/ 0:084 .4:77/ D 5:597:

The regression coefficients are 5.597, 0.308, and 0.084 for ˇ0 , ˇ1 , and ˇ2 , respectively.
The model is Yi D 5:577 C 0:31Xi1 C 0:087Xi 2 .
Interpretation: This implies that 1 unit increase in the level of education would lead to a
31% increase in compensation holding years of experience constant. Also, for every unit increase
in years of experience would lead to a 9% increase in compensation holding level of education
constant.

R Codes for the Solution of Example 11.3

Input the values of level of education (edu), year of employment (empl), and log of compensation
(comp).
edu<-c(1, 2, 2, 4, 3, 4, 5, 3, 4, 2, 2, 3, 3, 1, 3, 5, 5, 5, 4, 4,
2, 4, 3, 5, 3, 1, 4, 2, 2, 2, 5, 1, 1, 3, 4)

empl<-c(8, 4, 6, 2, 7, 5, 5, 3, 2, 6, 3, 5, 8, 5, 5, 5, 3, 4, 5, 5,
3, 9, 2, 4, 4, 7, 5, 3, 9, 5, 6, 2, 5, 3, 4)

comp<-c(6.43, 6.73, 6.76, 7, 7.10, 7.34, 7.53, 6.75, 6.83, 7, 6.51,

6.82, 6.94, 6.23, 6.82, 7.53, 7.31, 7.43, 7.34, 7.34, 6.51,
164 11. REGRESSION ANALYSIS AND CORRELATION
7.61, 6.71, 7.43, 6.79, 6.40, 7.34, 6.51, 7.17, 6.75, 7.55,
5.92, 6.23, 6.75, 7.31)
Combine the variables (edu, empl, and comp) as a dataframe and then regress comp on edu, and
empl.
mydata<-data.frame (edu, empl, comp)
multiple.reg<-lm(comp~edu+empl, mydata)
Obtain the regression coefficient and other statistics as follows:
summary(multiple.reg)

Call:
lm(formula = comp~edu + empl, data = mydata)

Residuals:
Min 1Q Median 3Q Max
-0.25870 -0.09077 -0.01283 0.07499 0.28368

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.57723 0.07656 72.844 < 2e-16 ***
edu 0.30803 0.01540 20.002 < 2e-16 ***
empl 0.08717 0.01069 8.151 2.61e-09 ***
---
Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1

Residual standard error: 0.1181 on 32 degrees of freedom

Multiple R-squared: 0.9308, Adjusted R-squared: 0.9265
F-statistic: 215.3 on 2 and 32 DF, p-value: < 2.2e-16
The regression coefficients are the same with Example 11.3 above and it is interpreted the same
manner.

11.3.1 SIGNIFICANCE TESTING OF EACH VARIABLE

In this section, we are testing if the independent variables in the model is useful in the model,
i.e., if the independent variable can usually help us predict the dependent variable (y ). Therefore,
 11.3. MULTIPLE LINEAR REGRESSION 165
to determine whether variable X1 is making a useful contribution in the model, we have to test
its significance by setting hypotheses as shown below.
Hypothesis: H0 W ˇ1 D 0 vs. H1 W ˇ1 ¤ 0.
ˇ1
Test statistic: t D se.ˇ 1/
 t1 ˛2 ;n 2 .
Decision rule: Reject H0 if jtj  t1 ˛2 ;n 2 .

Example 11.4
In Example 11.3, test whether the null hypothesis that ˇ1 is significantly different from zero
i.e., H0 W ˇ1 D 0 vs. H1 W ˇ1 ¤ 0.

Solution:
Hypothesis: H0 W ˇ1 D 0 vs. H1 W ˇ1 ¤ 0.
Test statistic: t D 0:30803
0:01540
D 20:002.
Critical value: t0:975;33 D 2:042.
Decision rule: Reject H0 if jtj  t0:975;33 , since 20:002  2:042, then we reject H0 .
Conclusion: The coefficient of level of education X1 is significantly different from zero.
Note: This result gives the same conclusion with the model in Example 11.3.

11.3.2 INTERPRETATION OF REGRESSION COEFFICIENTS AND

OTHER OUTPUT
(a) Regression coefficients: The magnitude of the coefficient of each independent variable
gives the size of the effect that the independent variable has on dependent variable and
the sign ( =C) on the regression coefficient tells the direction of the effect. In general, the
coefficient gives how much the dependent variable would change when the independent
variable change by 1 unit, keeping all other independent variables constant.
For example, if we fit a model, Y D 2:5 C 0:5X1 0:8X2 , this can be interpreted as: the
Y -intercept can be intepreted as the predicted value for Y when both X1 and X2 are zero.
Therefore, you would expect 2.5 unit when X1 D 0 and X2 D 0. For every unit increse in
Y would lead to 0.5 unit increase in X1 while X2 is held constant, also for every unit in Y
would lead to 0.8 unit decrease in X2 holding X1 constant.

(b) t -value: This is the ratio of the coefficient and the standard error of the coefficient. The
rule of thumb is that the absolute value of t must be 2 or more to show the significance
of the coefficient. T -value is used to determine the p -value corresponding to Student t -
distribution.

(c) P -value: It indicates that theprobability that the estimated coefficient is not reliable. The
less the p -value the more it is reliable under the significance level. For example, if the
level of significance is 5% or (10%) it means than the value of p is less than 5% or (10%)
166 11. REGRESSION ANALYSIS AND CORRELATION
indicates that the estimated coefficient is reliable, otherwise it is unreliable and it should
be discarded from the model.

(d) Multiple R-squared: This shows the fraction (percentage) of the variation in a response
variable that is accounted for by independent variables in the model. It indicates how
well the terms fit the data. In addition, the adjusted R-squared is used to adjust for the
number of terms in a model. As long as you add more independent variables to a model,
the R-squared continue to increase in value, even when the variable is useless in the model.
However, the adjusted R-squared will increase if you add useful independent variable in
the model, otherwise the value of adjusted R-squared will decrease. The R-squared rages
from 0 to 1 but the adjusted R-squared can dip down to the negative value.

(e) F-statistics: This test for the significance of the overall coefficients whether the regression
model provides a better fit to the data than a model with no independent variables.

(f ) Durbin–Watson: It is used to test for the autocorrelation assumption of the error terms.
That is, to make sure that no correlation between the error terms Cov ."i ; "i 1 / D 0. Auto-
correlation may be caused by omission of important explanatory variable, misspecification
of the model, and systematic error in measurement. The consequences of autocorrelation
include the least square estimators will be inefficient and the estimated variances of the
regression coefficients will be biased and inconsistent, thus hypothesis testing is no longer
valid. Furthermore, Durbin–Watson ranges from 0–4. The value of 2 indicates no auto-
correlation between the error terms, between 0 to < 2 is a positive autocorrelation and
between 2 and 4 is negative autocorrelation. A rule of thumb for Durbin–Watson is that
for a relatively normal data, the test statistic should fall within 1.5–2.5.

Regression Output in R
After inputting the values for dependent and independent variables and combined the series
into dataframe, then the R function lm() is used to regress the dependent variable on the set of
independent variables. The function summary() gives the result below, as explained in Exam-
ple 11.2. We shall discuss most important results in this output; the regression coefficients of
the model are 17.68 and 0.37 with the corresponding p -values of 2e-16 each. This is very close
to 0 and much less than 5%. This indicates that the coefficients are reliable. Adjusted R-squared
(0.93) indicates that 93% variation in GDP is explained by variation government expenditure.
This shows that the model fits well. Also, the p -value of F-statistics is 2.2e-16, indicating that
all the regression coefficients (intercept and slope) are jointly significant.

Call:
lm(formula = LNGDP~LNGEXP)

Residuals:
 11.4. PEARSON CORRELATION COEFFICIENT 167
Min 1Q Median 3Q Max
-0.25038 -0.06998 -0.01893 0.04646 0.40962

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 17.68058 0.39751 44.48 <2e-16 ***
LNGEXP 0.37273 0.01791 20.81 <2e-16 ***
---
Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1

Residual standard error: 0.1328 on 33 degrees of freedom

Multiple R-squared: 0.9292, Adjusted R-squared: 0.9271
F-statistic: 433.2 on 1 and 33 DF, p-value: < 2.2e-16

11.4 PEARSON CORRELATION COEFFICIENT

The simple linear regression analysis shows the relationship between two variables that are lin-
early related. The correlation analysis is a measure of strength or level of association between
variables. The correlation is measured by the Pearson correlation coefiicient and it is denoted by
“r”. The statistic “r” ranges from 1 to C1. If the value of “r” is zero, it implies that there is no
linear association between the variables. As the value of correlation coefficient “r” move close to
zero (0), the linear association between the variables becomes weaker. Conversely, as the value
of the correlation coefficient “r” far away from zero and approaches ˙1, the linear association
between the variables becomes stronger. When the correlation coefficient is exactly 1, it is called
a perfect positive correlation and when the correlation coefficient is exactly 1, then it is known
as a perfect negative correlation (see Fig. 11.2).

-1 0 1
Perfect Negative Negative No Association Positive Perfect Positive
Association Association Association Association

Figure 11.2: Pearson correlation coefficient “r”.

The significance of a relationship is determined whether the Pearson’s correlation coeffi-

cient “r” is a meaningful reflection of the linear relationship between the two variables or whether
the relationship occurred by chance. For a given significant value (˛ D 0:05), the probability that
Pearson’s correlation coefficient “r” value comes by chance is 5% or less.
168 11. REGRESSION ANALYSIS AND CORRELATION
Pearson’s correlation coefficient “r” between variables X and Y can be defined as:
Pn


i D1 Xi X Yi Y
rXY D q
2 qPn
2 : (11.16)
Pn
iD1 Xi X i D1 Yi Y
Alternatively,
Pn Pn
Pn

n i D1 Xi Yi i D1 Yi iD1 Xi Cov.X; Y /
rXY D rh D ; (11.17)
P Pn
2 i h P n Pn
2
i SX :SY
n niD1 X 2 iD1 Xi n i D1 Y 2
i D1 Yi

where X and Y are the means of X and Y , respectively.

Cov.X; Y / is the covariance between X and Y .
SX and SY are standard deviations of X and Y , respectively.

11.4.1 ASSUMPTIONS OF CORRELATION TEST

The followings are the assumptions under the use of Pearson correlation test.
1. There must be independent observations.
2. The population correlation assumed to be zero, i.e.,  D 0.
3. The bivariates are normally distributed in the population.

11.4.2 TYPES OF CORRELATION

See Fig. 11.3.

11.4.3 COEFFICIENT OF DETERMINATION

The coefficient of determination is defined as the square of the correlation coefficient and it is
denoted by r 2 . It measures the strength of the relationship between two variables. The coeffi-
cient of determination is a measure of ratio of proportion variance explained by the model. That
is, it measures the percentage of variation in dependent variable that is accounted for, by the
variation in the independent variable. For instance, if r D 0:9, then coefficient of determina-
tion r 2 D 0:81. This indicates that 81% variation in the dependent variable is accounted for by
the variation in indepedent variable. Thus, the remaining 19% variation in dependent variable
cannot be explained by the variation in th independent variable.

11.4.4 TEST FOR THE SIGNIFICANCE OF CORRELATION COEFFICIENT

(R)
This significance test is used to test the null hupothesis that rX Y D 0. We should bear in mind
that the sampling distribution of r is approximately normal when sample size is large (n  30),
and distributed t when the sample size is small (n < 30).
 11.4. PEARSON CORRELATION COEFFICIENT 169
Perfect Positive Low Degree of Low Degree of
Perfect Negative Positive Correlation
Y Correlation Y Y Y Negative Correlation
Correlation

O X O X O X O X

High Degree of High Degree of

Positive Correlation Negative Correlation
Y No Correlation Y No Correlation
Y

O X O XO X O X

Figure 11.3: Types of correlation. Source: https://www.slideshare.net/RamKumarshah/co

rrelation-and-regression-56561989.

The t -test statistics for the significance of the correlation coefficient r is defined as:
r
n 2
t Dr  t1 ˛2 ;n 2 : (11.18)
1 r2
However, when n is sufficiently large .n  30/, we use the standardized score for the r using
Fisher z -transformation (z 0 ) to test for the significance of the correlation coefficient r :

z 0 D 0:5 .ln .1 C r/ ln .1 r//  z.1 ˛

2/
: (11.19)

The result in (11.19) is compared with the appropriate normal table.

Example 11.5
Using the compensation data in Example 11.3,
(i) calculate the Pearson’s correlation coefficient between X2 and Y ;
(ii) compute coefficient of determination in (i); and
(iii) test for the significant of the Pearson’s correlation coefficient between X2 and Y .
170 11. REGRESSION ANALYSIS AND CORRELATION
Solution:
(i) From the data in Example 11.3, we obtained the following values:
X X X X
X22 D 921; X12 D 387; X1 Y D 759:61; X2 Y D 1165:55;
X X X
X1 X2 D 499; n D 35; X1 D 107; X2 D 167;
X X
Y D 242:76; Y 2 D 1690:23:

The formula for the correlation coefficient is given by:

P Pn
Pn

n niD1 X2i Yi i D1 Yi i D1 X2i
rX2 Y D rh :
P Pn
2 i h Pn Pn
2 i
n niD1 X2i 2
X
i D1 2i n i D1 Y 2 Y
i D1 i

Substituting for the values in the formula, we have

35.1165:55/ .242:76/ .167/
rX2 Y D rh ih i D 0:26:
2 2
35 .921/ .167/ 35 .1690:23/ .242:76/

(ii) Coefficient of determination r 2 D 0:07.

(iii) Hypothesis: H0 W rX2 Y D 0 vs. H1 W rX2 Y ¤ 0.
r r
n 2 33
t Dr 2
D 0:26 D 1:55:
1 r 0:93

Critical value: t0:975;33 D 2:042.

Decision rule: Reject H0 if t  t0:975;33 , since 1:55 < 2:042, then we do not reject null hypoth-
esis.
Conclusion: The correlation coefficient between X2 and Y are not significantly different from
zero.

The R Script for Testing Correlation Coefficient for Small Sample

In the code below, we will use R to obtain the correlation coefficient and CI for the correlation.
We compute the correlation between years of experience (empl) and compensation on level of
education (comp). We got our t -statistic to be 1.53 with corresponding p -value of 0.14. This
indicates that the correlation between empl and comp is not statistically significant. In addition,
at 95% confidence interval, the true correlation between the two variable should lie between
0:08 and 0.54.
# Pearson’s product-moment correlation using Example 11.5 compesation data
 11.4. PEARSON CORRELATION COEFFICIENT 171
cor.test(mydata$empl, mydata$comp, method='pearson')

Pearson's product-moment correlation

data: mydata$empl and mydata$comp

t = 1.5264, df = 33, p-value = 0.1364
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
-0.08359426 0.54353702
sample estimates:
cor
0.2568063

Example 11.6
Use the following data to test for the significance of correlation coefficient:
n D 50; r D 0:75 and ˛ D 1%.

Solution:
Hypothesis: H0 W  D 0 vs. H1 W  ¤ 0.
Test statistic:
z 0 D 0:5 .ln .1 C r/ ln .1 r// :
0
Substitute for the value of r to get z :
z 0 D 0:5  .ln .1 C 0:75/ ln .1 0:75// D 0:973:

Critical value: z.1 ˛2 / D z0:995 D 2:57 (from the normal table).

Decision rule: Reject H0 if jz 0 j  z0:975 , since 0:973 < 2:57, then we do not reject null hypoth-
esis.
Conclusion: The correlation coefficient is not significantly different from zero.

The R Script for Testing Correlation Coefficient for Large Sample

This is simplify version on how to compute Fisher z -transformation (z 0 ) for the test of signifi-
cance of the correlation coefficient, r .
The Fisher z -transformation (z 0 ) for r
r<-0.75
z.prime<-0.5*(log(1+0.75)-log(1-0.75))
z.prime
[1] 0.9729551
Compute the critical value at 0.01 significance level
172 11. REGRESSION ANALYSIS AND CORRELATION
alpha = 0.01
z.prime = qnorm(1-alpha/2)
z.prime
[1] 2.575829
This result is similar to the result we obtained in Example 11.6.

11.5 EXERCISES
11.1. (a) What do you understand about the concept of regression analysis?
(b) With the aid of an example, differentiate between simple linear regression and
multiple linear regression.
(c) Table 11.4 shows data on the GDP and the government investment in Nigeria
between 1981 and 2016. Regress GDP growth rate on the total investment (%
GDP) and interpret your result.
11.2. (a) List the types of regression analysis you know.
(b) What are the assumptions of the simple linear regression?
(c) Table 11.5 shows the log of space area (measured in sq. feet), log of number of
bedrooms, and the log of house price (measured in $’million).
(i) Perform a regression analysis to show the relationship between the three (3)
variables. Take a log of house prices as the dependent variable (Y ).
(ii) Interpret your result.
(iii) Comment on the output .
11.3. (a) What do you understand about correlation coefficient?
(b) What are the assumptions of the correlation test?
(c) Explain the coefficient of determination.
(d) The summary statistics for the two variables are given below:
X X X
XY D 216:24; X D 21:20; Y D 204:21;
X X
X 2 D 22:90; Y 2 D 2187:59; n D 20:
(i) Calculate the Pearson’s correlation coefficient (r).
(ii) Test for the significance of r.
11.4. (a) Show that the least square estimates for the simple linear regression are given as:
ˇ0 D Y ˇ1 X and
Cov.X; Y /
ˇ1 D :
Var.X/
 11.5. EXERCISES 173
Table 11.4: GDP and government investment in Nigeria

GDP Total GDP Total

Year Growth Investment Year Growth Investment
Rate (% GDP) Rate (% GDP)
1981 -13.13 35.22 1999 0.47 6.99
1982 -1.05 31.95 2000 5.32 7.02
1983 -5.05 23.01 2001 4.41 7.58
1984 -2.02 14.22 2002 3.78 7.01
1985 8.32 11.97 2003 10.35 9.90
1986 -8.75 15.15 2004 33.74 7.39
1987 -10.75 13.61 2005 3.44 5.46
1988 7.54 11.87 2006 8.21 8.27
1989 6.47 11.74 2007 6.83 9.25
1990 12.77 14.25 2008 6.27 8.32
1991 -0.62 13.73 2009 6.93 12.09
1992 0.43 12.75 2010 7.84 16.56
1993 2.09 13.55 2011 4.89 15.53
1994 0.91 11.17 2012 4.28 14.16
1995 -0.31 7.07 2013 5.39 14.17
1996 4.99 7.29 2014 6.31 15.08
1997 2.80 8.36 2015 2.65 14.83
1998 2.72 8.60 2016 -1.62 12.60

(b) With the aid of diagram, explain the different types of correlation you know.
11.5. The strength (MPa) of a steel and the diameter (mm) are measured, as shown in the
table below.

Diameter (mm) 75 80 85 90 95 100 105 110 115 120

Strength (MPa) 250 320 350 400 435 480 500 520 550 570

(a) Draw a scatter diagram to illustrate the information.

(b) Calculate the least square line of regression of Strength (MPa).
(c) Interpret the model.
(d) Estimate the strength (MPa) of the steel when the diameter (mm) is 150.
174 11. REGRESSION ANALYSIS AND CORRELATION
Table 11.5: Log table

Log of space area (sq. feet) 8.51 8.77 8.55 8.63 7.49 8.61 8.55 8.44 8.42 8.34
Log of number of bedrooms 1.61 1.39 1.39 0.69 1.10 1.61 1.39 1.39 1.39 1.61
Log of house price ($' m) 0.34 0.22 0.30 0.18 0.20 0.26 0.35 0.27 0.14 0.41

Log of space area (sq. feet) 8.39 8.52 8.69 8.60 8.40 8.39 8.25 8.71 8.55 8.55
Log of number of bedrooms 1.10 1.61 1.10 0.69 1.61 1.10 1.10 1.61 1.39 1.39
Log of house price ($' m) 0.17 0.24 0.22 0.22 0.35 0.19 0.10 0.48 0.43 0.35

Log of space area (sq. feet) 8.48 8.38 8.76 8.59 8.37 8.47 8.63 8.56 8.61 8.22
Log of number of bedrooms 1.61 1.10 1.10 0.69 1.10 1.10 1.10 1.10 1.61 1.39
Log of house price ($' m) 0.36 0.23 0.48 0.25 0.18 0.23 0.29 0.20 0.36 0.34

11.6. The following table shows the prices (N) of bread in a supermarket and the quantity
sold in a week.
Price (X ) 100 200 500 750 1000 1500 1800 2000
Quantity sold (Y ) 55 49 40 37 35 27 15 8
(a) Plot the data on a scatter diagram.
(b) Draw the regression line on your scatter diagram.
(c) From the regression equation obtained in (b), estimate the quantity to be sold if
the price (N) of the bread is N2,200.
(d) Calculate Pearson’s correlation coefficient between price of bread and quantity sold.
11.7. The weekly number of sales in the two complementary commodities (DVD player and
DVD disk) are given in the following table.

Week (X ) 1 2 3 4 5 6 7 8
DVD player 20 23 18 25 15 17 16 20
DVD disks 38 39 25 28 23 26 25 30

(a) Calculate Pearson’s correlation coefficient between the number of sales of a DVD
player and DVD disk.
(b) Give interpretation of your result in (a).
(c) Test hypothesis about r D 0.
 175

CHAPTER 12

Poisson Distribution
The Poisson distribution was developed by the French mathematician Simeon Denis Poisson in
1837. The Poisson distribution is a discrete probability distribution. It is used to approximate the
count of events that occur randomly and independently. The Poisson distribution may calculate
number of instances that should occur in a certain amount of time, distance, area, or volume.
For instance, the random variable could be used to estimate the number of radioactive decays in
a given period of time for a certain amount of the radioactive material. If you know the rate of
decay for that amount of the material you can use the Poisson distribution as a good estimate of
the amount of decays. We shall elaborate on Poisson statistical properties, the derivation of the
mean and the variance of a Poisson distribution, and an application of the Poisson distribution.
Its practicality shall be demonstrated in the worked examples.

12.1 POISSON DISTRIBUTION AND ITS PROPERTIES

A discrete random variable (X ) is said to be a Poisson distribution function if it has a pdf of the
form:
x e 
f .x/ D ; x D 0; 1; 2; : : : ; 1; (12.1)
xŠ
where  is the shape parameter that indicates the average occurrence per interval and e is the
Euler’s constant, approximately 2.7183.
In Fig. 12.1, it is shown that the Poisson data skewed right and the skewness is less pro-
nounced as the lambda, , which is the mean of the distribution increases. Poisson is closer to
symmetric as the mean () of the distribution increases.
This distribution is most applicable where we have a rare event and the number of trials
is large. It implies that the rate of occurrence is small, near to zero in a large number of trials.
Some examples of a Poisson distribution are the occurrence of the dropping of a fish from the
sky during rainfall, the occurrence of a fatal accident within a city, the number of automobiles
arriving at a traffic light with a specified period of time, the number of typographical errors in a
textbook, among others.
Poisson takes the values from zero to infinity, and the expected occurrences remain un-
changed throughout the experiment. For a Poisson random variable, these two conditions should
be met: the number of successes in two disjoint time intervals is independent and the probabil-
ity of success in a short time interval is proportional to the entire length of the time interval.
Alternatively, a Poisson random variable can also be applied to disjoint regions of space. Fig-
176 12. POISSON DISTRIBUTION

45
40
35
30
25

P(X = x)
20 lambda = 1
15 lambda = 2
10 lambda = 3
5 lambda = 10
0
-5
0 5 10 15 20
x

Figure 12.1: Poisson distributions.

Table of Poisson Probabilities

For a given value of λ, entry indicates the probability of a specifiec value of X.

Figure 12.2: Table of Poisson probabilities.

 12.2. MEAN AND VARIANCE OF A POISSON DISTRIBUTION 177
ure 12.2 shows the table of Poisson probabilities for a given value of  and the number of success
recorded. The values in this table are computed using the formula in Equation (12.1) above. This
Poisson table can be read directly without computational effort again. For example, suppose that
a car dealer sells on the average two exotic cars per week and we want to calculate the probability
that in a given week, he will sell one exotic car. We can check the table of Possion probabilities
where mean ( D 2) and the number of success .x D 1/. From the table of Poisson probabilities,
the required probability is 0.2707; this result is the same by substituting for  D 2 and x D 1 in
Equation (12.1) above. Using the formula, we have:
21 e 2
f .1/ D D 0:270671:
1Š

12.2 MEAN AND VARIANCE OF A POISSON

DISTRIBUTION
In this section, we derive the mean and variance of a Poisson distribution mathematically. Here,
you will see that the mean and variance of a Poisson distribution are equal.

Mean
The following is our standard form of the mean:
X
E .x/ D xf .x/:

Substituting the Poisson function we get

X1
x e 
E .x/ D x
xD0
xŠ
or
X1
x 1 e  
E .x/ D x
xD0
x.x 1/Š
and
X1
x 1 e 
E .x/ D 
xD0
.x 1/Š
since
X1
x 1 e 
D 1:
xD0
.x 1/Š
Therefore,
E .x/ D : (12.2)
So the mean of the Poisson distribution is lambda.
178 12. POISSON DISTRIBUTION
Variance
Using our standard formula for variance we have


variance D E x 2 .E .x//2 :

We will use the following clever fact to simplify the formula for variance of the Poission distri-
bution in the following few steps.
From x 2 D x
.x 1/ C x .
Then, E x 2 D E .x.x 1// C E .x/.
1
X
E .x.x 1// D x .x 1/ f .x/
xD0
X1
x e 
E .x.x 1// D x .x 1/
xD0
xŠ
X1
x 2 e  2
E .x.x 1// D x .x 1/
xD0
x.x 1/.x 2/Š
X1
x 2 e 
E .x.x 1// D 2
xD0
.x 2/Š
P x 2 e 
Since 1 xD0 .x 2/Š D 1.
Then E .x.x 1// D 2 .
And, therefore,
var.x/ D E .x.x 1// C E .x/ .E .x//2 :
And since E.x/ D  from our derivation of the mean above,

var.x/ D 2 C  2 :

Thus,
var.x/ D : (12.3)
Also, the variance of the Poisson distribution is lambda. This shows that the mean and variance
of a Poisson distribution are the same.

Example 12.1
A paper mill produces writing pads and the probability of a writing pad being defective is 0.02.
If a sample of 800 writing pads is selected, what is the probability that: (a) none are defective,
(b) one defective, (c) two are defectives, and (d) three or more are defective writing pads.
 12.2. MEAN AND VARIANCE OF A POISSON DISTRIBUTION 179
Solution:
In Example 11.1, the probability of a writing pad being defective .p/ D 0:02 and sample selected
n D 800.
Calculate the lambda (mean),

 D np D 800  0:02 D 16:

(a) The probability that none of the selected sample are defective:

160 e 16
16
f .0/ D De :
0Š

(b) The probability that one is defective:

161 e 16
16
f .1/ D D 16e :
1Š

(c) The probability that two writing pads are defectives:

162 e 16
16
f .2/ D D 128e :
2Š

(d) The probability that three or more writing pads are defectives:
16 16 16
f .x  3/ D 1 .f .0/ f .1/ f .2// D 1 e 16e 128e
16 16 16
f .x  3/ D 1 f .x < 3/ D 1 e 16e 128e D 0:999:

R Codes
The following are solutions to Example 12.1.
To calculate the probability of a Poisson distribution, the general format is:

ppois(q, lambda, lower.tail = TRUE, log.p = FALSE)

where
q vector of quantiles
lambda vector of positive means
lower.tail logical; if TRUE(default), probabilities are P .X  x/,
otherwise P .X > x/
log.p logical; if TRUE, probabilities P are given as log.p/.

1. The probability that none of the selected samples are defective:

180 12. POISSON DISTRIBUTION
n<-800
p<-0.02
lambda <-n*p
ppois(0, lambda=16, lower = TRUE)
[1] 1.125352e-07
16
This result is the same as e gotten in Example 12.1 above.
2. The probability that one is defective:
exact.1<-ppois(1, lambda=16)- ppois(0, lambda=16)
[1] 1.800563e-06
16
The evaluation of the quantity 16e gives 1.800563e-06.
3. The probability that two writing pads are defectives:
exact.2<-ppois(2, lambda=16) - ppois(1, lambda=16)
[1] 1.44045e-05
16
The answer 1.44045e-05 is the same as 128e .
4. The probability that three or more writing pads are defectives:
ppois(3, lambda=16, lower = FALSE) # upper tail
[1] 0.9999069

The upper tail is used here because we are considering the right side of the value 3 or value
of 3 and above. Thus, we are able to use R code to solve the problem in Example 12.1
above.

Example 12.2
The number of arrivals of customers into an eatery per minute has a Poisson with mean 3. Assum-
ing that the number of arrivals in two different minutes are independent. Find the probability
that:
1. no calls come in a period of a minute;
2. one call comes in a period of a minute; and
3. at least two customers will arrive in a given two-minute period.

Solution:
The mean of the Poisson distribution is 3.
 12.2. MEAN AND VARIANCE OF A POISSON DISTRIBUTION 181
1.  D 3.
Substitute for  D 3 and x D 0
30 e 3
f .0/ D D 0:0498:
0Š

2. The probability that one call comes in a period of a minute;

31 e 3
f .1/ D D 0:1494:
1Š

3. The probability that at least two customers will arrive in a given two-minute period:

f .x  2/ D 1 f .0/ f .1/
f .x  2/ D 1 0:0498 0:1494
f .x  2/ D 0:8009:

R Codes to Calculate the Probability of the Poisson Distribution in Example 12.2

lambda =3
p0<-ppois(0, lambda=3, lower = TRUE)
[1] 0.04978707

The probability that no calls come in a period of a minute is 0.0498.

exact.p1<- ppois(1, lambda) - ppois(0, lambda)
[1] 0.1493612

Despite omitting the option “lower” in the statement above, R assumes lower.tail = TRUE
by default.
p.greater2<- 1- p0 - exact.p1
[1] 0.8008517

The probability that at least two customers will arrive in a given two-minute period is 0.8009.
We got similar results with Example 12.2 above.

Example 12.3
The operational manager has the option of using one of the two machines (A and B) to produce
a particular product. He knew that the energy output of both machines is represented well by
a Poisson distribution with machine A having a mean of 8.25 and machine B has a mean of
7.50. Assuming that the energy input remains the same, the efficiency of machine A is f .x/ D
2x 2 8x C 6 and the efficiency of machine B is f .y/ D y 2 C 2y C 1. Which of the machines
has the maximum expected efficiency?
182 12. POISSON DISTRIBUTION
Solution:
Machine A expected efficiency is:

E.x/ D E.2x 2 8x C 6/
2


E .x/ D 2E x 8E.x/ C E.6/:

From a general equation,

Var .x/ D E x 2 .E .x//2 :
It implies that

E x 2 D Var .x/ C .E .x//2 :


Substitute for E x 2 to we get
h i
E .x/ D 2 Var .x/ C .E .x//2 8E.x/ C 6:

Also, substitute for the mean and variance of the distribution to get
 
E .x/ D 2 8:25 C 8:252 .8  8:25/ C 6 D 92:63:

Machine B’s expected efficiency is:

E.y/ D E.y 2 C 2y C 1/


E .y/ D E y 2 C 2E.y/ C E.1/
h i
E .y/ D Var .y/ C E .y/2 C 2E.y/ C 1;

since

E y 2 D V .y/ C E .y/2 :
Therefore,  
E .y/ D 7:50 C 7:502 C .2  7:50/ C 1 D 79:75:
Machine A is more efficient than machine B.

R Codes for the Example 12.3

We assigned for the values of mean and variance to be 8.25 each:
# compute efficiency for machine A
mean.x<-8.25
var.x<-8.25
h i
Referring to E .x/ D 2 Var .x/ C .E .x//2 8E.x/ C 6, we obtained efficiency for the ma-
chine A as eff.x.
 12.3. APPLICATION OF POISSON DISTRIBUTION 183
eff.x<-2*(var.x+mean.x**2)-8*mean.x+6
[1] 92.625

Also, we assigned for the values of mean and variance to be 7.50 each and then compute efficiency
for machine B:

# compute efficiency for machine B

mean.y<-7.50 # expected efficiency
var.y<-7.50 # variance efficiency
eff.y<- var.y+mean.y**2+2*mean.y+1
[1] 79.75

The efficiency of machine A is 92.63 and machine B is 79.75, therefore machine A is more
efficient.

12.3 APPLICATION OF POISSON DISTRIBUTION

Poisson distribution has its application in various aspects of life. A few areas shall be considered
in this section. In healthcare, the Poisson distribution is helpful in birth detection and genetic
mutations. It also helps to model the occurrence of rare disseases such as leukemia. In the area
of transport, Poisson is good at accounting and predicting the occurrence of accidents in a high-
way, monitoring the number of automobiles arriving at a traffic light within a one-hour period.
Poisson can also be applied in the telecomunication and communication sectors by determining
the number of calls received by the customer care center within a minute and the number of
network failures per day. In the area of production, it can be used to capture the failure of a
machine per month and the number of defective products in a production line.

12.4 POISSON TO APPROXIMATE THE BINOMIAL

The Poisson distribution is a special case of the binomial distribution and Poisson is useful in
handling rare events. In the event that the binomial random variable has an extremely large
number of trials (n) and the probability of success (p ) is very small, then the Poisson distribution
provides a good approximation of the binomial distribution when n  100 and np  10.

Proof. The Poisson () is an approximation to the binomial .n; p/ for a large n, small p , and
 D np .
Then p D n .
Substitute for p into the binomial distribution and then take the limit as n tends to infinity:
 k  n k
nŠ  
lim P .X D k/ D lim 1 : (12.4)
n!1 n!1 .n k/ŠkŠ n n
184 12. POISSON DISTRIBUTION
Take out the constant terms in (12.4):
 k  n   k
k nŠ 1  
lim P .X D k/ D lim 1 1 : (12.5)
n!1 kŠ n!1 .n k/Š n n n

We can take the limit of the RHS one term after the other in (12.5):
 k
nŠ 1
lim
n!1 .n k/Š n
 
n .n 1/ .n 2/ : : : .n k/.n k 1/ 1 k
lim
n!1 .n k/ .n k 1/ : : : .1/ n

n .n 1/ .n 2/ : : : .n k C 1/
lim : (12.6)
n!1 nk
As n ! 1, then k terms tend to 1.
Equation (12.6) can be written as:
    
n.n 1/.n 2/ : : : .n k C 1/  n  n 1 n 2 n kC1
lim ::: : (12.7)
n!1 nk n n n n

The second step is to take the limit of the middle term in (12.5):
 n

limn!1 1 : (12.8)
n

Since e D 2:718, then

 x
1
e D lim 1 : (12.9)
x!1 x
Let x D n .
Substitute for x into (12.8); then we have
 n  x. /
 1 
lim 1 D lim 1 De : (12.10)
n!1 n x!1 x

Take the limit of the last term of the RHS of (12.5):

  k

lim 1 : (12.11)
n!1 n
k
As n ! 1, then .1/ tends to 1.
 12.4. POISSON TO APPROXIMATE THE BINOMIAL 185
Putting all the results (12.7), (12.10) and (12.11) together into (12.5), we have
 k    
k nŠ 1  n  k
lim P .X D k/ D lim 1 1
n!1 kŠ n!1 .n k/Š n n n
!
 k  
D .1/ e  .1/: (12.12)
kŠ
 k 
Therefore, P .; k/ D  kŠ e
.
This gives Poisson distribution with k successes per period given parameter . 

Example 12.4
Suppose a random variable X has a binomial distribution with n D 120 and p D 0:01, and use
the Poisson distribution to calculate the following: (a) P .X D 0/, (b) P .X D 1/, (c) P .X D 2/,
and (d) P .X > 2/.

Solution:
(a)  D np D 1:2.
1:20 e 1:2
P .X D 0/ D D 0:3012:
0Š
1:21 e 1:2
(b) P .X D 1/ D 1Š
D 0:3614.
1:22 e 1:2
(c) P .X D 2/ D 2Š
D 0:2169.

(d) P .X > 2/ D 1 P .X  2/ D 1 P .X D 0/ P .X D 1/ P .X D 2/ D 0:1205.

R Codes for Example 12.4

Let’s write some short codes in R to solve the Example 12.4 above:
# Question 12.4a
n<-120
p<-0.01
lambda = n*p
p0<-ppois(0, lambda, lower = TRUE)
[1] 0.3011942
The probability that P .X D 0/ is 0.3012:
# Question 12.4b
p1<-ppois(1, lambda, lower = TRUE) - p0
[1] 0.3614331
186 12. POISSON DISTRIBUTION
The probability that P .X D 1/ is 0.3614:
# Question 12.4c
p2<-ppois(2, lambda)-ppois(1, lambda)
p2
[1] 0.2168598
The probability that P .X D 2/ is 0.2169:
# Question 12.4d
p.greater2<-1-ppois(2, lambda, lower=TRUE)
p.greater2
[1] 0.1205129
The probability that P .X > 2/ is 0.1205.

12.5 EXERCISES
12.1. (a) Define the Poission distribution and the properties of the distribution.
(b) Show that the mean and the variance of a Poisson distribution are equal.
(c) A production manager took a sample of 25 textbooks and examined them for the
number of defectives pages. The outcome of his findings is summarized in the table
below.
Find the probability of finding a textbook chosen at random that contains two or
more defective pages.

Number of defectives 0 1 2 3 4 5
Frequency 10 4 3 2 3 3

12.2. (a) Show that the Poisson ( ) is an approximation to the binomial .n; p/ as n ! 1
and p ! 0.
(b) Consider a random variable X that has a binomial distribution with n D 100 and
p D 0:005. Use the Poisson distribution to calculate the following: (i) P .X D 0/,
(ii) P .X D 1/, and (iii) P .X > 1/.
12.3. (a) Explain the area of life that Poisson process be applied.
(b) The ABC company supplies a supermarket with some groceries weekly. Experience
has shown that 3% of the total supply in a given week of the groceries is defective.
If the manager of the supermarket decides to check the number of defective items
for a particular grocery and he took a sample of 105 items, what is the probability
that: (i) none of the product is defective? (ii) one of the products is defective? (iii) 2
or more of the products are defective?
 12.5. EXERCISES 187
12.4. A specialist hospital recorded 200 deliveries in every 30 days and the management of
the hospital observed that most of the deliveries took place in the early hours of the
day, between 12:00 AM and 3:00 AM. Therefore, the management decided to make
as many staff as possible available during these time periods to show their dedication
to work. Using the Poisson distribution, find the probability of delivering: (a) no baby,
(b) one baby, (c) two babies, and (d) three babies in the early hours of the day. Hence,
how many days in a 30-day time period would 4 or more deliveries are expected?
12.5. An insurance company sells a special life insurance policy to people that are over the age
of 50. The actuarial probability that somebody of age 50 and above will die within one
year of the policy is 0.0008. If the special life insurance policy is sold to 7,500 people
of the same age group, this is an indication that there is possibility that 6 people aged
50 years or older will die within the next year. What is the probability that the insurance
company will pay exactly 6 claims on the 7,500 policies sold in the next year?
12.6. A car dealer claimed that he sold an average of three new brand cars per week. Assuming
that the sales follow a Poisson distribution, what is the probability that: (a) he will sell
exactly three new cars, (b) less than three new cars, or (c) more than three new cars in
a given week?
 189

CHAPTER 13

Uniform Distributions
13.1 UNIFORM DISTRIBUTION AND ITS PROPERTIES
Let X be a continuous random variable. Then X is said to be a uniform distribution over the
interval Œa; b if its probability density function is defined as:
1
f .x/ D ; a  x  b:
b a
The uniform distribution is denoted as X  U.a; b/. The uniform distribution is also known as
rectangular distribution.

f (x)

1/(b-a)

x
a b

Figure 13.1: A uniform distribution.

Figure 13.1 shows a rectangle with the length of the base .b a/ and a height of b 1 a . The
total area under the curve of pdf is the product of the height of the rectangle and length of the
base, thus the total area is 1. The area under f .x/ and between the points a and b is expected to
be 1 and f .x/ > 0, therefore, f .x/ is a probability density function. There is an infinite number
of possible values of a and b , thus there is an infinite number of possible uniform distributions.
The commonly used continuous uniform distribution is a D 0 and b D 1. Uniform distribution
is useful when every variable has an equal or exact chance of happening.

13.2 MEAN AND VARIANCE OF A UNIFORM

DISTRIBUTION
We shall derive the mean and then the variance of a uniform distribution in this section.
190 13. UNIFORM DISTRIBUTIONS
Mean
The expected value of X
Z b
E .x/ D xf .x/ dx:
a

Substituting for f .x/ yields

Z b
x
E .x/ D dx:
a b a
Solving the integral yields
 b
x2
E .x/ D
2.b a/ a
 
b2 a2 b 2 a2
E .x/ D D
2.b a/ 2.b a/ 2.b a/
.b C a/.b a/ .b C a/
E .x/ D D :
2.b a/ 2
.bCa/
So the mean is 2
and it is an average of the two limits (a and b ).

Variance
The variance of a uniform distribution is derived as follows:


Var.x/ D E x 2 .E .x//2
Z b


E x2 D x 2 f .x/ dx
a
Z b
2

1
E x D x2 dx
a b a
Z b

1
E x2 D x 2 dx:
b a a
Solving the integral we get
b 

x3 1
E x2 D
b a 3 a

1  3 b
E x2 D x a
3.b a/

1  3 
E x2 D b a3 ;
3.b a/
 
since b 3 a3 D .b 2 C ab C a2 /.b a/.
 13.2. MEAN AND VARIANCE OF A UNIFORM DISTRIBUTION 191

h i
Then, E x 2
D 3.b1 a/ .b 2 C ab C a2 /.b a/

1h 2 i
E x2 D .b C ab C a2 / ;
3
since


Var.x/ D E x 2 .E .x//2 :

Then,
 2
1h 2 i .b C a/
Var.x/ D .b C ab C a2 /
3 2
.b 2 C ab C a2 / .b C a/2
Var .x/ D
3 4
4.b C ab C a / 3.b C a/2
2 2
Var .x/ D
12
4b C 4ab C 4a2 3b 2 6ab
2
3a2
Var .x/ D
12
2 2
b 2ab C a
Var .x/ D
12
.b a/2
Var .x/ D :
12
.b a/2
Hence, the variance of a uniform distribution is 12
. That means, the square of the difference
of point a and b divided by 12.

Example 13.1
Upon arriving at his office building, John wanted to take an elevator to the 10th floor
where his office is located. Due to the congestion of the building, it takes between 0 and 60 s
before the elevator arrives at the ground floor. Assuming that the arrival of the elevator to the
ground floor is uniformly distributed. (a) What is the probability that the elevator takes less than
15 s to arrive at the ground floor? (b) Find the mean of the arrival of the elevator to the ground
floor. (c) Find the standard deviation of the arrival of the elevator to the ground floor.

Solution:
(a) Let the intervals in seconds be a D 0, b D 60, and c D 15.
192 13. UNIFORM DISTRIBUTIONS

Z 15
1
P .0  x  15/ D dx
0 b a
Z 15
1
P .0  x  15/ D dx
060 0
h x i15
P .0  x  15/ D
 60 0 
15 0
P .0  x  15/ D D 0:25:
60
The possibility that the elevator will arrive in 15 s is one fourth of the time.
(b) mean D E .x/ D .bCa/ 2
.
.60 C 0/
E .x/ D D 30 s:
2
The mean arrival time2
is 30 s.2
2
(c) Var .x/ D .b 12a/ D .60120/ D 60 12
.
s
602 60
SD .x/ D D p D 17:32 s:
12 12
The standard deviation is just over 17 s.

f (x)

1/60

0 60

Figure 13.2: Interval in seconds.

R Codes of a Uniform Distribution

The distribution function of a uniform distribution in R is generated by function punif. The
format is:
 13.2. MEAN AND VARIANCE OF A UNIFORM DISTRIBUTION 193
punif(q, min = 0, max = 1, lower.tail = TRUE, log.p = FALSE)

where
q vector of quintiles
min lower limit of the distribution
max upper limit of the distribution
lower.tail logical; if TRUE (default), probabilities are P[X  x],
otherwise, P[X > x]
log.p logical; if TRUE, probabilities p are given as log.p/.
The R code that provides the solutions to Example 13.1 is as follows.
# Solution to Example 13.1a
punif(15,min=0,max=60)
[1] 0.25

The probability that the elevator will arrive in 15 s is 0.25.

# Solution to Example 13.1b
min<-0
max<-60
punif(15,min=0,max=60)

# Solution to Example 13.1c

mean.x<-(max+min)/2
mean.x
[1] 30

The mean arrival time is 30 s.

# Solution to Example 13.1d
var.x<-((max-min)**2)/12
sd.x<-sqrt(var.x)
sd.x
[1] 17.32051

The standard deviation is about 17 s.

Example 13.2
In a debate competition that consists of 10 participants, the moderator gave each participant
5 min to talk on the debate topic. (a) Find the probability that the participants will finish their
194 13. UNIFORM DISTRIBUTIONS
talk within 3 min and 30 s. (b) How many of these participants can finish up the talk within
3 min and 30 s?

Solution:
Let’s convert all the time interval to seconds.
Interval (in second) of probability distribution D Œ0; 300.
1 1
f .x/ D D
b a 300

Z 210
1
P .0  x  210/ D dx
0 300
h x i210
P .0  x  210/ D
 300 0
210
P .0  x  210/ D D 0:7:
300

Therefore, the probability that the participants will finish their talk within 3 min and 30 s is 0.7.
The number of participants that can finish up the talk within 3 min and 30 s is: 0:7  10 D 7
participants.

R Codes

# Solution to Example 13.2a

p.210<-punif(210,min=0,max=300)
p.210
[1] 0.7

The possibility that the participants will finish their talk within 3 min and 30 s is 0.7.
# Solution to Example 13.2b
n.participants<- p.210*10
n.participants
[1] 7

Thus, 7 participants can finish up the talk within 3 min and 30 s.

Example 13.3
Two friends, Peter and Paul, agreed to meet at the school library for a study session. Both of
them randomly arrive between 5:45 PM and 6.00 PM. What is the probability that Peter arrives
at the venue at least 5 min before Paul?
 13.2. MEAN AND VARIANCE OF A UNIFORM DISTRIBUTION 195
Solution:
Let x be arrival time for Peter and y be arrival time for Paul.
Time interval D 6:00 PM 5:45 PM D 15 min.
Time interval of probability distribution D Œ0; 15.
1 1
f .x/ D D
b a 15
Z 15
1
P .x  5/ D dx
5 15
h x i15
P .x  5/ D
15 5
15 5 10
P .x  5/ D D D 0:67:
15 15 15
The possibility that Peter arrives at the venue at least 5 min before Paul is 0.67.

R Codes

# Solution to Example 13.3

p.greater5<-1-punif(5,min=0,max=15)
p.greater5
[1] 0.6666667

This result .0:67/ gives the possibility of Peter arriving the venue at least 5 min before Paul.

Example 13.4
A filling station manager claimed that the minimum volume of PMS sold per day is 5,000 L
and the maximum of the product sold is 6,500 L per daily. Assume that the service at the filling
station is uniform distribution. Find the probability that the volume of the product to be sold
per day will fall between 5,500 and 6,200 L.
196 13. UNIFORM DISTRIBUTIONS
Solution:
Given a D 5; 000 and b D 6; 500,
1 1
f .x/ D D
b a
1500
Z
6200
1
P .5500  x  6200/ D dx
5500 1500
h x i6200
P .5500  x  6500/ D
1500 5500
6200 5500
P .5500  x  6500/ D D 0:47:
1500 1500

R Codes

# Solution to Example 13.4

p.6200<-punif(6200,min=5000,max=6500)
[1] 0.8

The probability that the volume of the product to be sold per day is 6,200 L is 0.8.
p.5500<-punif(5500,min=5000,max=6500)
[1] 0.3333333

The probability that the volume of the product to be sold per day is 5,500 L is 0.33.
btw5500_6200<-p.6200-p.5500
btw5500_6200
[1] 0.4666667

The probability that the volume of the product to be sold per day will fall between 5,500 and
6,200 L is 0.47; this is the difference in the probability of sale of 6,200 L and 5,500 L.

13.3 EXERCISES
13.1. (a) Define the continous uniform distribution
(b) Suppose X has a uniform distribution on the interval Œ10; 50. That is, X 
U.10; 50/. Find the probability that: (i) P .X  25/, (ii) P .20  X  32/, and (iii)
P .X  32/.
.ˇ C˛/
13.2. (a) Let X  U.˛; ˇ/ and show that the mean of a uniform distribution is 2
and
.ˇ ˛/2
variance 12
.
2
(b) If f .x/ D 3x , 0  x  2. Find the expected of X .
 13.3. EXERCISES 197
13.3. The waiting time for a bus to arrive at a bus stop is 10 min. Find the probability that a
bus will come within 7 min of waiting at the bus stop. Asume that the waiting time is
uniformly distributed.
13.4. A recruitment agency conducted an interview for 20 job seekers and scheduled 5 min
for each of the job seekers to express themselves on what changes and contributions
they could bring into the company to which they were applying. Find the probability
that: (a) the job seekers finish within 3 min, and (b) the job seekers finish after 3 min.
(c) How many of these job seekers can express themselves for more than 3 min?
13.5. Suppose X  U.2; 10/. (a) What is the probability density function f .x/ of X ?
(b) Sketch the probability

density function f .x/ of X . (c) What is P .5  X  10/?
(d) What is E 2X 2 X ?
 199

APPENDIX A

Tables
We provide tables in the following pages.
200 A. TABLES
Standard Normal Distribution Values (z ≤ 0)
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.0 0.50000 0.49601 0.49202 0.48803 0.48405 0.48006 0.47608 0.47210 0.46812 0.46414
0.1 0.46017 0.45621 0.45224 0.44828 0.44433 0.44038 0.43644 0.43251 0.42858 0.42466
0.2 0.42074 0.41683 0.41294 0.40905 0.40517 0.40129 0.39743 0.39358 0.38974 0.38591
0.3 0.38209 0.37828 0.37448 0.37070 0.36693 0.36317 0.35942 0.35569 0.35197 0.34827
0.4 0.34458 0.34090 0.33724 0.33360 0.32997 0.32636 0.32276 0.31918 0.31561 0.31207
0.5 0.30854 0.30503 0.30153 0.29806 0.29460 0.29116 0.28774 0.28434 0.28096 0.27760
0.6 0.27425 0.27093 0.26763 0.26435 0.26109 0.25785 0.25463 0.25143 0.24825 0.24510
0.7 0.24196 0.23885 0.23576 0.23270 0.22965 0.22663 0.22363 0.22065 0.21770 0.21476
0.8 0.21186 0.20897 0.20611 0.20327 0.20045 0.19766 0.19489 0.19215 0.18943 0.18673
0.9 0.18406 0.18141 0.17879 0.17619 0.17361 0.17106 0.16853 0.16602 0.16354 0.16109
1.0 0.15866 0.15625 0.15386 0.15151 0.14917 0.14686 0.14457 0.14231 0.14007 0.13786
1.1 0.13567 0.13350 0.13136 0.12924 0.12714 0.12507 0.12302 0.12100 0.11900 0.11702
1.2 0.11507 0.11314 0.11123 0.10935 0.10749 0.10565 0.10384 0.10204 0.10027 0.09853
1.3 0.09680 0.09510 0.09342 0.09176 0.09012 0.08851 0.08692 0.08534 0.08379 0.08226
1.4 0.08076 0.07927 0.07780 0.07636 0.07493 0.07353 0.07215 0.07078 0.06944 0.06811
1.5 0.06681 0.06552 0.06426 0.06301 0.06178 0.06057 0.05938 0.05821 0.05705 0.05592
1.6 0.05480 0.05370 0.05262 0.05155 0.05050 0.04947 0.04846 0.04746 0.04648 0.04551
1.7 0.04457 0.04363 0.04272 0.04182 0.04093 0.04006 0.03920 0.03836 0.03754 0.03673
1.8 0.03593 0.03515 0.03438 0.03363 0.03288 0.03216 0.03144 0.03074 0.03005 0.02938
1.9 0.02872 0.02807 0.02743 0.02680 0.02619 0.02559 0.02500 0.02442 0.02385 0.02330
2.0 0.02275 0.02222 0.02169 0.02118 0.02068 0.02018 0.01970 0.01923 0.01876 0.01831
2.1 0.01786 0.01743 0.01700 0.01659 0.01618 0.01578 0.01539 0.01500 0.01463 0.01426
2.2 0.01390 0.01355 0.01321 0.01287 0.01255 0.01222 0.01191 0.01160 0.01130 0.01101
2.3 0.01072 0.01044 0.01017 0.00990 0.00964 0.00939 0.00914 0.00889 0.00866 0.00842
2.4 0.00820 0.00798 0.00776 0.00755 0.00734 0.00714 0.00695 0.00676 0.00657 0.00639
2.5 0.00621 0.00604 0.00587 0.00570 0.00554 0.00539 0.00523 0.00509 0.00494 0.00480
2.6 0.00466 0.00453 0.00440 0.00427 0.00415 0.00403 0.00391 0.00379 0.00368 0.00357
2.7 0.00347 0.00336 0.00326 0.00317 0.00307 0.00298 0.00289 0.00280 0.00272 0.00264
2.8 0.00256 0.00248 0.00240 0.00233 0.00226 0.00219 0.00212 0.00205 0.00199 0.00193
2.9 0.00187 0.00181 0.00175 0.00170 0.00164 0.00159 0.00154 0.00149 0.00144 0.00140
3.0 0.00135 0.00131 0.00126 0.00122 0.00118 0.00114 0.00111 0.00107 0.00104 0.00100
3.1 0.00097 0.00094 0.00090 0.00087 0.00085 0.00082 0.00079 0.00076 0.00074 0.00071
3.2 0.00069 0.00066 0.00064 0.00062 0.00060 0.00058 0.00056 0.00054 0.00052 0.00050
3.3 0.00048 0.00047 0.00045 0.00043 0.00042 0.00040 0.00039 0.00038 0.00036 0.00035
3.4 0.00034 0.00033 0.00031 0.00030 0.00029 0.00028 0.00027 0.00026 0.00025 0.00024
3.5 0.00023 0.00022 0.00022 0.00021 0.00020 0.00019 0.00019 0.00018 0.00017 0.00017
 A. TABLES 201
Standard Normal Distribution Values (z ≥ 0)
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.0 0.50000 0.50399 0.50798 0.51197 0.51595 0.51994 0.52392 0.52790 0.53188 0.53586
0.1 0.53983 0.54380 0.54776 0.55172 0.55567 0.55962 0.56356 0.56749 0.57142 0.57535
0.2 0.57926 0.58317 0.58706 0.59095 0.59483 0.59871 0.60257 0.60642 0.61026 0.61409
0.3 0.61791 0.62172 0.62552 0.62930 0.63307 0.63683 0.64058 0.64431 0.64803 0.65173
0.4 0.65542 0.65910 0.66276 0.66640 0.67003 0.67364 0.67724 0.68082 0.68439 0.68793
0.5 0.69146 0.69497 0.69847 0.70194 0.70540 0.70884 0.71226 0.71566 0.71904 0.72240
0.6 0.72575 0.72907 0.73237 0.73565 0.73891 0.74215 0.74537 0.74857 0.75175 0.75490
0.7 0.75804 0.76115 0.76424 0.76730 0.77035 0.77337 0.77637 0.77935 0.78230 0.78524
0.8 0.78814 0.79103 0.79389 0.79673 0.79955 0.80234 0.80511 0.80785 0.81057 0.81327
0.9 0.81594 0.81859 0.82121 0.82381 0.82639 0.82894 0.83147 0.83398 0.83646 0.83891
1.0 0.84134 0.84375 0.84614 0.84849 0.85083 0.85314 0.85543 0.85769 0.85993 0.86214
1.1 0.86433 0.86650 0.86864 0.87076 0.87286 0.87493 0.87698 0.87900 0.88100 0.88298
1.2 0.88493 0.88686 0.88877 0.89065 0.89251 0.89435 0.89617 0.89796 0.89973 0.90147
1.3 0.90320 0.90490 0.90658 0.90824 0.90988 0.91149 0.91308 0.91466 0.91621 0.91774
1.4 0.91924 0.92073 0.92220 0.92364 0.92507 0.92647 0.92785 0.92922 0.93056 0.93189
1.5 0.93319 0.93448 0.93574 0.93699 0.93822 0.93943 0.94062 0.94179 0.94295 0.94408
1.6 0.94520 0.94630 0.94738 0.94845 0.94950 0.95053 0.95154 0.95254 0.95352 0.95449
1.7 0.95543 0.95637 0.95728 0.95818 0.95907 0.95994 0.96080 0.96164 0.96246 0.96327
1.8 0.96407 0.96485 0.96562 0.96638 0.96712 0.96784 0.96856 0.96926 0.96995 0.97062
1.9 0.97128 0.97193 0.97257 0.97320 0.97381 0.97441 0.97500 0.97558 0.97615 0.97670
2.0 0.97725 0.97778 0.97831 0.97882 0.97932 0.97982 0.98030 0.98077 0.98124 0.98169
2.1 0.98214 0.98257 0.98300 0.98341 0.98382 0.98422 0.98461 0.98500 0.98537 0.98574
2.2 0.98610 0.98645 0.98679 0.98713 0.98745 0.98778 0.98809 0.98840 0.98870 0.98899
2.3 0.98928 0.98956 0.98983 0.99010 0.99036 0.99061 0.99086 0.99111 0.99134 0.99158
2.4 0.99180 0.99202 0.99224 0.99245 0.99266 0.99286 0.99305 0.99324 0.99343 0.99361
2.5 0.99379 0.99396 0.99413 0.99430 0.99446 0.99461 0.99477 0.99492 0.99506 0.99520
2.6 0.99534 0.99547 0.99560 0.99573 0.99585 0.99598 0.99609 0.99621 0.99632 0.99643
2.7 0.99653 0.99664 0.99674 0.99683 0.99693 0.99702 0.99711 0.99720 0.99728 0.99736
2.8 0.99744 0.99752 0.99760 0.99767 0.99774 0.99781 0.99788 0.99795 0.99801 0.99807
2.9 0.99813 0.99819 0.99825 0.99831 0.99836 0.99841 0.99846 0.99851 0.99856 0.99861
3.0 0.99865 0.99869 0.99874 0.99878 0.99882 0.99886 0.99889 0.99893 0.99896 0.99900
3.1 0.99903 0.99906 0.99910 0.99913 0.99916 0.99918 0.99921 0.99924 0.99926 0.99929
3.2 0.99931 0.99934 0.99936 0.99938 0.99940 0.99942 0.99944 0.99946 0.99948 0.99950
3.3 0.99952 0.99953 0.99955 0.99957 0.99958 0.99960 0.99961 0.99962 0.99964 0.99965
3.4 0.99966 0.99968 0.99969 0.99970 0.99971 0.99972 0.99973 0.99974 0.99975 0.99976
3.5 0.99977 0.99978 0.99978 0.99979 0.99980 0.99981 0.99981 0.99982 0.99983 0.99983
202 A. TABLES
Student's t-distribution

α
The table gives the values of tα; ν where
Pr(Tν > tα; ν ) = α , with ν degrees of freedom t α; ν
α
0.1 0.05 0.025 0.01 0.005 0.001 0.0005
ν
1 3.078 6.314 12.076 31.821 63.657 318.310 636.620
2 1.886 2.920 4.303 6.965 9.925 22.326 31.598
3 1.638 2.353 3.182 4.541 5.841 10.213 12.924
4 1.533 2.132 2.776 3.747 4.604 7.173 8.610
5 1.476 2.015 2.571 3.365 4.032 5.893 6.869
6 1.440 1.943 2.447 3.143 3.707 5.208 5.959
7 1.415 1.895 2.365 2.998 3.499 4.785 5.408
8 1.397 1.860 2.306 2.896 3.355 4.501 5.041
9 1.383 1.833 2.262 2.821 3.250 4.297 4.781
10 1.372 1.812 2.228 2.764 3.169 4.144 4.587
11 1.363 1.796 2.201 2.718 3.106 4.025 4.437
12 1.356 1.782 2.179 2.681 3.055 3.930 4.318
13 1.350 1.771 2.160 2.650 3.012 3.852 4.221
14 1.345 1.761 2.145 2.624 2.977 3.787 4.140
15 1.341 1.753 2.131 2.602 2.947 3.733 4.073
16 1.337 1.746 2.120 2.583 2.921 3.686 4.015
17 1.333 1.740 2.110 2.567 2.898 3.646 3.965
18 1.330 1.734 2.101 2.552 2.878 3.610 3.922
19 1.328 1.729 2.093 2.539 2.861 3.579 3.883
20 1.325 1.725 2.086 2.528 2.845 3.552 3.850
21 1.323 1.721 2.080 2.518 2.831 3.527 3.819
22 1.321 1.717 2.074 2.508 2.819 3.505 3.792
23 1.319 1.714 2.069 2.500 2.807 3.485 3.767
24 1.318 1.711 2.064 2.492 2.797 3.467 3.745
25 1.316 1.708 2.060 2.485 2.787 3.450 3.725
26 1.315 1.706 2.056 2.479 2.779 3.435 3.707
27 1.314 1.703 2.052 2.473 2.771 3.421 3.690
28 1.313 1.701 2.048 2.467 2.763 3.408 3.674
29 1.311 1.699 2.045 2.462 2.756 3.396 3.659
30 1.310 1.697 2.042 2.457 2.750 3.385 3.646
40 1.303 1.684 2.021 2.423 2.704 3.307 3.551
60 1.296 1.671 2.000 2.390 2.660 3.232 3.460
120 1.289 1.658 1.980 2.358 2.617 3.160 3.373
∞ 1.282 1.645 1.960 2.326 2.576 3.090 3.291
 A. TABLES 203
F-distribution (Upper tail probability = 0.05) Numerator df = 1 to 10
df2\df1 1 2 3 4 5 6 7 8 10
1 161.448 199.500 215.707 224.583 230.162 233.986 236.768 238.883 241.882
2 18.513 19.000 19.164 19.247 19.296 19.330 19.353 19.371 19.396
3 10.128 9.552 9.277 9.117 9.013 8.941 8.887 8.845 8.786
4 7.709 6.944 6.591 6.388 6.256 6.163 6.094 6.041 5.964
5 6.608 5.786 5.409 5.192 5.050 4.950 4.876 4.818 4.735
6 5.987 5.143 4.757 4.534 4.387 4.284 4.207 4.147 4.060
7 5.591 4.737 4.347 4.120 3.972 3.866 3.787 3.726 3.637
8 5.318 4.459 4.066 3.838 3.687 3.581 3.500 3.438 3.347
9 5.117 4.256 3.863 3.633 3.482 3.374 3.293 3.230 3.137
10 4.965 4.103 3.708 3.478 3.326 3.217 3.135 3.072 2.978
11 4.844 3.982 3.587 3.357 3.204 3.095 3.012 2.948 2.854
12 4.747 3.885 3.490 3.259 3.106 2.996 2.913 2.849 2.753
13 4.667 3.806 3.411 3.179 3.025 2.915 2.832 2.767 2.671
14 4.600 3.739 3.344 3.112 2.958 2.848 2.764 2.699 2.602
15 4.543 3.682 3.287 3.056 2.901 2.790 2.707 2.641 2.544
16 4.494 3.634 3.239 3.007 2.852 2.741 2.657 2.591 2.494
17 4.451 3.592 3.197 2.965 2.810 2.699 2.614 2.548 2.450
18 4.414 3.555 3.160 2.928 2.773 2.661 2.577 2.510 2.412
19 4.381 3.522 3.127 2.895 2.740 2.628 2.544 2.477 2.378
20 4.351 3.493 3.098 2.866 2.711 2.599 2.514 2.447 2.348
21 4.325 3.467 3.072 2.840 2.685 2.573 2.488 2.420 2.321
22 4.301 3.443 3.049 2.817 2.661 2.549 2.464 2.397 2.297
23 4.279 3.422 3.028 2.796 2.640 2.528 2.442 2.375 2.275
24 4.260 3.403 3.009 2.776 2.621 2.508 2.423 2.355 2.255
25 4.242 3.385 2.991 2.759 2.603 2.490 2.405 2.337 2.236
26 4.225 3.369 2.975 2.743 2.587 2.474 2.388 2.321 2.220
27 4.210 3.354 2.960 2.728 2.572 2.459 2.373 2.305 2.204
28 4.196 3.340 2.947 2.714 2.558 2.445 2.359 2.291 2.190
29 4.183 3.328 2.934 2.701 2.545 2.432 2.346 2.278 2.177
30 4.171 3.316 2.922 2.690 2.534 2.421 2.334 2.266 2.165
35 4.121 3.267 2.874 2.641 2.485 2.372 2.285 2.217 2.114
40 4.085 3.232 2.839 2.606 2.449 2.336 2.249 2.180 2.077
45 4.057 3.204 2.812 2.579 2.422 2.308 2.221 2.152 2.049
50 4.034 3.183 2.790 2.557 2.400 2.286 2.199 2.130 2.026
55 4.016 3.165 2.773 2.540 2.383 2.269 2.181 2.112 2.008
60 4.001 3.150 2.758 2.525 2.368 2.254 2.167 2.097 1.993
70 3.978 3.128 2.736 2.503 2.346 2.231 2.143 2.074 1.969
80 3.960 3.111 2.719 2.486 2.329 2.214 2.126 2.056 1.951
90 3.947 3.098 2.706 2.473 2.316 2.201 2.113 2.043 1.938
100 3.936 3.087 2.696 2.463 2.305 2.191 2.103 2.032 1.927
110 3.927 3.079 2.687 2.454 2.297 2.182 2.094 2.024 1.918
120 3.920 3.072 2.680 2.447 2.290 2.175 2.087 2.016 1.910
130 3.914 3.066 2.674 2.441 2.284 2.169 2.081 2.010 1.904
140 3.909 3.061 2.669 2.436 2.279 2.164 2.076 2.005 1.899
150 3.904 3.056 2.665 2.432 2.274 2.160 2.071 2.001 1.894
160 3.900 3.053 2.661 2.428 2.271 2.156 2.067 1.997 1.890
180 3.894 3.046 2.655 2.422 2.264 2.149 2.061 1.990 1.884
200 3.888 3.041 2.650 2.417 2.259 2.144 2.056 1.985 1.878
220 3.884 3.037 2.646 2.413 2.255 2.140 2.051 1.981 1.874
240 3.880 3.033 2.642 2.409 2.252 2.136 2.048 1.977 1.870
260 3.877 3.031 2.639 2.406 2.249 2.134 2.045 1.974 1.867
280 3.875 3.028 2.637 2.404 2.246 2.131 2.042 1.972 1.865
300 3.873 3.026 2.635 2.402 2.244 2.129 2.040 1.969 1.862
400 3.865 3.018 2.627 2.394 2.237 2.121 2.032 1.962 1.854
500 3.860 3.014 2.623 2.390 2.232 2.117 2.028 1.957 1.850
600 3.857 3.011 2.620 2.387 2.229 2.114 2.025 1.954 1.846
700 3.855 3.009 2.618 2.385 2.227 2.112 2.023 1.952 1.844
800 3.853 3.007 2.616 2.383 2.225 2.110 2.021 1.950 1.843
900 3.852 3.006 2.615 2.382 2.224 2.109 2.020 1.949 1.841
1000 3.851 3.005 2.614 2.381 2.223 2.108 2.019 1.948 1.840
∞ 3.841 2.996 2.605 2.372 2.214 2.099 2.010 1.938 1.831
204 A. TABLES
F-distribution (Upper tail probability = 0.05) Numerator df = 12 to 40
df2\df1 12 14 16 18 20 24 28 32 36 40
1 243.906 245.364 246.464 247.323 248.013 249.052 249.797 250.357 250.793 251.143
2 19.413 19.424 19.433 19.440 19.446 19.454 19.460 19.464 19.468 19.471
3 8.745 8.715 8.692 8.675 8.660 8.639 8.623 8.611 8.602 8.594
4 5.912 5.873 5.844 5.821 5.803 5.774 5.754 5.739 5.727 5.717
5 4.678 4.636 4.604 4.579 4.558 4.527 4.505 4.488 4.474 4.464
6 4.000 3.956 3.922 3.896 3.874 3.841 3.818 3.800 3.786 3.774
7 3.575 3.529 3.494 3.467 3.445 3.410 3.386 3.367 3.352 3.340
8 3.284 3.237 3.202 3.173 3.150 3.115 3.090 3.070 3.055 3.043
9 3.073 3.025 2.989 2.960 2.936 2.900 2.874 2.854 2.839 2.826
10 2.913 2.865 2.828 2.798 2.774 2.737 2.710 2.690 2.674 2.661
11 2.788 2.739 2.701 2.671 2.646 2.609 2.582 2.561 2.544 2.531
12 2.687 2.637 2.599 2.568 2.544 2.505 2.478 2.456 2.439 2.426
13 2.604 2.554 2.515 2.484 2.459 2.420 2.392 2.370 2.353 2.339
14 2.534 2.484 2.445 2.413 2.388 2.349 2.320 2.298 2.280 2.266
15 2.475 2.424 2.385 2.353 2.328 2.288 2.259 2.236 2.219 2.204
16 2.425 2.373 2.333 2.302 2.276 2.235 2.206 2.183 2.165 2.151
17 2.381 2.329 2.289 2.257 2.230 2.190 2.160 2.137 2.119 2.104
18 2.342 2.290 2.250 2.217 2.191 2.150 2.119 2.096 2.078 2.063
19 2.308 2.256 2.215 2.182 2.155 2.114 2.084 2.060 2.042 2.026
20 2.278 2.225 2.184 2.151 2.124 2.082 2.052 2.028 2.009 1.994
21 2.250 2.197 2.156 2.123 2.096 2.054 2.023 1.999 1.980 1.965
22 2.226 2.173 2.131 2.098 2.071 2.028 1.997 1.973 1.954 1.938
23 2.204 2.150 2.109 2.075 2.048 2.005 1.973 1.949 1.930 1.914
24 2.183 2.130 2.088 2.054 2.027 1.984 1.952 1.927 1.908 1.892
25 2.165 2.111 2.069 2.035 2.007 1.964 1.932 1.908 1.888 1.872
26 2.148 2.094 2.052 2.018 1.990 1.946 1.914 1.889 1.869 1.853
27 2.132 2.078 2.036 2.002 1.974 1.930 1.898 1.872 1.852 1.836
28 2.118 2.064 2.021 1.987 1.959 1.915 1.882 1.857 1.837 1.820
29 2.104 2.050 2.007 1.973 1.945 1.901 1.868 1.842 1.822 1.806
30 2.092 2.037 1.995 1.960 1.932 1.887 1.854 1.829 1.808 1.792
35 2.041 1.986 1.942 1.907 1.878 1.833 1.799 1.773 1.752 1.735
40 2.003 1.948 1.904 1.868 1.839 1.793 1.759 1.732 1.710 1.693
45 1.974 1.918 1.874 1.838 1.808 1.762 1.727 1.700 1.678 1.660
50 1.952 1.895 1.850 1.814 1.784 1.737 1.702 1.674 1.652 1.634
55 1.933 1.876 1.831 1.795 1.764 1.717 1.681 1.653 1.631 1.612
60 1.917 1.860 1.815 1.778 1.748 1.700 1.664 1.636 1.613 1.594
70 1.893 1.836 1.790 1.753 1.722 1.674 1.637 1.608 1.585 1.566
80 1.875 1.817 1.772 1.734 1.703 1.654 1.617 1.588 1.564 1.545
90 1.861 1.803 1.757 1.720 1.688 1.639 1.601 1.572 1.548 1.528
100 1.850 1.792 1.746 1.708 1.676 1.627 1.589 1.559 1.535 1.515
110 1.841 1.783 1.736 1.698 1.667 1.617 1.579 1.549 1.524 1.504
120 1.834 1.775 1.728 1.690 1.659 1.608 1.570 1.540 1.516 1.495
130 1.827 1.769 1.722 1.684 1.652 1.601 1.563 1.533 1.508 1.488
140 1.822 1.763 1.716 1.678 1.646 1.595 1.557 1.526 1.502 1.481
150 1.817 1.758 1.711 1.673 1.641 1.590 1.552 1.521 1.496 1.475
160 1.813 1.754 1.707 1.669 1.637 1.586 1.547 1.516 1.491 1.470
180 1.806 1.747 1.700 1.661 1.629 1.578 1.539 1.508 1.483 1.462
200 1.801 1.742 1.694 1.656 1.623 1.572 1.533 1.502 1.476 1.455
220 1.796 1.737 1.690 1.651 1.618 1.567 1.528 1.496 1.471 1.450
240 1.793 1.733 1.686 1.647 1.614 1.563 1.523 1.492 1.466 1.445
260 1.790 1.730 1.683 1.644 1.611 1.559 1.520 1.488 1.463 1.441
280 1.787 1.727 1.680 1.641 1.608 1.556 1.517 1.485 1.459 1.438
300 1.785 1.725 1.677 1.638 1.606 1.554 1.514 1.482 1.456 1.435
400 1.776 1.717 1.669 1.630 1.597 1.545 1.505 1.473 1.447 1.425
500 1.772 1.712 1.664 1.625 1.592 1.539 1.499 1.467 1.441 1.419
600 1.768 1.708 1.660 1.621 1.588 1.536 1.495 1.463 1.437 1.414
700 1.766 1.706 1.658 1.619 1.586 1.533 1.492 1.460 1.434 1.412
800 1.764 1.704 1.656 1.617 1.584 1.531 1.490 1.458 1.432 1.409
900 1.763 1.703 1.655 1.615 1.582 1.529 1.489 1.457 1.430 1.408
1000 1.762 1.702 1.654 1.614 1.581 1.528 1.488 1.455 1.429 1.406
∞ 1.752 1.692 1.644 1.604 1.571 1.517 1.476 1.444 1.417 1.394
 A. TABLES 205
Cumulative Binomial Distribution - 1
206 A. TABLES
Cumulative Binomial Distribution - 2
 A. TABLES 207
Cumulative Binomial Distribution - 3
208 A. TABLES
Cumulative Binomial Distribution - 4
 A. TABLES 209
Cumulative Binomial Distribution - 5
210 A. TABLES
Cumulative Binomial Distribution - 6
 A. TABLES 211
Cumulative Binomial Distribution - 7
212 A. TABLES
Cumulative Binomial Distribution - 8
 213

Bibliography
[1] Robert I. Kobacoff, R in Action: Data Analysis and Graphics with R, 2nd ed., Manning
Publications, 2015.
[2] http://www.businessdictionary.com/definition/sampling-distribution.htm
l

[3] https://www.emathzone.com/tutorials/basic-statistics/simple-random-
sampling.html#ixzz5EhfEhbLUhttps://www.emathzone.com/tutorials/basic-
statistics/simple-random-sampling.html#ixzz5EhaygQsf

[4] https://www.britannica.com/science/statistics/Sample-survey-methods#re
f367539

[5] https://www.intmath.com/counting-probability/13-poisson-probability-
distribution.php#mean_

[6] https://www.umass.edu/wsp/resources/poisson/#end_
[7] https://www.isixsigma.com/tools-templates/control-charts/a-guide-to-
control-charts/

[8] https://businessjargons.com/chi-square-distribution.html
[9] https://www.statisticshowto.datasciencecentral.com/probability-and-
statistics/chi-square/
 215

Author’s Biography
MUSTAPHA AKINKUNMI
Dr. Mustapha Akinkunmi is a Financial Economist and
Technology Strategist. He has over 25 years of experi-
ence in estimation, planning, and forecasting using sta-
tistical and econometric methods, with particular expertise
in risk, expected utility, discounting, binomial-tree valua-
tion methods, financial econometrics models, Monte Carlo
simulations, macroeconomics, and exchange rate modeling.
Dr. Akinkunmi has performed extensive software develop-
ment for quantitative analysis of capital markets, revenue and
payment gateway, predictive analytics, data science, and credit
risk management.
He has a record of success in identifying and implement-
ing change management programs and institutional develop-
ment initiatives in both public and private sector organizations. He has been in high profile po-
sitions as a Consultant, Financial Advisor, Project Manager, and Business Strategist to AT&T,
Salomon Brothers, Goldman Sachs, Phibro Energy, First Boston (Credit Suisse First Boston),
World Bank, and Central Bank of Nigeria. He is an internationally recognized co-author (In-
troduction to Strategic Financial Management, May 2013) and leader in demand analysis, special-
izing in working with very large databases. Furthermore, he has conducted teaching and applied
research in areas that include analyses of expenditure patterns, inflation and exchange rate mod-
eling for Manhattan College, Riverdale, NY, Fordham University, New York, NY, University
of Lagos, Lagos, Nigeria, State University of New York-FIT, New York, NY, Montclair State
University, Montclair, NJ, and American University, Yola, Nigeria.
In 1990, he founded Technology Solutions Incorporated (TSI) in New York, which fo-
cused on data science and software application development for clients including major fi-
nancial services institutions. After ten years of successful operations and rapid growth under
Dr. Akinkunmi’s leadership, TSI was acquired by a publicly traded technology company based
in the U.S. in a value-creating transaction. Dr. Akinkunmi was the former Honorable Com-
missioner for Finance, Lagos State, Nigeria. He is now an Associate Professor of Finance and
Chair of the Accounting and Finance Department at the American University of Nigeria, Yola,
Nigeria.