# SPSS Survival Manual 4th Edition A Step By Guide To Data Analysis Using The Program 2010

### SPSS_Survival_Manual_4th_Edition_0335242391_ manual pdf -FilePursuit

User Manual: manual pdf -FilePursuit

Open the PDF directly: View PDF .

Page Count: 359 [warning: Documents this large are best viewed by clicking the View PDF Link!]

- Title Page
- Contents
- Preface
- Data files and website
- Introduction and overview
- Part One: Getting started
- Part Two: Preparing the data file
- Part Three: Preliminary analyses
- Part Four: Statistical techniques to explore relationships among variables
- Part Five: Statistical techniques to compare groups
- Appendix: Details of data files
- Recommended reading
- References
- Index

For the SPSS Survival Manual website, go to www.allenandunwin.com/spss

This is what readers from around the world say about the SPSS Survival Manual:

‘Best book ever written. My ability to work the maze of statistics and my sanity has been SAVED by this

book.’

Natasha Davison, Doctorate of Health Psychology, Deakin University, Australia

‘I just wanted to say how much I value Julie Pallant’s SPSS Survival Manual. It’s quite the best text on

SPSS I’ve encountered and I recommend it to anyone who’s listening!’

Professor Carolyn Hicks, Health Sciences, Birmingham University, UK

‘This book was responsible for an A on our educational research project. This is the perfect book for

people who are bafﬂ ed by statistical analysis, but still have to understand and accomplish it.’

Becky, Houston, Texas, USA

‘Truly a survival manual. This was highly recommended to me and was well worth it. I had no difﬁ culty

following the steps as they were so well laid out and included screen shots. This book takes the majority

of the anxiety out of statistical analysis.’

C. Wright, amazon.com

‘Having perceived myself as one who was not conﬁ dent in anything statistical, I worked my way through

the book and with each turn of the page gained more and more conﬁ dence until I was running off

analyses with (almost) glee. I now enjoy using SPSS and this book is the reason for that.’

Dr Marina Harvey, Centre for Professional Development, Macquarie University, Australia

‘I have had several courses in advanced statistics, but unfortunately none of them went too “in depth”

into SPSS. This book does just that, in a clear “how to” format that gets right to the point and tells you

what you need to know.’

John Ryan, Atlanta, Georgia, USA

‘This book really lives up to its name . . . I highly recommend this book to any MBA student carrying

out a dissertation project, or anyone who needs some basic help with using SPSS and data analysis

techniques.’

Business student, UK

‘I must say how much I value SPSS Survival Manual. It is so clearly written and helpful. I ﬁ nd myself

using it constantly and also ask any students doing a thesis or dissertation to obtain a copy.’

Associate Professor Sheri Bauman, Department of Educational Psychology, University of Arizona, USA

‘This book is simple to understand, easy to read and very concise. Those who have a general fear or dislike

for statistics or statistics and computers should enjoy reading this book.’

Lloyd G. Waller PhD, Jamaica

‘There are several SPSS manuals published and this one really does “do what it says on the tin” . . . Whether

you are a beginner doing your BSc or struggling with your PhD research (or beyond!), I wholeheartedly

recommend this book.’

British Journal of Occupational Therapy, UK

‘I love SPSS Survival Manual . . . I can’t imagine teaching without it. After seeing my copy and hearing me

talk about it many of my other colleagues are also utilising it.’

Wendy Close PhD, Psychology Department, Wisconsin Lutheran College, USA

‘. . . being an external student so much of the time is spent teaching myself. But this has been made easier

with your manual as I have found much of the content very easy to follow. I only wish I had discovered

it earlier.’

Anthropology student, Australia

‘This book is a “must have” introduction to SPSS. Brilliant and highly recommended.’

Dr Joe, South Africa

‘The strength of this book lies in the explanations that accompany the descriptions of tests and I predict

great popularity for this text among teachers, lecturers and researchers.’

Roger Watson, Journal of Advanced Nursing

‘This is the one. If you need to do statistics for a thesis, dissertation, course, etc. but aren’t quite sure

where to start or what to do, this is the book you have been looking for. I don’t know how I would’ve

completed my dissertation without this book. EXTREMELY helpful and easy to understand without

being “dumbed down”.’

Thomas A. Delaney, Eugene, Oregon, USA

‘This book is the absolute bible for SPSS users and the book’s cover picture says it all—a true life saver.

Without this book I would not be graduating with a doctoral degree.’

A. Preston, Hawaii

‘Pallant’s excellent book has all the ingredients to take interested students,including the statistically naive

and the algebraically challenged, to a new level of skill and understanding.’

Geoffrey N. Molloy, Behaviour Change journal

‘I have four SPSS manuals and have found that this is the only manual that explains the issues clearly

and is easy to follow. SPSS is evil and anything that makes it less so is fabulous.’

Helen Scott, Psychology Honours Student, University of Queensland, Australia

‘To any students who have found themselves faced with the horror of SPSS when they had signed up for

a degree in psychology—this is a god send.’

Psychology student, Ireland

‘This is the best SPSS manual I’ve had. It’s comprehensive and easy to follow. I really enjoy it.’

Norshidah Mohamed, Kuala Lumpur, Malaysia

‘Julie Pallant saved my life with this book. OK, slight exaggeration but this book really is a life saver . . . If

the mere thought of statistics gives you a headache, then this book is for you.’

Statistics student, UK

‘Simply the best book on introductory SPSS that exists. I know nothing about the author but having

bought this book in the middle of a statistics open assignment I can conﬁ dently say that I love her and

want to marry her. There must be dozens of books that claim to be beginners’ guides to SPSS. This one

actually does what it says: totally brilliant.’

J. Sutherland, amazon.co.uk

SPSS

SURVIVAL MANUAL

A step by step guide to

data analysis using SPSS

4th edition

Julie Pallant

This fourth edition ﬁ rst published in 2011

Copyright © Julie Pallant 2002, 2005, 2007, 2011

All rights reserved. No part of this book may be reproduced or transmitted in any form or by any

means, electronic or mechanical, including photocopying, recording or by any information storage

and retrieval system, without prior permission in writing from the publisher. The Australian Copyright

Act 1968 (the Act) allows a maximum of one chapter or 10 per cent of this book, whichever is the

greater, to be photocopied by any educational institution for its educational purposes provided that

the educational institution (or body that administers it) has given a remuneration notice to Copyright

Agency Limited (CAL) under the Act.

Allen & Unwin

83 Alexander Street

Crows Nest NSW 2065

Australia

Phone: (61 2) 8425 0100

Fax: (61 2) 9906 2218

Email: info@allenandunwin.com

Web: www.allenandunwin.com

Cataloguing-in-Publication details are available

from the National Library of Australia

www.librariesaustralia.nla.gov.au

ISBN 978 1 74237 392 8

Set in 11/13.5 pt Minion by Midland Typesetters, Australia

Printed in China at Everbest Printing Co

10 9 8 7 6 5 4 3 2 1

Contents

Preface vii

Data ﬁ les and website viii

Introduction and overview x

Part One Getting started 1

1 Designing a study 3

2 Preparing a codebook 11

3 Getting to know SPSS 14

Part Two Preparing the data ﬁ le 25

4 Creating a data ﬁ le and entering data 27

5 Screening and cleaning the data 43

Part Three Preliminary analyses 51

6 Descriptive statistics 53

7 Using graphs to describe and explore the data 66

8 Manipulating the data 83

9 Checking the reliability of a scale 97

10 Choosing the right statistic 102

Part Four Statistical techniques to explore relationships among variables 121

11 Correlation 128

12 Partial correlation 143

13 Multiple regression 148

14 Logistic regression 168

15 Factor analysis 181

Part Five Statistical techniques to compare groups 203

16 Non-parametric statistics 213

17 T-tests 239

vi Contents

18 One-way analysis of variance 249

19 Two-way between-groups ANOVA 265

20 Mixed between-within subjects analysis of variance 274

21 Multivariate analysis of variance 283

22 Analysis of covariance 297

Appendix: Details of data ﬁ les 319

Recommended reading 334

References 337

Index 341

Preface

For many students, the thought of completing a statistics subject, or using statistics

in their research, is a major source of stress and frustration. The aim of the original

SPSS Survival Manual (published in 2000) was to provide a simple, step-by-step guide

to the process of data analysis using SPSS. Unlike other statistical titles it did not

focus on the mathematical underpinnings of the techniques, but rather on the appro-

priate use of SPSS as a tool. Since the publication of the three editions of the SPSS

Survival Manual, I have received many hundreds of emails from students who have

been grateful for the helping hand (or lifeline).

The same simple approach has been incorporated in this fourth edition. Since

the last edition, however, SPSS has undergone a number of changes—including

a brief period when it changed name. During 2009 version 18 of the program was

renamed PASW Statistics, which stands for Predictive Analytics Software. The name

was changed again in 2010 to IBM SPSS. To prevent confusion I have referred to

the program as SPSS throughout the book, but all the material applies to programs

labelled both PASW and IBM SPSS. All chapters in this edition have been updated to

suit version 18 of the package (although most of the material is also suitable for users

of earlier versions).

I have resisted urges from students, instructors and reviewers to add too many

extra topics, but instead have upgraded and expanded the existing material. This

book is not intended to cover all possible statistical procedures available in SPSS, or

to answer all questions researchers might have about statistics. Instead, it is designed

to get you started with your research and to help you gain conﬁ dence in the use of the

program to analyse your data. There are many other excellent statistical texts avail-

able that you should refer to—suggestions are made throughout each chapter in the

book. Additional material is also available on the book’s website (details in the next

section).

vii

Data ﬁ les and website

Throughout the book, you will see examples of research that are taken from a number

of data ﬁ les included on the website that accompanies this book. This website is at:

www.allenandunwin.com/spss

From this site you can download the data ﬁ les to your hard drive or memory stick

by following the instructions on screen. Then you should start SPSS and open the data

ﬁ les. These ﬁ les can be opened only in SPSS.

The survey4ED.sav data ﬁ le is a ‘real’ data ﬁ le, based on a research project that

was conducted by one of my graduate diploma classes. So that you can get a feel for

the research process from start to ﬁ nish, I have also included in the Appendix a copy

of the questionnaire that was used to generate this data and the codebook used to

code the data. This will allow you to follow along with the analyses that are presented

in the book, and to experiment further using other variables.

The second data ﬁ le (error4ED.sav) is the same ﬁ le as the survey4ED.sav, but I

have deliberately added some errors to give you practice in Chapter 5 at screening and

cleaning your data ﬁ le.

The third data ﬁ le (experim4ED.sav) is a manufactured (fake) data ﬁ le, constructed

and manipulated to illustrate the use of a number of techniques covered in Part Five

of the book (e.g. Paired Samples t-test, Repeated Measures ANOVA). This ﬁ le also

includes additional variables that will allow you to practise the skills learnt through-

out the book. Just don’t get too excited about the results you obtain and attempt to

replicate them in your own research!

The fourth ﬁ le used in the examples in the book is depress4ED.sav. This is used

in Chapter 16, on non-parametric techniques, to illustrate some techniques used in

health and medical research.

Two other data ﬁ les have been included, giving you the opportunity to complete

some additional activities with data from different discipline areas. The sleep4ED.sav

ﬁ le is a real data ﬁ le from a study conducted to explore the prevalence and impact of

sleep problems on aspects of people’s lives. The staffsurvey4ED.sav ﬁ le comes from a

staff satisfaction survey conducted for a large national educational institution.

viii

Data ﬁ les and websites ix

See the Appendix for further details of these ﬁ les (and associated materials). Apart

from the data ﬁ les, the SPSS Survival Manual website also contains a number of useful

items for students and instructors, including:

• guidelines for preparing a research report

• practice exercises

• updates on changes to SPSS as new versions are released

• useful links to other websites

• additional reading

• an instructor’s guide.

Introduction and overview

This book is designed for students completing research design and statistics courses

and for those involved in planning and executing research of their own. Hopefully this

guide will give you the conﬁ dence to tackle statistical analyses calmly and sensibly, or

at least without too much stress!

Many of the problems that students experience with statistical analysis are due to

anxiety and confusion from dealing with strange jargon, complex underlying theories

and too many choices. Unfortunately, most statistics courses and textbooks encourage

both of these sensations! In this book I try to translate statistics into a language that

can be more easily understood and digested.

The SPSS Survival Manual is presented in a structured format, setting out step

by step what you need to do to prepare and analyse your data. Think of your data as

the raw ingredients in a recipe. You can choose to cook your ‘ingredients’ in different

ways—a ﬁ rst course, main course, dessert. Depending on what ingredients you have

available, different options may, or may not, be suitable. (There is no point planning

to make beef stroganoff if all you have is chicken.) Planning and preparation are an

important part of the process (both in cooking and in data analysis). Some things you

will need to consider are:

• Do you have the correct ingredients in the right amounts?

• What preparation is needed to get the ingredients ready to cook?

• What type of cooking approach will you use (boil, bake, stir-fry)?

• Do you have a picture in your mind of how the end result (e.g. chocolate cake) is

supposed to look?

• How will you tell when it is cooked?

• Once it is cooked, how should you serve it so that it looks appetising?

The same questions apply equally well to the process of analysing your data. You

must plan your experiment or survey so that it provides the information you need,

in the correct format. You must prepare your data ﬁ le properly and enter your

data carefully. You should have a clear idea of your research questions and how

x

Introduction and overview xi

you might go about addressing them. You need to know what statistical techniques

are available, what sort of variables are suitable and what are not. You must be

able to perform your chosen statistical technique (e.g. t-test) correctly and interpret

the output. Finally, you need to relate this ‘output’ back to your original research

question and know how to present this in your report (or in cooking terms, should

you serve your chocolate cake with cream or ice-cream, or perhaps some berries and

a sprinkle of icing sugar on top?).

In both cooking and data analysis, you can’t just throw in all your ingredients

together, shove it in the oven (or SPSS, as the case may be) and hope for the best.

Hopefully this book will help you understand the data analysis process a little better

and give you the conﬁ dence and skills to be a better ‘cook’.

STRUCTURE OF THIS BOOK

This SPSS Survival Manual consists of 22 chapters, covering the research process from

designing a study through to the analysis of the data and presentation of the results.

It is broken into ﬁ ve main parts. Part One (Getting started) covers the preliminar-

ies: designing a study, preparing a codebook and becoming familiar with SPSS. In

Part Two (Preparing the data ﬁ le) you will be shown how to prepare a data ﬁ le, enter

your data and check for errors. Preliminary analyses are covered in Part Three, which

includes chapters on the use of descriptive statistics and graphs; the manipulation of

data; and the procedures for checking the reliability of scales. You will also be guided,

step by step, through the sometimes difﬁ cult task of choosing which statistical tech-

nique is suitable for your data.

In Part Four the major statistical techniques that can be used to explore relation-

ships are presented (e.g. correlation, partial correlation, multiple regression, logistic

regression and factor analysis). These chapters summarise the purpose of each tech-

nique, the underlying assumptions, how to obtain results, how to interpret the output,

and how to present these results in your thesis or report.

Part Five discusses the statistical techniques that can be used to compare groups.

These include non-parametric techniques, t-tests, analysis of variance, multivariate

analysis of variance and analysis of covariance.

USING THIS BOOK

To use this book effectively as a guide to SPSS, you need some basic computer skills.

In the instructions and examples provided throughout the text I assume that you are

already familiar with using a personal computer, particularly the Windows functions.

I have listed below some of the skills you will need. Seek help if you have difﬁ culty

with any of these operations. You will need to be able to:

xii Introduction and overview

• use the Windows drop-down menus

• use the left and right buttons on the mouse

• use the click and drag technique for highlighting text

• minimise and maximise windows

• start and exit programs from the Start menu or from Windows Explorer

• move between programs that are running simultaneously

• open, save, rename, move and close ﬁ les

• work with more than one ﬁ le at a time, and move between ﬁ les that are open

• use Windows Explorer to copy ﬁ les from a memory stick to the hard drive, and

back again

• use Windows Explorer to create folders and to move ﬁ les between folders.

This book is not designed to ‘stand alone’. It is assumed that you have been exposed to

the fundamentals of statistics and have access to a statistics text. It is important that

you understand some of what goes on ‘below the surface’ when using SPSS. SPSS is

an enormously powerful data analysis package that can handle very complex statis-

tical procedures. This manual does not attempt to cover all the different statistical

techniques available in the program. Only the most commonly used statistics are

covered. It is designed to get you started and to develop your conﬁ dence in using the

program.

Depending on your research questions and your data, it may be necessary to tackle

some of the more complex analyses available in SPSS. There are many good books

available covering the various statistical techniques in more detail. Read as widely as

you can. Browse the shelves in your library, look for books that explain statistics in a

language that you understand (well, at least some of it anyway!). Collect this material

together to form a resource to be used throughout your statistics classes and your

research project. It is also useful to collect examples of journal articles where statisti-

cal analyses are explained and results are presented. You can use these as models for

your ﬁ nal write-up.

The SPSS Survival Manual is suitable for use as both an in-class text, where you

have an instructor taking you through the various aspects of the research process,

and as a self-instruction book for those conducting an individual research project.

If you are teaching yourself, be sure to actually practise using SPSS by analysing the

data that is included on the website accompanying this book (see p. viii for details).

The best way to learn is by actually doing, rather than just reading. ‘Play’ with the data

ﬁ les from which the examples in the book are taken before you start using your own

data ﬁ le. This will improve your conﬁ dence and also allow you to check that you are

performing the analyses correctly.

Sometimes you may ﬁ nd that the output you obtain is different from that presented

in the book. This is likely to occur if you are using a different version of SPSS from that

Introduction and overview xiii

used throughout this book (SPSS Statistics 18). SPSS regularly updates its products,

which is great in terms of improving the program but it can lead to confusion for

students who ﬁ nd that what is on the screen differs from what is in the book. Usually

the difference is not too dramatic, so stay calm and play detective. The information

may be there, but just in a different form. For information on changes to the SPSS

products you may like to go to the SPSS website (www.spss.com).

RESEARCH TIPS

If you are using this book to guide you through your own research project, there are a

few additional tips I would like to recommend.

• Plan your project carefully. Draw on existing theories and research to guide the

design of your project. Know what you are trying to achieve and why.

• Think ahead. Anticipate potential problems and hiccups—every project has them!

Know what statistics you intend to employ and use this information to guide the

formulation of data collection materials. Make sure that you will have the right

sort of data to use when you are ready to do your statistical analyses.

• Get organised. Keep careful notes of all relevant research, references etc. Work out

an effective ﬁ ling system for the mountain of journal articles you will acquire and,

later on, the output from SPSS. It is easy to become disorganised, overwhelmed

and confused.

• Keep good records. When using SPSS to conduct your analyses, keep careful

records of what you do. I recommend to all my students that they buy a spiral-

bound exercise book to record every session they spend on SPSS. You should

record the date, new variables you create, all analyses you perform and the names

of the ﬁ les where you have saved the output. If you have a problem or something

goes horribly wrong with your data ﬁ le, this information can be used by your

supervisor to help rescue you!

• Stay calm! If this is your ﬁ rst exposure to SPSS and data analysis, there may be

times when you feel yourself becoming overwhelmed. Take some deep breaths

and use some positive self-talk. Just take things step by step—give yourself

permission to make mistakes and become confused sometimes. If it all gets too

much then stop, take a walk and clear your head before you tackle it again. Most

students ﬁ nd SPSS quite easy to use, once they get the hang of it. Like learning

any new skill, you just need to get past that ﬁ rst feeling of confusion and lack of

conﬁ dence.

• Give yourself plenty of time. The research process, particularly the data entry

and data analysis stages, always takes longer than expected, so allow plenty of time

for this.

• Work with a friend. Make use of other students for emotional and practical

support during the data analysis process. Social support is a great buffer against

stress!

ADDITIONAL RESOURCES

There are a number of different topic areas covered throughout this book, from

the initial design of a study, questionnaire construction, basic statistical techniques

(t-tests, correlation), through to advanced statistics (multivariate analysis of variance,

factor analysis). Further reading and resource material is recommended throughout

the different chapters in the book. You should try to read as broadly as you can, par-

ticularly if tackling some of the more complex statistical procedures.

xiv Introduction and overview

PART ONE

Getting started

Data analysis is only one part of the research process. Before you can use SPSS to

analyse your data, there are a number of things that need to happen. First, you have

to design your study and choose appropriate data collection instruments. Once you

have conducted your study, the information obtained must be prepared for entry into

SPSS (using something called a ‘codebook’). To enter the data you must understand

how SPSS works and how to talk to it appropriately. Each of these steps is discussed

in Part One.

Chapter 1 provides some tips and suggestions for designing a study, with the aim

of obtaining good-quality data. Chapter 2 covers the preparation of a codebook to

translate the information obtained from your study into a format suitable for SPSS.

Chapter 3 takes you on a guided tour of the program, and some of the basic skills that

you will need are discussed. If this is your ﬁ rst time using SPSS, it is important that

you read the material presented in Chapter 3 before attempting any of the analyses

presented later in the book.

1

This page intentionally left blank

3

1

Designing a study

Although it might seem a bit strange to discuss research design in a book on SPSS, it is

an essential part of the research process that has implications for the quality of the data

collected and analysed. The data you enter must come from somewhere—responses to

a questionnaire, information collected from interviews, coded observations of actual

behaviour, or objective measurements of output or performance. The data are only

as good as the instrument that you used to collect them and the research framework

that guided their collection.

In this chapter a number of aspects of the research process are discussed that have

an impact on the potential quality of the data. First, the overall design of the study

is considered; this is followed by a discussion of some of the issues to consider when

choosing scales and measures; and ﬁ nally, some guidelines for preparing a question-

naire are presented.

PLANNING THE STUDY

Good research depends on the careful planning and execution of the study. There

are many excellent books written on the topic of research design to help you

with this process—from a review of the literature, formulation of hypotheses,

choice of study design, selection and allocation of participants, recording of obser-

vations and collection of data. Decisions made at each of these stages can affect the

quality of the data you have to analyse and the way you address your research ques-

tions. In designing your own study I would recommend that you take your time

working through the design process to make it the best study that you can produce.

Reading a variety of texts on the topic will help. A few good, easy-to-follow titles

are Stangor (2006), Goodwin (2007) and, if you are working in the area of market

research, Boyce (2003). A good basic overview for health and medical research is

Peat (2001).

4 Getting Started

To get you started, consider these tips when designing your study:

• Consider what type of research design (e.g. experiment, survey, observation) is the

best way to address your research question. There are advantages and disadvan-

tages to all types of research approaches; choose the most appropriate approach

for your particular research question. Have a good understanding of the research

that has already been conducted in your topic area.

• If you choose to use an experiment, decide whether a between-groups design

(different cases in each experimental condition) or a repeated measures design

(same cases tested under all conditions) is the more appropriate for your research

question. There are advantages and disadvantages to each approach (see Stangor

2006), so weigh up each approach carefully.

• In experimental studies, make sure you include enough levels in your indepen-

dent variable. Using only two levels (or groups) means fewer participants are

required, but it limits the conclusions that you can draw. Is a control group necess-

ary or desirable? Will the lack of control group limit the conclusions that you

can draw?

• Always select more participants than you need, particularly if you are using a sample

of humans. People are notoriously unreliable—they don’t turn up when they are

supposed to, they get sick, drop out and don’t ﬁ ll out questionnaires properly! So

plan accordingly. Err on the side of pessimism rather than optimism.

• In experimental studies, check that you have enough participants in each of

your groups (and try to keep them equal when possible). With small groups, it is

difﬁ cult to detect statistically signiﬁ cant differences between groups (an issue of

power, discussed in the introduction to Part Five). There are calculations you can

perform to determine the sample size that you will need. See, for example, Stangor

(2006), or consult other statistical texts under the heading ‘power’.

• Wherever possible, randomly assign participants to each of your experimental

conditions, rather than using existing groups. This reduces the problem associated

with non-equivalent groups in between-groups designs. Also worth considering

is taking additional measurements of the groups to ensure that they don’t differ

substantially from one another. You may be able to statistically control for differ-

ences that you identify (e.g. using analysis of covariance).

• Choose appropriate dependent variables that are valid and reliable (see discussion

on this point later in this chapter). It is a good idea to include a number of differ-

ent measures—some measures are more sensitive than others. Don’t put all your

eggs in one basket.

• Try to anticipate the possible inﬂ uence of extraneous or confounding variables.

These are variables that could provide an alternative explanation for your results.

Sometimes they are hard to spot when you are immersed in designing the study

Designing a study 5

yourself. Always have someone else (supervisor, fellow researcher) check over

your design before conducting the study. Do whatever you can to control for these

potential confounding variables. Knowing your topic area well can also help you

identify possible confounding variables. If there are additional variables that you

cannot control, can you measure them? By measuring them, you may be able to

control for them statistically (e.g. using analysis of covariance).

• If you are distributing a survey, pilot-test it ﬁ rst to ensure that the instructions,

questions and scale items are clear. Wherever possible, pilot-test on the same type

of people who will be used in the main study (e.g. adolescents, unemployed youth,

prison inmates). You need to ensure that your respondents can understand the

survey or questionnaire items and respond appropriately. Pilot-testing should

also pick up any questions or items that may offend potential respondents.

• If you are conducting an experiment, it is a good idea to have a full dress rehearsal

and to pilot-test both the experimental manipulation and the dependent measures

you intend to use. If you are using equipment, make sure it works properly. If you

are using different experimenters or interviewers, make sure they are properly

trained and know what to do. If different observers are required to rate behaviours,

make sure they know how to appropriately code what they see. Have a practice run

and check for inter-rater reliability (i.e. how consistent scores are from different

raters). Pilot-testing of the procedures and measures helps you identify anything

that might go wrong on the day and any additional contaminating factors that

might inﬂ uence the results. Some of these you may not be able to predict (e.g.

workers doing noisy construction work just outside the lab’s window), but try to

control those factors that you can.

CHOOSING APPROPRIATE SCALES AND MEASURES

There are many different ways of collecting ‘data’, depending on the nature of your

research. This might involve measuring output or performance on some objective

criteria, or rating behaviour according to a set of speciﬁ ed criteria. It might also

involve the use of scales that have been designed to ‘operationalise’ some underly-

ing construct or attribute that is not directly measurable (e.g. self-esteem). There are

many thousands of validated scales that can be used in research. Finding the right one

for your purpose is sometimes difﬁ cult. A thorough review of the literature in your

topic area is the ﬁ rst place to start. What measures have been used by other research-

ers in the area? Sometimes the actual items that make up the scales are included in

the appendix to a journal article; otherwise you may need to trace back to the original

article describing the design and validation of the scale you are interested in. Some

scales have been copyrighted, meaning that to use them you need to purchase ‘ofﬁ cial’

copies from the publisher. Other scales, which have been published in their entirety

6 Getting Started

in journal articles, are considered to be ‘in the public domain’, meaning that they

can be used by researchers without charge. It is very important, however, to properly

acknowledge each of the scales you use, giving full reference details.

In choosing appropriate scales there are two characteristics that you need

to be aware of: reliability and validity. Both of these factors can inﬂ uence the

quality of the data you obtain. When reviewing possible scales to use, you should

collect information on the reliability and validity of each of the scales. You will

need this information for the ‘Method’ section of your research report. No matter

how good the reports are concerning the reliability and validity of your scales, it

is important to pilot-test them with your intended sample. Sometimes scales are

reliable with some groups (e.g. adults with an English-speaking background), but

are totally unreliable when used with other groups (e.g. children from non-English-

speaking backgrounds).

Reliability

The reliability of a scale indicates how free it is from random error. Two frequently

used indicators of a scale’s reliability are test-retest reliability (also referred to as

‘temporal stability’) and internal consistency. The test-retest reliability of a scale

is assessed by administering it to the same people on two different occasions, and

calculating the correlation between the two scores obtained. High test-retest corre-

lations indicate a more reliable scale. You need to take into account the nature of the

construct that the scale is measuring when considering this type of reliability. A scale

designed to measure current mood states is not likely to remain stable over a period

of a few weeks. The test-retest reliability of a mood scale, therefore, is likely to be low.

You would, however, hope that measures of stable personality characteristics would

stay much the same, showing quite high test-retest correlations.

The second aspect of reliability that can be assessed is internal consistency. This

is the degree to which the items that make up the scale are all measuring the same

underlying attribute (i.e. the extent to which the items ‘hang together’). Internal

consistency can be measured in a number of ways. The most commonly used statistic

is Cronbach’s coefﬁ cient alpha (available using SPSS, see Chapter 9). This statistic

provides an indication of the average correlation among all of the items that make up

the scale. Values range from 0 to 1, with higher values indicating greater reliability.

While different levels of reliability are required, depending on the nature and

purpose of the scale, Nunnally (1978) recommends a minimum level of .7. Cronbach

alpha values are dependent on the number of items in the scale. When there are a

small number of items in the scale (fewer than 10), Cronbach alpha values can be

quite small. In this situation it may be better to calculate and report the mean inter-

item correlation for the items. Optimal mean inter-item correlation values range from

.2 to .4 (as recommended by Briggs & Cheek 1986).

Designing a study 7

Validity

The validity of a scale refers to the degree to which it measures what it is supposed to

measure. Unfortunately, there is no one clear-cut indicator of a scale’s validity. The

validation of a scale involves the collection of empirical evidence concerning its use.

The main types of validity you will see discussed are content validity, criterion validity

and construct validity.

Content validity refers to the adequacy with which a measure or scale has sampled

from the intended universe or domain of content. Criterion validity concerns the

relationship between scale scores and some speciﬁ ed, measurable criterion. Construct

validity involves testing a scale not against a single criterion but in terms of theoretically

derived hypotheses concerning the nature of the underlying variable or construct. The

construct validity is explored by investigating its relationship with other constructs,

both related (convergent validity) and unrelated (discriminant validity). An easy-to-

follow summary of the various types of validity is provided in Stangor (2006) and in

Streiner and Norman (2008).

If you intend to use scales in your research, it would be a good idea to read further

on this topic: see Kline (2005) for information on psychological tests, and Streiner and

Norman (2008) for health measurement scales. Bowling also has some great books on

health and medical scales.

PREPARING A QUESTIONNAIRE

In many studies it is necessary to collect information from your participants or respon-

dents. This may involve obtaining demographic information from participants prior

to exposing them to some experimental manipulation. Alternatively, it may involve the

design of an extensive survey to be distributed to a selected sample of the population. A

poorly planned and designed questionnaire will not give good data with which to address

your research questions. In preparing a questionnaire, you must consider how you intend

to use the information; you must know what statistics you intend to use. Depending on

the statistical technique you have in mind, you may need to ask the question in a particular

way, or provide different response formats. Some of the factors you need to consider in the

design and construction of a questionnaire are outlined in the sections that follow.

This section only brieﬂ y skims the surface of questionnaire design, so I would

suggest that you read further on the topic if you are designing your own study. A really

great book for this purpose is De Vaus (2002) or, if your research area is business,

Boyce (2003).

Question types

Most questions can be classiﬁ ed into two groups: closed or open-ended. A closed

question involves offering respondents a number of deﬁ ned response choices. They are

8 Getting Started

asked to mark their response using a tick, cross, circle, etc. The choices may be a simple

Yes/No, Male/Female, or may involve a range of different choices. For example:

What is the highest level of education you have completed (please tick)?

❐ 1. Primary school

❐ 2. Some secondary school

❐ 3. Completed secondary school

❐ 4. Trade training

❐ 5. Undergraduate university

❐ 6. Postgraduate university

Closed questions are usually quite easy to convert to the numerical format required

for SPSS. For example, Yes can be coded as a 1, No can be coded as a 2; Males as 1,

Females as 2. In the education question shown above, the number corresponding to

the response ticked by the respondent would be entered. For example, if the respon-

dent ticked Undergraduate university, this would be coded as a 5. Numbering each of

the possible responses helps with the coding process. For data entry purposes, decide

on a convention for the numbering (e.g. in order across the page, and then down),

and stick with it throughout the questionnaire.

Sometimes you cannot guess all the possible responses that respondents might

make—it is therefore necessary to use open-ended questions. The advantage here is

that respondents have the freedom to respond in their own way, not restricted to the

choices provided by the researcher. For example:

What is the major source of stress in your life at the moment?

___________________________________________________________________

___________________________________________________________________

Responses to open-ended questions can be summarised into a number of different

categories for entry into SPSS. These categories are usually identiﬁ ed after looking

through the range of responses actually received from the respondents. Some possi-

bilities could also be raised from an understanding of previous research in the area.

Each of these response categories is assigned a number (e.g. work=1, ﬁ nances=2, rela-

tionships=3), and this number is entered into SPSS. More details on this are provided

in the section on preparing a codebook in Chapter 2.

Sometimes a combination of both closed and open-ended questions works best.

This involves providing respondents with a number of deﬁ ned responses, and also

an additional category (other) that they can tick if the response they wish to give is

not listed. A line or two is provided so that they can write the response they wish to

Designing a study 9

give. This combination of closed and open-ended questions is particularly useful in

the early stages of research in an area, as it gives an indication of whether the deﬁ ned

response categories adequately cover all the responses that respondents wish to give.

Response format

In asking respondents a question, you also need to decide on a response format. The

type of response format you choose can have implications when you come to do your

statistical analysis. Some analyses (e.g. correlation) require scores that are continuous,

from low through to high, with a wide range of scores. If you had asked respondents

to indicate their age by giving them a category to tick (e.g. less than 30, between 31

and 50 and over 50), these data would not be suitable to use in a correlational analysis.

So, if you intend to explore the correlation between age and, say, self-esteem, you

will need to ensure that you ask respondents for their actual age in years. Be warned

though, some people don’t like giving their exact age (e.g. women over 30!).

Try to provide as wide a choice of responses to your questions as possible. You can

always condense things later if you need to (see Chapter 8). Don’t just ask respondents

whether they agree or disagree with a statement—use a Likert-type scale, which can

range from strongly disagree to strongly agree:

strongly disagree 1 2 3 4 5 6 strongly agree

This type of response scale gives you a wider range of possible scores, and increases the

statistical analyses that are available to you. You will need to make a decision concern-

ing the number of response steps (e.g. 1 to 6) that you use. DeVellis (2003) has a good

discussion concerning the advantages and disadvantages of different response scales.

Whatever type of response format you choose, you must provide clear instructions. Do

you want your respondents to tick a box, circle a number, make a mark on a line? For

some respondents, this may be the ﬁ rst questionnaire that they have completed. Don’t

assume they know how to respond appropriately. Give clear instructions, provide an

example if appropriate, and always pilot-test on the type of people that will make up

your sample. Iron out any sources of confusion before distributing hundreds of your

questionnaires. In designing your questions, always consider how a respondent might

interpret the question and all the possible responses a person might want to make.

For example, you may want to know whether people smoke or not. You might ask the

question:

Do you smoke? (please tick) ❐ Yes ❐ No

In trialling this questionnaire, your respondent might ask whether you mean ciga-

rettes, cigars or marijuana. Is knowing whether they smoke enough? Should you also

10 Getting Started

ﬁ nd out how much they smoke (two or three cigarettes, versus two or three packs),

and/or how often they smoke (every day or only on social occasions)? The message

here is to consider each of your questions, what information they will give you and

what information might be missing.

Wording the questions

There is a real art to designing clear, well-written questionnaire items. Although there

are no clear-cut rules that can guide this process, there are some things you can do to

improve the quality of your questions, and therefore your data. Try to avoid:

• long complex questions

• double negatives

• double-barrelled questions

• jargon or abbreviations

• culture-speciﬁ c terms

• words with double meanings

• leading questions

• emotionally loaded words.

When appropriate, you should consider including a response category for ‘Don’t

know’ or ‘Not applicable’. For further suggestions on writing questions, see De Vaus

(2002) and Kline (2005).

11

2

Preparing a codebook

Before you can enter the information from your questionnaire, interviews or experi-

ment into SPSS, it is necessary to prepare a ‘codebook’. This is a summary of the

instructions you will use to convert the information obtained from each subject or

case into a format that SPSS can understand. The steps involved will be demonstrated

in this chapter using a data ﬁ le that was developed by a group of my graduate diploma

students. A copy of the questionnaire, and the codebook that was developed for this

questionnaire, can be found in the Appendix. The data ﬁ le is provided on the website

that accompanies this book. The provision of this material allows you to see the whole

process, from questionnaire development through to the creation of the ﬁ nal data ﬁ le

ready for analysis. Although I have used a questionnaire to illustrate the steps involved

in the development of a codebook, a similar process is also necessary in experimental

studies, or when retrieving information from existing records (e.g. hospital medical

records).

Preparing the codebook involves deciding (and documenting) how you will go

about:

• deﬁ ning and labelling each of the variables

• assigning numbers to each of the possible responses.

All this information should be recorded in a book or computer ﬁ le. Keep this some-

where safe; there is nothing worse than coming back to a data ﬁ le that you haven’t

used for a while and wondering what the abbreviations and numbers refer to.

In your codebook you should list all of the variables in your questionnaire, the

abbreviated variable names that you will use in SPSS and the way in which you will

code the responses. In this chapter simpliﬁ ed examples are given to illustrate the various

steps. In the ﬁ rst column of Table 2.1 you have the name of the variable (in English,

rather than in computer talk). In the second column you write the abbreviated name

12 Getting Started

for that variable that will appear in SPSS (see conventions below), and in the third

column you detail how you will code each of the responses obtained.

Variable names

Each question or item in your questionnaire must have a unique variable name. Some

of these names will clearly identify the information (e.g. sex, age). Other questions,

such as the items that make up a scale, may be identiﬁ ed using an abbreviation (e.g.

op1, op2, op3 is used to identify the items that make up the Optimism Scale).

There are a number of conventions you must follow in assigning names to your

variables in SPSS. These are set out in the ‘Rules for naming of variables’ box. In

earlier versions of SPSS (prior to Version 12), you could use only eight characters

for your variable names. The later versions of the program allow you longer variable

names, but very long names can make the output rather hard to read so keep them as

concise as possible.

Rules for naming of variables

Variable names:

• must be unique (i.e. each variable in a data set must have a different name)

• must begin with a letter (not a number)

• cannot include full stops, spaces or symbols (! , ? * “)

• cannot include words used as commands by SPSS (all, ne, eq, to, le, lt, by,

or, gt, and, not, ge, with)

• cannot exceed 64 characters.

Variable SPSS variable name Coding instructions

Identiﬁ cation number ID Number assigned to each survey

Sex Sex 1 = Males

2 = Females

Age Age Age in years

Marital status Marital 1 = single

2 = steady relationship

3 = married for the ﬁ rst time

4 = remarried

5 = divorced/separated

6 = widowed

Optimism Scale op1 to op6 Enter the number circled from

items 1 to 6 1 (strongly disagree) to

5 (strongly agree)

Table 2.1

Example of a

codebook

Preparing a codebook 13

The ﬁ rst variable in any data set should be ID—that is, a unique number that

identiﬁ es each case. Before beginning the data entry process, go through and assign a

number to each of the questionnaires or data records. Write the number clearly on the

front cover. Later, if you ﬁ nd an error in the data set, having the questionnaires or data

records numbered allows you to check back and ﬁ nd where the error occurred.

CODING RESPONSES

Each response must be assigned a numerical code before it can be entered into SPSS.

Some of the information will already be in this format (e.g. age in years); other vari-

ables such as sex will need to be converted to numbers (e.g. 1=males, 2=females). If

you have used numbers in your questions to label your responses (see, for example,

the education question in Chapter 1), this is relatively straightforward. If not, decide

on a convention and stick to it. For example, code the ﬁ rst listed response as 1, the

second as 2 and so on across the page.

What is your current marital status? (please tick)

❐ single ❐ in a relationship ❐ married ❐ divorced

To code responses to the question above: if a person ticked single, they would

be coded as 1; if in a relationship, they would be coded 2; if married, 3; and if

divorced, 4.

CODING OPEN-ENDED QUESTIONS

For open-ended questions (where respondents can provide their own answers), coding

is slightly more complicated. Take, for example, the question: What is the major source

of stress in your life at the moment? To code responses to this, you will need to scan

through the questionnaires and look for common themes. You might notice a lot of

respondents listing their source of stress as related to work, ﬁ nances, relationships,

health or lack of time. In your codebook you list these major groups of responses under

the variable name stress, and assign a number to each (work=1, spouse/partner=2 and

so on). You also need to add another numerical code for responses that did not fall

into these listed categories (other=99). When entering the data for each respondent,

you compare his/her response with those listed in the codebook and enter the appro-

priate number into the data set under the variable stress.

Once you have drawn up your codebook, you are almost ready to enter your data.

First you need to get to know SPSS (Chapter 3), and then you need to set up a data ﬁ le

and enter your data (Chapter 4).

14

3

Getting to know SPSS

SPSS operates using a number of different screens, or ‘windows’, designed to do differ-

ent things. Before you can access these windows, you need to either open an existing

data ﬁ le or create one of your own. So, in this chapter we will cover how to open and

close SPSS; how to open and close existing data ﬁ les; and how to create a data ﬁ le from

scratch. We will then go on to look at the different windows SPSS uses.

STARTING SPSS

There are a number of different ways to start SPSS:

• The simplest way is to look for an SPSS icon on your desktop. Place your cursor

on the icon and double-click.

• You can also start SPSS by clicking on Start, move your cursor to All Programs,

and then across to the list of programs available. See if you have a folder labelled

SPSS Inc, which should contain the option SPSS Statistics 18. This may vary

depending on your computer and the SPSS licence that you have.

• SPSS will also start up if you double-click on an SPSS data ﬁ le listed in Windows

Explorer—these ﬁ les have a .sav extension.

When you open SPSS, you may encounter a front cover screen asking ‘What would

you like to do?’ It is easier to close this screen (click on the cross in the top right-hand

corner) and to use the menus.

OPENING AN EXISTING DATA FILE

If you wish to open an existing data ﬁ le (e.g. survey4ED.sav, one of the ﬁ les included

on the website that accompanies this book—see p. viii), click on File from the menu

Getting to know SPSS 15

across the top of the screen, and then choose Open, and then slide across to Data. The

Open File dialogue box will allow you to search through the various directories on

your computer to ﬁ nd where your data ﬁ le is stored.

You should always open data ﬁ les from the hard drive of your computer. If you

have data on a memory stick or ﬂ ash drive, transfer it to a folder on the hard drive

of your computer before opening it. Find the ﬁ le you wish to use and click on Open.

Remember, all SPSS data ﬁ les have a .sav extension. The data ﬁ le will open in front of

you in what is labelled the Data Editor window (more on this window later).

WORKING WITH DATA FILES

In SPSS, you are allowed to have more than one data ﬁ le open at any one time. This

can be useful, but also potentially confusing. You must keep at least one data ﬁ le open

at all times. If you close a data ﬁ le, SPSS will ask if you would like to save the ﬁ le

before closing. If you don’t save it, you will lose any data you may have entered and

any recoding or computing of new variables that you may have done since the ﬁ le was

opened.

Saving a data ﬁ le

When you ﬁ rst create a data ﬁ le or make changes to an existing one (e.g. creating new

variables), you must remember to save your data ﬁ le. This does not happen automati-

cally. If you don’t save regularly and there is a power blackout or you accidentally press

the wrong key (it does happen!), you will lose all of your work. So save yourself the

heartache and save regularly.

To save a ﬁ le you are working on, go to the File menu (top left-hand corner) and

choose Save. Or, if you prefer, you can also click on the icon that looks like a ﬂ oppy

disk, which appears on the toolbar at the top left of your screen. This will save your

ﬁ le to whichever drive you are currently working on. This should always be the hard

drive—working from a ﬂ ash drive is a recipe for disaster! I have had many students

come to me in tears after corrupting their data ﬁ le by working from an external drive

rather than from the hard disk.

When you ﬁ rst save a new data ﬁ le, you will be asked to specify a name for the

ﬁ le and to indicate a directory and a folder in which it will be stored. Choose the

directory and then type in a ﬁ le name. SPSS will automatically give all data ﬁ le names

the extension .sav. This is so that it can recognise it as a data ﬁ le. Don’t change this

extension, otherwise SPSS won’t be able to ﬁ nd the ﬁ le when you ask for it again later.

Opening a different data ﬁ le

If you ﬁ nish working on a data ﬁ le and wish to open another one, click on File, select

Open, and then slide across to Data. Find the directory where your second ﬁ le is

16 Getting Started

stored. Click on the desired ﬁ le and then click the Open button. This will open the

second data ﬁ le, while still leaving the ﬁ rst data ﬁ le open in a separate window. It

is a good idea to close ﬁ les that you are not currently working on—it can get very

confusing having multiple ﬁ les open.

Starting a new data ﬁ le

Starting a new data ﬁ le is easy. Click on File, then, from the drop-down menu, click

on New and then Data. From here you can start deﬁ ning your variables and entering

your data. Before you can do this, however, you need to understand a little about

the windows and dialogue boxes that SPSS uses. These are discussed in the next

section.

WINDOWS

The main windows you will use in SPSS are the Data Editor, the Viewer, the Pivot

Table Editor, the Chart Editor and the Syntax Editor. These windows are summarised

here, but are discussed in more detail in later sections of this book.

When you begin to analyse your data, you will have a number of these windows

open at the same time. Some students ﬁ nd this idea very confusing. Once you get the

hang of it, it is really quite simple. You will always have the Data Editor open because

this contains the data ﬁ le that you are analysing. Once you start to do some analyses,

you will have the Viewer window open because this is where the results of all your

analyses are displayed, listed in the order in which you performed them.

The different windows are like pieces of paper on your desk—you can shufﬂ e

them around, so that sometimes one is on top and at other times another. Each of

the windows you have open will be listed along the bottom of your screen. To change

windows, just click on whichever window you would like to have on top. You can also

click on Window on the top menu bar. This will list all the open windows and allow

you to choose which you would like to display on the screen.

Sometimes the windows that SPSS displays do not initially ﬁ ll the screen. It is

much easier to have the Viewer window (where your results are displayed) enlarged

on top, ﬁ lling the entire screen. To do this, look on the top right-hand area of your

screen. There should be three little buttons or icons. Click on the middle button to

maximise that window (i.e. to make your current window ﬁ ll the screen). If you wish

to shrink it again, just click on this middle button.

Data Editor window

The Data Editor window displays the contents of your data ﬁ le, and in this window

you can open, save and close existing data ﬁ les, create a new data ﬁ le, enter data, make

changes to the existing data ﬁ le, and run statistical analyses (see Figure 3.1).

Getting to know SPSS 17

Viewer window

When you start to do analyses, the Viewer window should open automatically (see

Figure 3.2). If it does not open automatically, click on Window from the menu and this

should be listed. This window displays the results of the analyses you have conducted,

including tables and charts. In this window you can modify the output, delete it, copy

it, save it, or even transfer it into a Word document.

The Viewer screen consists of two parts. On the left is an outline or navigation

pane, which gives you a full list of all the analyses you have conducted. You can use

this side to quickly navigate your way around your output (which can become very

long). Just click on the section you want to move to and it will appear on the right-

hand side of the screen. On the right-hand side of the Viewer window are the results

of your analyses, which can include tables and graphs (also referred to as charts in

SPSS).

Saving output

When you save the output from SPSS, it is saved in a separate ﬁ le with a .spv exten-

sion, to distinguish it from data ﬁ les, which have a .sav extension. If you are using a

version of SPSS prior to version 18, your output will be given a .spo extension. To

Figure 3.1

Example of a Data

Editor window

18 Getting Started

read these older ﬁ les in SPSS Statistics 18, you will need to download a Legacy Viewer

program from the SPSS website.

To save the results of your analyses, you must have the Viewer window open on

the screen in front of you. Click on File from the menu at the top of the screen. Click

on Save. Choose the directory and folder in which you wish to save your output,

and then type in a ﬁ le name that uniquely identiﬁ es your output. Click on Save. To

name my ﬁ les, I use an abbreviation that indicates the data ﬁ le I am working on and

the date I conducted the analyses. For example, the ﬁ le survey8may2009.spv would

contain the analyses I conducted on 8 May 2009 using the survey data ﬁ le. I keep a log

book that contains a list of all my ﬁ le names, along with details of the analyses that

were performed. This makes it much easier for me to retrieve the results of speciﬁ c

analyses. When you begin your own research, you will ﬁ nd that you can very quickly

accumulate a lot of different ﬁ les containing the results of many different analyses.

To prevent confusion and frustration, get organised and keep good records of the

analyses you have done and of where you have saved the results.

It is important to note that the output ﬁ le (with a .spv extension) can only be

opened in SPSS. This can be a problem if you, or someone that needs to read the

output, does not have SPSS available. To get around this problem, you may choose to

‘export’ your SPSS results. If you wish to save the entire output, select File from the

Figure 3.2

Example of Viewer

window

Getting to know SPSS 19

menu and then choose Export. You can choose the format that you would like to use

(e.g. pdf, Word/rtf). Saving as a Word/rtf ﬁ le means that you will be able to modify the

tables in Word. Use the Browse button to identify the folder you wish to save the ﬁ le

into, specify a suitable name in the Save File pop-up box that appears and then click

on Save and then OK.

If you don’t want to save the whole ﬁ le, you can select speciﬁ c parts of the output

to export. Select these in the Viewer window using the left-hand navigation pane.

With the selections highlighted, select File from the menu and choose Export. In the

Export Output dialog box you will need to tick the box at the top labelled Selected

and then select the format of the ﬁ le and the location you wish to save to.

Printing output

You can use the navigation pane (left-hand side) of the Viewer window to select

particular sections of your results to print out. To do this, you need to highlight the

sections that you want. Click on the ﬁ rst section you want, hold down the Ctrl key

on your keyboard and then just click on any other sections you want. To print these

sections, click on the File menu (from the top of your screen) and choose Print. SPSS

will ask whether you want to print your selected output or the whole output.

Pivot Table Editor window

The tables you see in the Viewer window (which SPSS calls pivot tables) can be

modiﬁ ed to suit your needs. To modify a table you need to double-click on it, which

takes you into what is known as the Pivot Table Editor. You can use this editor to

change the look of your table, the size, the fonts used and the dimensions of the

columns—you can even swap the presentation of variables around (transpose rows

and columns).

If you click the right mouse button on a table in the Viewer window, a pop-up

menu of options that are speciﬁ c to that table will appear. If you double-click on a

table and then click on your right mouse button even more options appear, including

the option to Create Graph using these results. You may need to highlight the part

of the table that you want to graph by holding down the Ctrl key while you select the

parts of the table you want to display.

Chart Editor window

When you ask SPSS to produce a histogram, bar graph or scatterplot, it initially displays

these in the Viewer window. If you wish to make changes to the type or presentation

of the chart, you need to go into the Chart Editor window by double-clicking on your

chart. In this window you can modify the appearance and format of your graph, change

the fonts, colours, patterns and line markers (see Figure 3.3). The procedure to generate

charts and to use the Chart Editor is discussed further in Chapter 7.

20 Getting Started

Syntax Editor window

In the ‘good old days’, all SPSS commands were given using a special command

language or syntax. SPSS still creates these sets of commands to run each of the

programs, but all you usually see are the Windows menus that ‘write’ the commands

for you. Although the options available through the SPSS menus are usually all that

most undergraduate students need to use, there are some situations when it is useful

to go behind the scenes and to take more control over the analyses that you wish to

conduct.

Syntax is a good way of keeping a record of what commands you have used,

particularly when you need to do a lot of recoding of variables or computing new

variables (demonstrated in Chapter 8). It is also useful when you need to repeat a lot

of analyses or generate a number of similar graphs.

You can use the normal SPSS menus to set up the basic commands of a partic-

ular statistical technique and then ‘paste’ these to the Syntax Editor using a Paste

button provided with each procedure (see Figure 3.4). It allows you to copy and paste

commands, and to make modiﬁ cations to the commands generated by SPSS. Quite

complex commands can also be written to allow more sophisticated recoding and

manipulation of the data. SPSS has a Command Syntax Reference under the Help

menu if you would like additional information. (Warning: this is not for beginners—

it is quite complex to follow.)

The commands pasted to the Syntax Editor are not executed until you choose to

run them. To run the command, highlight the speciﬁ c command (making sure you

include the ﬁ nal full stop), or select it from the left-hand side of the screen, and then

click on the Run menu option or the arrow icon from the menu bar. Extra comments

can be added to the syntax ﬁ le by starting them with an asterisk (see Figure 3.4).

Syntax is stored in a separate text ﬁ le with a .sps extension. Make sure you have

the syntax editor open in front of you and then select File from the menu. Select the

Save option from the drop-down menu, choose the location you wish to save the ﬁ le

to and then type in a suitable ﬁ le name. Click on the Save button.

The syntax ﬁ le (with the extension .sps) can only be opened using SPSS. Some-

times it may be useful to copy and paste the syntax text from the Syntax Editor into

a Word document so that you (or others) can view it even if SPSS is not available. To

Figure 3.3

Example of a Chart

Editor window

Getting to know SPSS 21

do this, hold down the left mouse button and drag the cursor over the syntax you wish

to save. Choose Edit from the menu and then select Copy from the drop-down menu.

Open a Word document and paste this material using the Edit, Paste option or hold

the Ctrl key down and press V on the keyboard.

MENUS

Within each of the windows described above, SPSS provides you with quite a bewilder-

ing array of menu choices. These choices are displayed in drop-down menus across the

top of the screen, and also as icons. Try not to become overwhelmed; initially, just learn

the key ones, and as you get a bit more conﬁ dent you can experiment with others.

DIALOGUE BOXES

Once you select a menu option, you will usually be asked for further information. This

is done in a dialogue box. Figure 3.5 shows the dialogue box that appears when you

use the Frequencies procedure to get some descriptive statistics. To see this, click on

Analyze from the menu at the top of the screen, and then select Descriptive Statistics

and then slide across and select Frequencies. This will display a dialogue box asking

you to nominate which variables you want to use (see Figure 3.5).

Selecting variables in a dialogue box

To indicate which variables you want to use you need to highlight the selected vari-

ables in the list provided (by clicking on them), then click on the arrow button in

Figure 3.4

Example of a Syntax

Editor window

22 Getting Started

the centre of the screen to move them into the empty box labelled Variable(s). You

can select variables one at a time, clicking on the arrow each time, or you can select a

group of variables. If the variables you want to select are all listed together, just click

on the ﬁ rst one, hold down the Shift key on your keyboard and press the down arrow

key until you have highlighted all the desired variables. Click on the arrow button and

all of the selected variables will move across into the Variable(s) box.

If the variables you want to select are spread throughout the variable list, you should

click on the ﬁ rst variable you want, hold down the Ctrl key, move the cursor down to

the next variable you want and then click on it, and so on. Once you have all the desired

variables highlighted, click on the arrow button. They will move into the box.

To remove a variable from the box, you just reverse the process. Click on the

variable in the Variable(s) box that you wish to remove, click on the arrow button,

and it shifts the variable back into the original list. You will notice the direction of the

arrow button changes, depending on whether you are moving variables into or out of

the Variable(s) box.

Dialogue box buttons

In most dialogue boxes you will notice a number of standard buttons (OK, Paste,

Reset, Cancel and Help; see Figure 3.5). The uses of each of these buttons are:

• OK: Click on this button when you have selected your variables and are ready to

run the analysis or procedure.

Figure 3.5

Example of a

Frequencies

dialogue box

Getting to know SPSS 23

• Paste: This button is used to transfer the commands that SPSS has generated in

this dialogue box to the Syntax Editor. This is useful if you wish to keep a record

of the command or repeat an analysis a number of times.

• Reset: This button is used to clear the dialogue box of all the previous commands

you might have given when you last used this particular statistical technique or

procedure. It gives you a clean slate to perform a new analysis, with different

variables.

• Cancel: Clicking on this button closes the dialogue box and cancels all of the

commands you may have given in relation to that technique or procedure.

• Help: Click on this button to obtain information about the technique or

procedure you are about to perform.

Although I have illustrated the use of dialogue boxes in Figure 3.5 by using Frequen-

cies, all dialogue boxes work on the same basic principle. Each will have a series of

buttons with a variety of options relating to the speciﬁ c procedure or analysis. These

buttons will open subdialogue boxes that allow you to specify which analyses you wish

to conduct or which statistics you would like displayed.

CLOSING SPSS

When you have ﬁ nished your SPSS session and wish to close the program down, click

on the File menu at the top left of the screen. Click on Exit. SPSS will prompt you to

save your data ﬁ le and a ﬁ le that contains your output. You should not rely on the fact

that SPSS will prompt you to save when closing the program. It is important that you

save both your output and your data ﬁ le regularly throughout your session. If SPSS

crashes or there is a power cut you will lose all your work.

GETTING HELP

If you need help while using SPSS or don’t know what some of the options refer to,

you can use the in-built Help menu. Click on Help from the menu bar and a number

of choices are offered. You can ask for speciﬁ c topics, work through a Tutorial, or

consult a Statistics Coach. This takes you step by step through the decision-making

process involved in choosing the right statistic to use. This is not designed to replace

your statistics books, but it may prove a useful guide.

Within each of the major dialogue boxes there is an additional Help menu that

will assist you with the procedure you have selected.

This page intentionally left blank

PART TWO

Preparing the

data ﬁ le

Preparation of the data ﬁ le for analysis involves a number of steps. These include

creating the data ﬁ le and entering the information obtained from your study in a

format deﬁ ned by your codebook (covered in Chapter 2). The data ﬁ le then needs to

be checked for errors, and these errors corrected. Part Two of this book covers these

two steps. In Chapter 4, the procedures required to create a data ﬁ le and enter the

data are discussed. In Chapter 5, the process of screening and cleaning the data ﬁ le is

covered.

25

This page intentionally left blank

27

4

Creating a data ﬁ le and

entering data

There are a number of stages in the process of setting up a data ﬁ le and analysing the

data. The ﬂ ow chart shown on the next page outlines the main steps that are needed.

In this chapter I will lead you through the process of creating a data ﬁ le and entering

the data.

To prepare a data ﬁ le, three key steps are covered in this chapter:

• Step 1. The ﬁ rst step is to check and modify, where necessary, the options that

SPSS uses to display the data and the output that is produced.

• Step 2. The next step is to set up the structure of the data ﬁ le by ‘deﬁ ning’ the

variables.

• Step 3. The ﬁ nal step is to enter the data—that is, the values obtained from each

participant or respondent for each variable.

To illustrate these procedures I have used the data ﬁ le survey4ED.sav, which is

described in the Appendix. The codebook used to generate these data is also provided

in the Appendix.

Data ﬁ les can also be ‘imported’ from other spreadsheet-type programs (e.g.

Excel). This can make the data entry process much more convenient, particularly

for students who don’t have SPSS on their home computers. You can set up a basic

data ﬁ le on Excel and enter the data at home. When complete, you can then import

the ﬁ le into SPSS and proceed with the data manipulation and data analysis stages.

The instructions for using Excel to enter the data are provided later in this chapter.

28 Preparing the data ﬁ le

CHANGING THE SPSS ‘OPTIONS’

Before you set up your data ﬁ le, it is a good idea to check the SPSS options that govern

the way your data and output are displayed. The options allow you to deﬁ ne how your

variables will be displayed, the type of tables that will be displayed in the output and

many other aspects of the program. Some of this will seem confusing at ﬁ rst, but once

you have used the program to enter data and run some analyses you may want to refer

back to this section.

Flow chart of data analysis process

Prepare codebook (Chapter 2)

Set up structure of data ﬁ le (Chapter 4)

Enter data (Chapter 4)

Screen data ﬁ le for errors (Chapter 5)

Explore data using descriptive statistics and graphs (Chapters 6 and 7)

Modify variables for further analyses (Chapter 8)

Conduct statistical analyses to Conduct statistical analyses to

explore relationships (Part 4) compare groups (Part 5)

Correlation (Chapter 11) Non-parametric techniques (Chapter 16)

Partial correlation (Chapter 12) T-tests (Chapter 17)

Multiple regression (Chapter 13) Analysis of variance (Chapters 18, 19, 20)

Logistic regression (Chapter 14) Multivariate analysis of variance (Chapter 21)

Factor analysis (Chapter 15) Analysis of covariance (Chapter 22)

Creating a data ﬁ le and entering data 29

If you are sharing a computer with other people (e.g. in a computer lab), it is worth

being aware of these options. Sometimes other students will change these options,

which can dramatically inﬂ uence how the program appears. It is useful to know how

to change things back to the way you want them.

To open the Options screen, click on Edit from the menu at the top of the screen

and then choose Options. The screen shown in Figure 4.1 should appear. There are

a lot of choices listed, many of which you won’t need to change. I have described the

key ones below, organised by the tab they appear under. To move between the various

tabs, just click on the one you want. Don’t click on OK until you have ﬁ nished all the

changes you want to make, across all the tabs.

General tab

When you come to do your analyses, you can ask for your variables to be listed in

alphabetical order or by the order in which they appear in the ﬁ le. I always use the Figure 4.1

Example of an

Options screen

30 Preparing the data ﬁ le

ﬁ le order, because this is consistent with the order of the questionnaire items and the

codebook. To keep the variables in ﬁ le order, just make sure the option File in

the Variable Lists section is selected.

In the Output section on the right-hand side, place a tick in the box No scientiﬁ c

notation for small numbers in tables. This will stop you getting some very strange

numbers in your output for the statistical analyses. In the Notiﬁ cation section, make

sure the options Raise viewer window and Scroll to new output are selected. This

means that when you conduct an analysis the Viewer window will appear, and the

new output will be displayed on the screen.

Data tab

Click on the Data tab to make changes to the way that your data ﬁ le is displayed. If

your variables do not involve values with decimal places, you may like to change the

display format for all your variables. In the section labelled Display Format for New

Numeric Variables, change the Decimal Places value to 0. This means that all new

variables will not display any decimal places. This reduces the size of your data ﬁ le and

simpliﬁ es its appearance.

Output Labels tab

The options in this section allow you to customise how you want the variable names

and value labels displayed in your output. In the very bottom section under Variable

values in labels are shown as: choose Values and Labels from the drop-down options.

This will allow you to see both the numerical values and the explanatory labels in the

tables that are generated in the Viewer window.

Charts tab

Click on the Charts tab if you wish to change the appearance of your charts. You can

alter the Chart Aspect Ratio if you wish. You can also make other changes to the way

in which the chart is displayed (e.g. font, colour, lines).

Pivot Tables tab

SPSS presents most of the results of the statistical analyses in tables called pivot tables.

Under the Pivot Tables tab you can choose the format of these tables from an extensive

list. It is a matter of experimenting to ﬁ nd a style that best suits your needs. When I am

ﬁ rst doing my analyses, I use a style called CompactBoxed. This saves space (and paper

when printing). However, this style is not suitable for importing into documents that

are being sent for publication in a journal because it includes vertical lines. The styles

listed as ‘Academic’ may be more suitable here as they do not use vertical lines.

You can change the table styles as often as you like—just remember that you have

to change the style before you run the analysis. You cannot change the style of the

Creating a data ﬁ le and entering data 31

tables after they appear in your output, but you can modify many aspects (e.g. font

sizes, column width) by using the Pivot Table Editor. This can be activated by double-

clicking on the table that you wish to modify.

Once you have made all the changes you wish to make on the various Options

tabs, click on OK. You can then proceed to deﬁ ne your variables and enter your data.

DEFINING THE VARIABLES

Before you can enter your data, you need to tell SPSS about your variable names and

coding instructions. This is called ‘deﬁ ning the variables’. You will do this in the Data

Editor window (see Figure 4.2). The Data Editor window consists of two different

views: Data View and Variable View. You can move between these two views using the

little tabs at the bottom left-hand side of the screen.

You will notice that in the Data View window each of the columns is labelled var.

These will be replaced with the variable names that you listed in your codebook (see

Figure 4.2). Down the side you will see the numbers 1, 2, 3 and so on. These are the

case numbers that SPSS assigns to each of your lines of data. These are not the same

as your ID numbers, and these case numbers change if you sort your ﬁ le or split your

ﬁ le to analyse subsets of your data.

Procedure for deﬁ ning your variables

To de ﬁ ne each of the variables that make up your data ﬁ le, you ﬁ rst need to click

on the Variable View tab at the bottom left of your screen. In this view (see

Figure 4.3) the variables are listed down the side, with their characteristics listed

along the top (name, type, width, decimals, label etc.).

Your job now is to deﬁ ne each of your variables by specifying the required infor-

mation for each variable listed in your codebook. Some of the information you will

need to provide yourself (e.g. name); other bits are provided automatically using

Figure 4.2

Data Editor window

32 Preparing the data ﬁ le

default values. These default values can be changed if necessary. The key pieces of infor-

mation that are needed are described below. The headings I have used correspond to

the column headings displayed in the Variable View. I have provided the simple step-

by-step procedures below; however, there are a number of shortcuts that you can use

once you are comfortable with the process. These are listed later, in the section headed

‘Optional shortcuts’. You should become familiar with the basic techniques ﬁ rst.

Name

In this column, type in the brief variable name that will be used to identify each of the

variables in the data ﬁ le (listed in your codebook). Keep these variable names as short as

possible, not exceeding 64 characters. They must follow the naming conventions speci-

ﬁ ed by SPSS (listed in Chapter 2). Each variable name must be unique, must start with

a letter, and cannot contain spaces or symbols. For ideas on how to label your variables,

have a look at the codebooks provided in the Appendix. These list the variable names

used in data ﬁ les that accompany this book (see p. viii for details of these ﬁ les).

Type

The default value for Type that will appear automatically as you enter your ﬁ rst variable

name is Numeric. For most purposes, this is all you will need to use. There are some

circumstances where other options may be appropriate. For example, if you need to enter

text information (e.g. a person’s surname), you need to change the type to String. A Date

option is also available if your data includes dates. To change the variable type, click in the

cell and a box with three dots should appear giving you the options available. You can also

use this window to adjust the width of the variable and the number of decimal places.

Width

The default value for Width is 8. This is usually sufﬁ cient for most data. If your

variable has very large values (or you have requested a string variable), you may need

to change this default value; otherwise, leave it as is.

Figure 4.3

Variable View

Creating a data ﬁ le and entering data 33

Decimals

The default value for Decimals is usually 2 (however, this can be changed using the

Options facility described earlier in this chapter). If your variable has decimal places,

change this to suit your needs.

Label

The Label column allows you to provide a longer description for your variable than

used in the Name column. This will be used in the output generated from the analyses

conducted by SPSS. For example, you may wish to give the label Total Optimism to

your variable TOPTIM.

Values

In the Values column you can deﬁ ne the meaning of the values you have used to code

your variables. I will demonstrate this process for the variable Sex.

1. Click on the three dots on the right-hand side of the cell. This opens the

Value Label dialogue box.

2. Click in the box marked Value. Type in 1.

3. Click in the box marked Label. Type in Male.

4. Click on Add. You will then see in the summary box: 1=Male.

5. Repeat for Females: Value: enter 2, Label: enter Female. Add.

6. When you have ﬁ nished deﬁ ning all the possible values (as listed in your

codebook), click on OK.

Missing

Sometimes researchers assign speciﬁ c values to indicate missing values for their data.

This is not essential—SPSS will recognise any blank cell as missing data. So if you

intend to leave a blank when a piece of information is not available, it is not necessary

to do anything with this Variable View column.

If you do intend to use speciﬁ c missing value codes (e.g. 99=not applicable), you

must specify this value in the Missing section, otherwise SPSS will use the value as a

legitimate value in any statistical analyses. Click in the cell and then on the shaded box

with three dots that appears. Choose the option Discrete missing values and type the

value (e.g. 99) in the space provided. Up to three values can be speciﬁ ed. Click on OK.

If you are using these special codes, it is also a good idea to go back and label these

values in the Values column.

Columns

The default column width is usually set at 8, which is sufﬁ cient for most purposes.

Change it only if necessary to accommodate your values or long variable names.

34 Preparing the data ﬁ le

Align

The alignment of the columns is usually set at ‘right’ alignment. There is no need to

change this.

Measure

The column heading Measure refers to the level of measurement of each of your vari-

ables. The default is Scale, which refers to continuous data measured at interval or ratio

level of measurement. If your variable consists of categories (e.g. sex), click in the cell and

then on the arrow key that appears. Choose Nominal for categorical data and Ordinal if

your data involve rankings or ordered values (e.g. level of education completed).

Optional shortcuts

The process described above can be rather tedious if you have a large number of

variables in your data ﬁ le. There are a number of shortcuts you can use to speed up

the process. If you have a number of variables that have the same ‘attributes’ (e.g.

type, width, decimals), you can set the ﬁ rst variable up correctly and then copy these

attributes to one or more other variables.

Copying variable deﬁ nition attributes to one other variable

1. In Variable View, click on the cell that has the attribute you wish to copy

(e.g. Width).

2. From the menu, click on Edit and then Copy.

3. Click on the same attribute cell for the variable you wish to apply this to.

4. From the menu, click on Edit and then Paste.

Copying variable deﬁ nition attributes to a number of other

variables

1. In Variable View, click on the cell that has the attribute you wish to copy

(e.g. Width).

2. From the menu, click on Edit and then Copy.

3. Click on the same attribute cell for the ﬁ rst variable you wish to copy to

and then, holding your left mouse button down, drag the cursor down

the column to highlight all the variables you wish to copy to.

4. From the menu, click on Edit and then Paste.

Setting up a series of new variables all with the same attributes

If your data consists of scales made up of a number of individual items, you can

create the new variables and deﬁ ne the attributes of all of these items in one go. The

Creating a data ﬁ le and entering data 35

procedure is detailed below, using the six items of the Optimism Scale as an example

(optim1 to optim6). If you want to practise this as an exercise, you should start a new

data ﬁ le (File, New, Data).

1. In Variable View, deﬁ ne the attributes of the ﬁ rst variable (optim1)

following the instructions provided earlier. This would involve deﬁ ning

the value labels 1=strongly disagree, 2=disagree, 3=neutral, 4=agree,

5=strongly agree.

2. With the Variable View selected, click on the row number of this variable

(this should highlight the whole row).

3. From the menu, select Edit and then Copy.

4. Click on the row number of the next empty row.

5. From the menu, select Edit and then Paste Variables.

6. In the dialogue box that appears, enter the number of additional

variables you want to add (in this case, 5). Enter the preﬁ x you wish to

use (optim) and the number you wish the new variables to start on (in this

case, 2). Click on OK.

This will give you ﬁ ve new variables (optim2, optim3, optim4, optim5 and optim6).

To set up all of the items in other scales, just repeat the process detailed above

(e.g. sest1 to sest10 for the self-esteem items). Remember, this procedure is suitable

only for items that have all the same attributes; it is not appropriate if the items have

different response scales (e.g. if some are categorical and others continuous), or if the

values are coded differently.

ENTERING DATA

Once you have deﬁ ned each of your variable names and given them value labels (where

appropriate), you are ready to enter your data. Make sure you have your codebook ready.

Procedure for entering data

1. To enter data, you need to have the Data View active. Click on the Data

View tab at the bottom left-hand side of the screen of the Data Editor

window. A spreadsheet should appear with your newly deﬁ ned variable

names listed across the top.

2. Click on the ﬁ rst cell of the data set (ﬁ rst column, ﬁ rst row).

3. Type in the number (if this variable is ID, this should be 1).

4. Press the right arrow key on your keyboard; this will move the cursor into

the second cell, ready to enter your second piece of information for case

number 1.

36 Preparing the data ﬁ le

5. Move across the row, entering all the information for case 1, making sure

that the values are entered in the correct columns.

6. To move back to the start, press the Home key on your keyboard (on some

computers you may need to hold the Ctrl key or the Fn key down and

then press the Home key). Press the down arrow to move to the second

row, and enter the data for case 2.

7. If you make a mistake and wish to change a value, click in the cell that

contains the error. Type in the correct value and then press the right

arrow key.

After you have deﬁ ned your variables and entered your data, your Data Editor window

should look something like that shown previously in Figure 3.1.

If you have entered value labels for some of your variables (e.g. Sex: 1=male,

2=female), you can choose to have these labels displayed in the Data Editor window

instead of just the numbers. To do this, click on View from the menu and select the

option Va lu e Labels. This option can also be activated during the data entry process so

that you can choose an option from a drop-down menu, rather than typing a number

in each cell. This is slower, but does ensure that only valid numbers are entered. To

turn this option off, go to View and click on Value Labels again to remove the tick.

MODIFYING THE DATA FILE

After you have created a data ﬁ le, you may need to make changes to it (e.g. to add,

delete or move variables, or to add or delete cases). Make sure you have the Data

Editor window open on the screen, showing Data View.

Delete a case

Move down to the case (row) you wish to delete. Position your cursor in the shaded

section on the left-hand side that displays the case number. Click once to highlight

the row. Press the Delete button on your computer keyboard. You can also click on the

Edit menu and click on Clear.

Insert a case between existing cases

Move your cursor to a cell in the case (row) immediately below where you would like

the new case to appear. Click on the Edit menu and choose Insert Cases. An empty

row will appear in which you can enter the data of the new case.

Delete a variable

Position your cursor in the shaded section (which contains the variable name) above the

column you wish to delete. Click once to highlight the whole column. Press the Delete

button on your keyboard. You can also click on the Edit menu and click on Clear.

Creating a data ﬁ le and entering data 37

Insert a variable between existing variables

Position your cursor in a cell in the column (variable) to the right of where you would

like the new variable to appear. Click on the Edit menu and choose Insert Variable.

An empty column will appear in which you can enter the data of the new variable.

Move an existing variable(s)

In the Data Editor window, have the Variable View showing. Highlight the variable

you wish to move by clicking in the left-hand margin. Click and hold your left mouse

button and then drag the variable to the new position (a red line will appear as you

drag). Release the left mouse button when you get to the desired spot.

DATA ENTRY USING EXCEL

Data ﬁ les can be prepared in the Microsoft Excel program and then imported into

SPSS for analysis. This is great for students who don’t have access to SPSS at

home. Excel usually comes as part of the Microsoft Ofﬁ ce package—check under All

Programs in your Start menu. The procedure for creating a data ﬁ le in Excel and

then importing it into SPSS is described below. If you intend to use this option you

should have at least a basic understanding of Excel, as this will not be covered here.

Warning: Excel can cope with only 256 columns of data (or variables). If your data

ﬁ le is likely to be larger than this, it is probably easier to set it up in SPSS rather than

convert from Excel to SPSS later. Alternatively, you can use different Excel spread-

sheets (each with the ID as the ﬁ rst variable), convert each to SPSS separately, then

merge the ﬁ les in SPSS later (see instructions in the next section).

Step 1: Set up the variable names

Set up an Excel spreadsheet with the variable names in the ﬁ rst row across

the page. The variable names must conform to the SPSS rules for naming

variables (see Chapter 2).

Step 2: Enter the data

1. Enter the information for the ﬁ rst case on one line across the page, using

the appropriate columns for each variable.

2. Repeat for each of the remaining cases. Don’t use any formulas or other

Excel functions. Remember to save your ﬁ le regularly.

3. Click on File, Save. In the section marked Save as Type, make sure

Microsoft Excel Workbook is selected. Type in an appropriate ﬁ le name.

38 Preparing the data ﬁ le

Step 3: Converting to SPSS

1. After you have entered the data, save your ﬁ le and then close Excel.

2. Start SPSS and select File, Open, Data from the menu at the top of the

screen.

3. In the section labelled Files of type, choose Excel. Excel ﬁ les have a .xls or

.xlsx extension. Find the ﬁ le that contains your data. Click on it so that it

appears in the File name section.

4. Click on the Open button. A screen will appear labelled Opening Excel

Data Source. Make sure there is a tick in the box Read variable names

from the ﬁ rst row of data. Click on OK.

The data will appear on the screen with the variable names listed across the top. You

will then need to save this new SPSS ﬁ le.

Step 4: Saving as an SPSS ﬁ le

1. Choose File, and then Save As from the menu at the top of the screen.

2. Type in a suitable ﬁ le name. Make sure that the Save as Type is set at

SPSS Statistics (*.sav). Click on Save.

3. In the Data Editor, Variable view, you will now need to deﬁ ne each of

the Labels, Values and Measure information (see instructions presented

earlier). You may also want to reduce the width of the columns as they

often come in from Excel with a width of 11.

When you wish to open this ﬁ le later to analyse your data using SPSS, make sure you

choose the ﬁ le that has a .sav extension (not your original Excel ﬁ le that has a .xls

extension).

MERGE FILES

There are times when it is necessary to merge different data ﬁ les. SPSS allows you to

merge ﬁ les by adding additional cases at the end of your ﬁ le, or to merge additional

variables for each of the cases in an existing data ﬁ le (e.g. when Time 2 data becomes

available). This second option is also useful when you have Excel ﬁ les with infor-

mation spread across different spreadsheets that need to be merged by ID.

To merge ﬁ les by adding cases

This procedure will allow you to merge ﬁ les that have the same variables, but differ-

ent cases; for example, where the same information is recorded at two different sites

Creating a data ﬁ le and entering data 39

(e.g. clinic settings) or entered by two different people. The two ﬁ les should have the

same variable names for the data you wish to merge (although other non-equivalent

information can exist in each ﬁ le).

If the ID numbers used in each ﬁ le are the same (starting at ID=1, 2, 3), you will

need to change the ID numbers in one of the ﬁ les before merging so that each case

is still uniquely identiﬁ ed. To do this, open one of the ﬁ les, choose Transform from

the menu, and then Compute Variable. Type ID in the Target Variable box, and then

ID + 1000 in the Numeric Expression box (or some number that is bigger than the

number of cases in the ﬁ le). Click on the OK button, and then on OK in the dialogue

box that asks if you wish to change the variable. This will create new ID numbers

for this ﬁ le starting at 1001, 1002 and so on. Note this in your codebook for future

reference. Then you are ready to merge the ﬁ les.

1. Open the ﬁ rst ﬁ le that you wish to merge.

2. Go to the Data menu, choose Merge Files and then Add Cases.

3. In the dialogue box, click on An external SPSS data ﬁ le and choose the

ﬁ le that you wish to merge with. (If your second ﬁ le is already open it will

be listed in the top box, An open dataset.)

4. Click on Continue and then on OK. Save the new data ﬁ le using a

different name (File, Save As).

To merge ﬁ les by adding variables

This option is useful when adding additional information for each case (with the

matching IDs). Each ﬁ le must start with the ID number.

1. Sort each ﬁ le in ascending order by ID by clicking on the Data menu,

choose Sort Cases and choose ID.

2. Go to the Data menu, choose Merge ﬁ les and then Add Variables.

3. In the dialogue box, click on An external SPSS data ﬁ le and choose the

ﬁ le that you wish to merge with. (If your second ﬁ le is already open it will

be listed in the top box, An open dataset.)

4. In the Excluded variables box, you should see the ID variable listed

(because it exists in both data ﬁ les). (If you have any other variables listed

here, you will need to click on the Rename button to change the variable

name so that it is unique.)

5. Click on the ID variable, and then on the box Match cases on key variables

and on the arrow button to move ID into the Key Variables box. This means

that all information will be matched by ID. Click on Continue and then OK.

6. Save your merged ﬁ le under a different name (File, Save As).

40 Preparing the data ﬁ le

USEFUL SPSS FEATURES

There are many useful features of SPSS that can be used to help with analyses, and to

save you time and effort. I have highlighted a few of the main ones in the following

sections.

Sort the data ﬁ le

You can ask SPSS to sort your data ﬁ le according to values on one of your variables

(e.g. sex, age).

1. Click on the Data menu, choose Sort Cases and specify which variable will

be used to sort by. Choose either Ascending or Descending. Click on OK.

2. To return your ﬁ le to its original order repeat the process, asking SPSS to

sort the ﬁ le by ID.

Split the data ﬁ le

Sometimes it is necessary to split your ﬁ le and to repeat analyses for groups (e.g. males and

females) separately. This procedure does not physically alter your ﬁ le in any permanent

manner. It is an option you can turn on and off as it suits your purposes. The order in

which the cases are displayed in the data ﬁ le will change, however. You can return the data

ﬁ le to its original order (by ID) by using the Sort Cases command described above.

1. Click on the Data menu and choose the Split File option.

2. Click on Compare groups and specify the grouping variable (e.g. sex).

Click on OK.

For the analyses that you perform after this split ﬁ le procedure, the two groups (in this

case, males and females) will be analysed separately.

Important: when you have ﬁ nished the analyses, you need to go back and turn the

Split File option off.

1. Click on the Data menu and choose the Split File option.

2. Click on the ﬁ rst dot (Analyze all cases, do not create groups). Click on OK.

Select cases

For some analyses, you may wish to select a subset of your sample (e.g. only males).

1. Click on the Data menu and choose the Select Cases option.

2. Click on the If condition is satisﬁ ed button.

3. Click on the button labelled IF.

Creating a data ﬁ le and entering data 41

4. Choose the variable that deﬁ nes the group that you are interested in (e.g. sex).

5. Click on the arrow button to move the variable name into the box. Click

on the = key from the keypad displayed on the screen.

6. Type in the value that corresponds to the group you are interested in

(check with your codebook). For example, males in this sample are coded

1, therefore you would type in 1. The command line should read: sex=1.

7. Click on Continue and then OK.

For the analyses (e.g. correlation) that you perform after this Select Cases procedure,

only the group that you selected (e.g. males) will be included.

Important: when you have ﬁ nished the analyses, you need to go back and turn the

Select Cases option off, otherwise it will apply to all analyses conducted.

1. Click on the Data menu and choose Select Cases option.

2. Click on the ﬁ rst All cases option. Click on OK.

USING SETS

With large data ﬁ les, it can be a pain to have to scroll through lots of variable names

in SPSS dialogue boxes to reach the ones that you want to analyse. SPSS allows you to

deﬁ ne and use ‘sets’ of variables. This is particularly useful in the survey4ED.sav data

ﬁ le, where there are lots of individual items that are added to give total scores, which

are located at the end of the ﬁ le. In the following example, I will establish a set that

includes only the demographic variables and the scale totals.

1. Click on Utilities from the menu and choose Deﬁ ne Variable Sets.

2. Choose the variables you want in your set from the list. Include ID, the

demographic variables (sex through to smoke number), and then all the

totals at the end of the data ﬁ le from Total Optimism onwards. Move

these into the Variables in Set box.

3. In the box Set Name, type an appropriate name for your set (e.g. Totals).

4. Click on the Add Set button and then on Close.

To use the sets you have created, you need to activate them.

1. Click on Utilities and on Use Variable Sets.

2. In the list of variable sets, tick the set you have created (Totals) and then

go up and untick the ALLVARIABLES option, as this would display all

variables. Leave NEWVARIABLES ticked. Click on OK.

42 Preparing the data ﬁ le

With the sets activated, only the selected variables will be displayed in the data ﬁ le and

in the dialogue boxes used to conduct statistical analyses.

To turn the option off

1. Click on Utilities and on Use Variable Sets.

2. Tick the ALLVARIABLES option and click OK.

Data ﬁ le comments

Under the Utilities menu, SPSS provides you with the chance to save descriptive

comments with a data ﬁ le.

1. Select Utilities and Data File Comments.

2. Type in your comments, and if you would like them recorded in the

output ﬁ le, click on the option Display comments in output. Comments

are saved with the date they were made.

Display values labels in data ﬁ le

When the data ﬁ le is displayed in the Data Editor window, the numerical values for

all variables are usually shown. If you would like the value labels (e.g. male, female)

displayed instead, go to the View menu and choose Value Labels. To turn this option

off, go to the View menu and click on Value Labels again to remove the tick.

43

5

Screening and

cleaning the data

Before you start to analyse your data, it is essential that you check your data set for errors.

It is very easy to make mistakes when entering data, and unfortunately some errors can

completely mess up your analyses. For example, entering 35 when you mean to enter 3

can distort the results of a correlation analysis. Some analyses are very sensitive to what

are known as ‘outliers’; that is, values that are well below or well above the other scores.

So it is important to spend the time checking for mistakes initially, rather than trying to

repair the damage later. Although boring, and a threat to your eyesight if you have large

data sets, this process is essential and will save you a lot of heartache later!

The data screening process involves a number of steps:

• Step 1: Checking for errors. First, you need to check each of your variables for

scores that are out of range (i.e. not within the range of possible scores).

• Step 2: Finding and correcting the error in the data ﬁ le. Second, you need to ﬁ nd

where in the data ﬁ le this error occurred (i.e. which case is involved) and correct

or delete the value.

To give you the chance to practise these steps, I have created a modiﬁ ed data ﬁ le

(error4ED.sav) provided on the website accompanying this book (this is based on the

main ﬁ le survey4ED.sav—see details on p. viii and in the Appendix). To follow along,

you will need to start SPSS and open the error4ED.sav ﬁ le. In working through each of

the steps on the computer you will become more familiar with the use of menus, inter-

preting the output from SPSS analyses and manipulating your data ﬁ le. For each of the

procedures, I have included the SPSS syntax. For more information on the use of the

Syntax Editor for recording and saving the SPSS commands, see Chapter 3.

Before you start, you should go to the Edit menu and choose Options. Under the

Output Labels tab, go down to the ﬁ nal box (Variable values in labels shown as:) and

choose Values and Labels. This will allow you to display both the values and labels

used for each of your categorical variables—making identiﬁ cation of errors easier.

44 Preparing the data ﬁ le

STEP 1: CHECKING FOR ERRORS

When checking for errors, you are primarily looking for values that fall outside the

range of possible values for a variable. For example, if sex is coded 1=male, 2=female,

you should not ﬁ nd any scores other than 1 or 2 for this variable. Scores that fall

outside the possible range can distort your statistical analyses—so it is very important

that all these errors are corrected before you start. To check for errors, you will need

to inspect the frequencies for each of your variables. This includes all of the individual

items that make up the scales. Errors must be corrected before total scores for these

scales are calculated. It is a good idea to keep a log book where you record any errors

that you detect and any changes that you make to your data ﬁ le.

There are a number of different ways to check for errors using SPSS. I will illus-

trate two different ways, one that is more suitable for categorical variables (e.g. sex)

and the other for continuous variables (e.g. age).

Checking categorical variables

In this section, the procedure for checking categorical variables for errors is presented.

In the example shown below, I will illustrate the process using the error4ED.sav

data ﬁ le (included on the website accompanying this book—see p. viii), checking for

errors on the variables Sex, Marital status and Highest education completed. Some

deliberate errors have been introduced in the error4ED.sav data ﬁ le so that you

can get practice spotting them—they are not present in the main survey4ED.sav

data ﬁ le.

Procedure for checking categorical variables

1. From the main menu at the top of the screen, click on Analyze, then click

on Descriptive Statistics, then Frequencies.

2. Choose the variables that you wish to check (e.g. sex, marital, educ.).

3. Click on the arrow button to move these into the Variable box.

4. Click on the Statistics button. Tick Minimum and Maximum in the

Dispersion section.

5. Click on Continue and then on OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

FREQUENCIES

VARIABLES=sex marital educ

/STATISTICS=MINIMUM MAXIMUM

/ORDER= ANALYSIS .

Screening and cleaning the data 45

Selected output generated using this procedure is displayed as follows.

There are two parts to the output. The ﬁ rst table provides a summary of each of the

variables you requested. The remaining tables give you a breakdown, for each variable,

of the range of responses. (These are listed using the value label and the code number

that was used if you changed the Options as suggested earlier in this chapter.)

46 Preparing the data ﬁ le

• Check your Minimum and Maximum values. Do they make sense? Are they

within the range of possible scores on that variable? You can see from the ﬁ rst

table (labelled Statistics) that, for the variable Sex, the minimum value is 1 and

the maximum is 3. This value is incorrect, as the maximum value should only be

2 according to the codebook in the Appendix. For marital status, the scores are

within the appropriate range of 1 to 8. The maximum value for highest educ is 22,

indicating an error, as the maximum value should only be 6.

• Check the number of Valid and Missing cases. If there are a lot of missing cases, you

need to ask why. Have you made errors in entering the data (e.g. put the data in the

wrong columns)? Sometimes extra cases appear at the bottom of the data ﬁ le, where

you may have moved your cursor too far down and accidentally created some ‘empty’

cases. If this occurs, open your Data Editor window, move down to the empty case

row, click in the shaded area where the case number appears and press Delete on your

keyboard. Rerun the Frequencies procedure again to get the correct values.

• Other tables are also presented in the output, corresponding to each of the vari-

ables that were investigated. In these tables, you can see how many cases fell into

each of the legitimate categories. It also shows how many cases have out-of-range

values. There is one case with a value of 3 for sex, and one person with a value of

22 for education. We will need to ﬁ nd out where these errors occurred, but ﬁ rst

we will demonstrate how to check for errors in some of the continuous variables

in the data ﬁ le.

Checking continuous variables

Procedure for checking continuous variables

1. From the menu at the top of the screen, click on Analyze, then click on

Descriptive statistics, then Descriptives.

2. Click on the variables that you wish to check. Click on the arrow button to

move them into the Variables box (e.g. age).

3. Click on the Options button. You can ask for a range of statistics. The

main ones at this stage are mean, standard deviation, minimum and

maximum. Click on the statistics you wish to generate.

4. Click on Continue, and then on OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

DESCRIPTIVES

VARIABLES=age

/STATISTICS=MEAN STDDEV MIN MAX .

Screening and cleaning the data 47

The output generated from this procedure is shown as follows.

• Check the Minimum and Maximum values. Do these make sense? In this case, the

ages range from 2 to 82. The minimum value suggests an error (given this was an

adult-only sample).

• Does the Mean score make sense? If there is an out-of-range value in the data ﬁ le,

this will distort the mean value. If the variable is the total score on a scale, is the

mean value what you expected from previous research on this scale?

STEP 2: FINDING AND CORRECTING THE ERROR IN THE

DATA FILE

So what do we do if we ﬁ nd some out-of-range responses (e.g. a value of 3 for sex)?

First, we need to ﬁ nd the error in the data ﬁ le. Don’t try to scan through your entire

data set looking for the error—there are a number of different ways to ﬁ nd an error

in a data ﬁ le. I will illustrate two approaches.

Method 1

1. Click on the Data menu and choose Sort Cases.

2. In the dialogue box that pops up, click on the variable that you know

has an error (e.g. sex) and then on the arrow to move it into the Sort By

box. Click on either ascending or descending (depending on whether you

want the higher values at the top or the bottom). For sex, we want to

ﬁ nd the person with the value of 3, so we would choose descending.

3. Click on OK.

In the Data Editor window, make sure that you have selected the Data View tab

so that you can see your data values. The case with the error for your selected variable

(e.g. sex) should now be located at the top of your data ﬁ le. Look across to the sex

variable column. In this example, you will see that the ﬁ rst case listed (ID=103) has

a value of 3 for sex. If this was your data, you would need to access the original ques-

tionnaires and check whether the person with an identiﬁ cation number of 103 was

a male or female. You would then delete the value of 3 and type in the correct value.

Record this information in your log book. If you don’t have access to the original data,

48 Preparing the data ﬁ le

you should delete the value and let SPSS replace it with the system missing code (it

will show as a full stop—this happens automatically, don’t type a full stop).

When you ﬁ nd an error in your data ﬁ le, it is important that you check for other

errors in the surrounding columns. In this example, notice that the inappropriate

value of 2 for age is also for person ID=103.

Shown below is another way that we could have found the case that had an error

for sex.

Method 2

1. Make sure that the Data Editor window is open and on the screen with

the data showing.

2. Click on the variable name in which the error has occurred (e.g. sex).

3. Click once to highlight the column.

4. Click on Edit from the menu across the top of the screen. Click on Find.

5. In the Find box, type in the incorrect value that you are looking for (e.g. 3).

6. Click on Find Next. SPSS will scan through the ﬁ le and will stop at the ﬁ rst

occurrence of the value that you speciﬁ ed. Take note of the ID number of

this case (from the ﬁ rst column). You will need this to check your records

or questionnaires to ﬁ nd out what the value should be.

7. Click on Find Next again if you need to continue searching for other

cases with the same incorrect value. In this example, we know from the

Frequencies output that there is only one incorrect value of 3.

8. Click on Close when you have ﬁ nished searching.

After you have corrected your errors, it is essential to repeat Frequencies to double-

check. Sometimes, in correcting one error you may have accidentally caused another

error. Although this process is tedious, it is very important that you start with a clean,

error-free data set. The success of your research depends on it. Don’t cut corners!

CASE SUMMARIES

One other aspect of SPSS that may be useful in this data screening process is Sum-

marize Cases. This allows you to select and display speciﬁ c pieces of information for

each case.

1. Click on Analyze, go to Reports and choose Case Summaries.

2. Choose the ID variable and other variables you are interested in (e.g. sex,

child, smoker). Remove the tick from the Limit cases to ﬁ rst 100.

3. Click on the Statistics button and remove Number of cases from the Cell

Statistics box. Click on Continue.

Screening and cleaning the data 49

4. Click on the Options button and remove the tick from Subheadings for

totals.

5. Click on Continue and then on OK (or on Paste to save to Syntax Editor).

The syntax from this procedure is:

SUMMARIZE

/TABLES=id sex child smoke

/FORMAT=VALIDLIST NOCASENUM NOTOTAL

/TITLE=‘Case Summaries’

/MISSING=VARIABLE

/CELLS=NONE.

Part of the output is shown below.

In this chapter, we have checked for errors in only a few of the variables in the data

ﬁ le to illustrate the process. For your own research, you would obviously check every

variable in the data ﬁ le. If you would like some more practice ﬁ nding errors, repeat

the procedures described above for all the variables in the error4ED.sav data ﬁ le. I

have deliberately included a few errors to make the process more meaningful. Refer

to the codebook in the Appendix for survey4ED.sav to ﬁ nd out what the legitimate

values for each variable should be.

For additional information on the screening and cleaning process, I would strongly

recommend you read Chapter 4 in Tabachnick and Fidell (2007).

This page intentionally left blank

PART THREE

Preliminary

analyses

Once you have a clean data ﬁ le, you can begin the process of inspecting your data ﬁ le and

exploring the nature of your variables. This is in readiness for conducting speciﬁ c statis-

tical techniques to address your research questions. There are ﬁ ve chapters that make

up Part Three of this book. In Chapter 6, the procedures required to obtain descriptive

statistics for both categorical and continuous variables are presented. This chapter also

covers checking the distribution of scores on continuous variables in terms of normality

and possible outliers. Graphs can be useful tools when getting to know your data. Some

of the more commonly used graphs available through SPSS are presented in Chapter 7.

Sometimes manipulation of the data ﬁ le is needed to make it suitable for speciﬁ c

analyses. This may involve calculating the total score on a scale, by adding up the scores

obtained on each of the individual items. It may also involve collapsing a continuous

variable into a smaller number of categories. These data manipulation techniques are

covered in Chapter 8. In Chapter 9, the procedure used to check the reliability (internal

consistency) of a scale is presented. This is particularly important in survey research,

or in studies that involve the use of scales to measure personality characteristics, atti-

tudes, beliefs etc. Also included in Part Three is a chapter that helps you through the

decision-making process in deciding which statistical technique is suitable to address

your research question. In Chapter 10, you are provided with an overview of some of

the statistical techniques available in SPSS and led step by step through the process

of deciding which one would suit your needs. Important aspects that you need to con-

sider (e.g. type of question, data type, characteristics of the variables) are highlighted.

51

This page intentionally left blank

53

6

Descriptive statistics

Once you are sure there are no errors in the data ﬁ le (or at least no out-of-range values

on any of the variables), you can begin the descriptive phase of your data analysis.

Descriptive statistics have a number of uses. These include to:

• describe the characteristics of your sample in the Method section of your report

• check your variables for any violation of the assumptions underlying the statisti-

cal techniques that you will use to address your research questions

• address speciﬁ c research questions.

The two procedures outlined in Chapter 5 for checking the data will also give you

information for describing your sample in the Method section of your report.

In studies involving human participants, it is useful to collect information on the

number of people or cases in the sample, the number and percentage of males and

females in the sample, the range and mean of ages, education level, and any other

relevant background information. Prior to doing many of the statistical analyses (e.g.

t-test, ANOVA, correlation), it is important to check that you are not violating any

of the ‘assumptions’ made by the individual tests. (These are covered in detail in Part

Four and Part Five of this book.)

Testing of assumptions usually involves obtaining descriptive statistics on your

variables. These descriptive statistics include the mean, standard deviation, range

of scores, skewness and kurtosis. Descriptive statistics can be obtained a number of

different ways, providing a variety of information.

If all you want is a quick summary of the characteristics of the variables in your

data ﬁ le, you can use a relatively new feature of SPSS (this may not be available if using

earlier versions of SPSS).

54 Preliminary analyses

Procedure for obtaining codebook

1. Click on Analyze, go to Reports and choose Codebook.

2. Select the variables you want and move them into the Codebook

Variables box.

3. Click on the Output tab and untick (by clicking on the box with a tick) all

the options except Label, Value Labels and Missing Values.

4. Click on the Statistics tab and make sure that all the options in both

sections are ticked.

5. Click on Continue, and then OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

CODEBOOK sex [n] age [s]

/VARINFO LABEL VALUELABELS MISSING

/OPTIONS VARORDER=VARLIST SORT=ASCENDING MAXCATS=200

/STATISTICS COUNT PERCENT MEAN STDDEV QUARTILES.

The output is shown below.

The output from the procedure shown above gives you a quick summary of the

cases in your data ﬁ le. Often, however, you need more detailed information. This

Descriptive statistics 55

can be obtained using the Frequencies, Descriptives or Explore procedures. These

are all procedures listed under the Analyze, Descriptive Statistics drop-down menu.

There are, however, different procedures depending on whether you have a categori-

cal or continuous variable. Some of the statistics (e.g. mean, standard deviation) are

not appropriate if you have a categorical variable. The different approaches to be

used with categorical and continuous variables are presented in the following two

sections. If you would like to follow along with the examples in this chapter, open the

survey4ED.sav ﬁ le.

CATEGORICAL VARIABLES

To obtain descriptive statistics for categorical variables, you should use Frequencies.

This will tell you how many people gave each response (e.g. how many males, how

many females). It doesn’t make any sense asking for means, standard deviations etc.

for categorical variables, such as sex or marital status.

Procedure for obtaining descriptive statistics for categorical variables

1. From the menu click on Analyze, then click on Descriptive Statistics, then

Frequencies.

2. Choose and highlight the categorical variables you are interested in

(e.g. sex). Move these into the Variables box.

3. Click on OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

FREQUENCIES

VARIABLES=sex

/ORDER= ANALYSIS .

The output is shown below.

56 Preliminary analyses

Interpretation of output from Frequencies

From the output shown above, we know that there are 185 males (42.1 per cent)

and 254 females (57.9 per cent) in the sample, giving a total of 439 respondents. It is

important to take note of the number of respondents you have in different subgroups

in your sample. For some analyses (e.g. ANOVA), it is easier to have roughly equal

group sizes. If you have very unequal group sizes, particularly if the group sizes are

small, it may be inappropriate to run some analyses.

CONTINUOUS VARIABLES

For continuous variables (e.g. age) it is easier to use Descriptives, which will provide

you with ‘summary’ statistics such as mean, median and standard deviation. You

certainly don’t want every single value listed, as this may involve hundreds of values

for some variables. You can collect the descriptive information on all your continu-

ous variables in one go; it is not necessary to do it variable by variable. Just transfer

all the variables you are interested in into the box labelled Variables. If you have a lot

of variables, however, your output will be extremely long. Sometimes it is easier to do

them in chunks and tick off each group of variables as you do them.

Procedure for obtaining descriptive statistics for continuous variables

1. From the menu click on Analyze, then select Descriptive Statistics, then

Descriptives.

2. Click on all the continuous variables that you wish to obtain descriptive

statistics for. Click on the arrow button to move them into the Variables

box (e.g. age, Total perceived stress: tpstress).

3. Click on the Options button. Make sure mean, standard deviation,

minimum, maximum are ticked and then click on skewness, kurtosis.

4. Click on Continue, and then OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

DESCRIPTIVES

VARIABLES=age tpstress

/STATISTICS=MEAN STDDEV MIN MAX KURTOSIS SKEWNESS .

The output generated from this procedure is shown below.

Descriptive statistics 57

Interpretation of output from Descriptives

In the output presented above, the information we requested for each of the vari-

ables is summarised. For example, for the variable age we have information from

439 respondents, ranging in age from 18 to 82 years, with a mean of 37.44 and

standard deviation of 13.20. This information may be needed for the Method section

of a report to describe the characteristics of the sample.

Descriptives also provides some information concerning the distribution of

scores on continuous variables (skewness and kurtosis). This information may

be needed if these variables are to be used in parametric statistical techniques

(e.g. t-tests, analysis of variance). The Skewness value provides an indication of

the symmetry of the distribution. Kurtosis, on the other hand, provides infor-

mation about the ‘peakedness’ of the distribution. If the distribution is perfectly

normal, you would obtain a skewness and kurtosis value of 0 (rather an uncommon

occurrence in the social sciences).

Positive skewness values indicate positive skew (scores clustered to the left at the

low values). Negative skewness values indicate a clustering of scores at the high end

(right-hand side of a graph). Positive kurtosis values indicate that the distribution is

rather peaked (clustered in the centre), with long thin tails. Kurtosis values below 0

indicate a distribution that is relatively ﬂ at (too many cases in the extremes). With

reasonably large samples, skewness will not ‘make a substantive difference in the

analysis’ (Tabachnick & Fidell 2007, p. 80). Kurtosis can result in an underestimate of

the variance, but this risk is also reduced with a large sample (200+ cases: see Tabach-

nick & Fidell 2007, p. 80).

While there are tests that you can use to evaluate skewness and kurtosis values,

these are too sensitive with large samples. Tabachnick and Fidell (2007, p. 81) rec-

ommend inspecting the shape of the distribution (e.g. using a histogram). The proce-

dure for further assessing the normality of the distribution of scores is provided later

in this section.

58 Preliminary analyses

MISSING DATA

When you are doing research, particularly with human beings, it is rare that you will

obtain complete data from every case. It is important that you inspect your data ﬁ le

for missing data. Run Descriptives and ﬁ nd out what percentage of values is missing

for each of your variables. If you ﬁ nd a variable with a lot of unexpected missing data,

you need to ask yourself why. You should also consider whether your missing values

are happening randomly, or whether there is some systematic pattern (e.g. lots of

women over 30 years of age failing to answer the question about their age!).

You also need to consider how you will deal with missing values when you come

to do your statistical analyses. The Options button in many of the SPSS statistical

procedures offers you choices for how you want to deal with missing data. It is impor-

tant that you choose carefully, as it can have dramatic effects on your results. This is

particularly important if you are including a list of variables and repeating the same

analysis for all variables (e.g. correlations among a group of variables, t-tests for a

series of dependent variables).

• The Exclude cases listwise option will include cases in the analysis only if they

have full data on all of the variables listed in your Variables box for that case. A

case will be totally excluded from all the analyses if it is missing even one piece of

information. This can severely, and unnecessarily, limit your sample size.

• The Exclude cases pairwise option, however, excludes the case (person) only

if they are missing the data required for the speciﬁ c analysis. They will still be

included in any of the analyses for which they have the necessary information.

• The Replace with mean option, which is available in some SPSS statistical proce-

dures (e.g. multiple regression), calculates the mean value for the variable and

gives every missing case this value. This option should never be used, as it can

severely distort the results of your analysis, particularly if you have a lot of missing

values.

Always press the Options button for any statistical procedure you conduct, and check

which of these options is ticked (the default option varies across procedures). I would

suggest that you use pairwise exclusion of missing data, unless you have a pressing

reason to do otherwise. The only situation where you might need to use listwise ex-

clusion is when you want to refer only to a subset of cases that provided a full set of

results.

For more experienced users, there are more advanced and complex options avail-

able in SPSS for estimating missing values (e.g. imputation). These are included in

the Missing Value Analysis procedure. This can also be used to detect patterns within

missing data.

Descriptive statistics 59

ASSESSING NORMALITY

Many of the statistical techniques presented in Part Four and Part Five of this book

assume that the distribution of scores on the dependent variable is ‘normal’. Normal

is used to describe a symmetrical, bell-shaped curve, which has the greatest frequency

of scores in the middle with smaller frequencies towards the extremes (see Gravet-

ter & Wallnau 2004, p. 48). Normality can be assessed to some extent by obtaining

skewness and kurtosis values (as described in the previous section). However, other

techniques are also available in SPSS using the Explore option of the Descriptive

Statistics menu. This procedure is detailed below. In this example, I will assess the

normality of the distribution of scores for Total perceived stress for the sample as a

whole. You also have the option of doing this separately for different groups in your

sample by specifying an additional categorical variable (e.g. sex) in the Factor List

option that is available in the Explore dialogue box.

Procedure for assessing normality using Explore

1. From the menu at the top of the screen click on Analyze, then select

Descriptive Statistics, then Explore.

2. Click on the variable(s) you are interested in (e.g. Total perceived stress:

tpstress). Click on the arrow button to move them into the Dependent

List box.

3. In the Label Cases by: box, put your ID variable.

4. In the Display section, make sure that Both is selected.

5. Click on the Statistics button and click on Descriptives and Outliers. Click

on Continue.

6. Click on the Plots button. Under Descriptive, click on Histogram. Click on

Normality plots with tests. Click on Continue.

7. Click on the Options button. In the Missing Values section, click on

Exclude cases pairwise. Click on Continue and then OK (or on Paste to

save to Syntax Editor).

The syntax generated is:

EXAMINE

VARIABLES=tpstress

/ID= id

/PLOT BOXPLOT HISTOGRAM NPPLOT

/COMPARE GROUP

/STATISTICS DESCRIPTIVES EXTREME

/CINTERVAL 95

60 Preliminary analyses

/MISSING PAIRWISE

/NOTOTAL.

Selected output generated from this procedure is shown below.

Tests of normality

Descriptive statistics 61

62 Preliminary analyses

Descriptive statistics 63

Interpretation of output from Explore

Quite a lot of information is generated as part of this output. This tends to be a bit

overwhelming until you know what to look for. I will take you through the output

step by step.

• In the table labelled Descriptives, you are provided with descriptive statistics and

other information concerning your variables. If you speciﬁ ed a grouping variable

in the Factor List, this information will be provided separately for each group,

rather than for the sample as a whole. Some of this information you will recognise

(mean, median, std deviation, minimum, maximum etc.).

One statistic you may not know is the 5% Trimmed Mean. To obtain this

value, SPSS removes the top and bottom 5 per cent of your cases and calculates a

new mean value. If you compare the original mean (26.73) and this new trimmed

mean (26.64), you can see whether your extreme scores are having a strong inﬂ u-

ence on the mean. If these two mean values are very different, you may need to

investigate these data points further. The ID values of the most extreme cases are

shown in the Extreme Values table.

• Skewness and kurtosis values are also provided as part of this output, giving

information about the distribution of scores for the two groups (see discussion of

the meaning of these values in the previous section).

• In the table labelled Tests of Normality, you are given the results of the

Kolmogorov-Smirnov statistic. This assesses the normality of the distribution of

scores. A non-signiﬁ cant result (Sig. value of more than .05) indicates normal-

ity. In this case, the Sig. value is .000, suggesting violation of the assumption of

normality. This is quite common in larger samples.

• The actual shape of the distribution for each group can be seen in the Histograms.

In this example, scores appear to be reasonably normally distributed. This is also

supported by an inspection of the normal probability plots (labelled Normal Q-Q

Plot). In this plot, the observed value for each score is plotted against the expected

value from the normal distribution. A reasonably straight line suggests a normal

distribution.

• The Detrended Normal Q-Q Plots are obtained by plotting the actual deviation

of the scores from the straight line. There should be no real clustering of points,

with most collecting around the zero line.

• The ﬁ nal plot that is provided in the output is a boxplot of the distribution of

scores for the two groups. The rectangle represents 50 per cent of the cases, with

the whiskers (the lines protruding from the box) going out to the smallest and

largest values. Sometimes you will see additional circles outside this range—these

are classiﬁ ed by SPSS as outliers. The line inside the rectangle is the median value.

Boxplots are discussed further in the next section on detecting outliers.

64 Preliminary analyses

In the example given above, the distribution of scores was reasonably ‘normal’. Often this

is not the case. Many scales and measures used in the social sciences have scores that are

skewed, either positively or negatively. This does not necessarily indicate a problem with

the scale, but rather reﬂ ects the underlying nature of the construct being measured. Life

satisfaction measures, for example, are often negatively skewed, with most people being

reasonably happy with their lot in life. Clinical measures of anxiety or depression are often

positively skewed in the general population, with most people recording relatively few

symptoms of these disorders. Some authors in this area recommend that, with skewed

data, the scores be ‘transformed’ statistically. This issue is discussed further in Chapter 8.

CHECKING FOR OUTLIERS

Many of the statistical techniques covered in this book are sensitive to outliers (cases

with values well above or well below the majority of other cases). The techniques

described in the previous section can also be used to check for outliers.

• First, have a look at the Histogram. Look at the tails of the distribution. Are there

data points sitting on their own, out on the extremes? If so, these are potential

outliers. If the scores drop away in a reasonably even slope, there is probably not

too much to worry about.

• Second, inspect the Boxplot. Any scores that SPSS considers are outliers appear

as little circles with a number attached (this is the ID number of the case). SPSS

deﬁ nes points as outliers if they extend more than 1.5 box-lengths from the edge

of the box. Extreme points (indicated with an asterisk, *) are those that extend

more than three box-lengths from the edge of the box. In the example above there

are no extreme points, but there are two outliers: ID numbers 24 and 157. If you

ﬁ nd points like this, you need to decide what to do with them.

• It is important to check that the outlier’s score is genuine, not just an error. Check

the score and see whether it is within the range of possible scores for that variable.

Sometimes it is worth checking back with the questionnaire or data record to see

if there was a mistake in entering the data. If it is an error, correct it, and repeat

the boxplot. If it turns out to be a genuine score, you then need to decide what

you will do about it. Some statistics writers suggest removing all extreme outliers

from the data ﬁ le. Others suggest changing the value to a less extreme value, thus

including the person in the analysis but not allowing the score to distort the

statistics (for more advice on this, see Chapter 4 in Tabachnick & Fidell 2007).

• The information in the Descriptives table can give you an indication of how

much of a problem these outlying cases are likely to be. The value you are inter-

ested in is the 5% Trimmed Mean. If the trimmed mean and mean values are very

different, you may need to investigate these data points further. In this example,

the two mean values (26.73 and 26.64) are very similar. Given this, and the fact

Descriptive statistics 65

that the values are not too different from the remaining distribution, I will retain

these cases in the data ﬁ le.

• If you wish to change or remove values in your ﬁ le, go to the Data Editor window,

sort the data ﬁ le in descending order (to ﬁ nd the people with the highest values)

or ascending if you are concerned about cases with very low values. The cases you

need to look at in more detail are then at the top of the data ﬁ le. Move across to

the column representing that variable and modify or delete the value of concern.

Always record changes to your data ﬁ le in a log book.

ADDITIONAL EXERCISES

Business

Data ﬁ le: staffsurveysav. See Appendix for details of the data ﬁ le.

1. Follow the procedures covered in this chapter to generate appropriate descriptive

statistics to answer the following questions.

(a) What percentage of the staff in this organisation are permanent employees?

(Use the variable employstatus.)

(b) What is the average length of service for staff in the organisation? (Use the

variable service.)

(c) What percentage of respondents would recommend the organisation to others

as a good place to work? (Use the variable recommend.)

2. Assess the distribution of scores on the Total Staff Satisfaction Scale (totsatis) for

employees who are permanent versus casual (employstatus).

(a) Are there any outliers on this scale that you would be concerned about?

(b) Are scores normally distributed for each group?

Health

Data ﬁ le: sleep4ED.sav. See Appendix for details of the data ﬁ le.

1. Follow the procedures covered in this chapter to generate appropriate descriptive

statistics to answer the following questions.

(a) What percentage of respondents are female (gender)?

(b) What is the average age of the sample?

(c) What percentage of the sample indicated that they had a problem with their

sleep (probsleeprec)?

(d) What is the median number of hours sleep per weeknight (hourweeknight)?

2. Assess the distribution of scores on the Sleepiness and Associated Sensations Scale

(totSAS) for people who feel that they do/don’t have a sleep problem (probsleeprec).

(a) Are there any outliers on this scale that you would be concerned about?

(b) Are scores normally distributed for each group?

66

While the numerical values obtained in Chapter 6 provide useful information

concerning your sample and your variables, some aspects are better explored visually.

SPSS provides a number of different types of graphs (also referred to as charts). In this

chapter, I’ll cover the basic procedures to obtain the following graphs:

• histograms

• bar graphs

• line graphs

• scatterplots

• boxplots.

In SPSS there are a number of different ways of generating graphs, using the Graph

menu option. These include Chart Builder, Graphboard Template Chooser, and

Legacy Dialogs.

In this chapter I will demonstrate the Legacy Dialogs approach, which I ﬁ nd the

easiest way to generate graphs. Spend some time playing with each of the different

graphs and exploring their possibilities. In this chapter only a brief overview is given

to get you started. To illustrate the various graphs I have used the survey4ED.sav

data ﬁ le, which is included on the website accompanying this book (see p. viii and the

Appendix for details). If you wish to follow along with the procedures described in

this chapter, you will need to start SPSS and open the ﬁ le labelled survey4ED.sav.

At the end of this chapter, instructions are also given on how to edit a graph

to better suit your needs. This may be useful if you intend to use the graph in your

research paper. The procedure for importing graphs directly into Microsoft Word is

also detailed.

7

Using graphs to describe

and explore the data

Using graphs to describe and explore the data 67

HISTOGRAMS

Histograms are used to display the distribution of a single continuous variable (e.g.

age, perceived stress scores).

Procedure for creating a histogram

1. From the menu click on Graphs, then select Legacy Dialogs. Choose

Histogram.

2. Click on your variable of interest and move it into the Variable box. This

should be a continuous variable (e.g. Total perceived stress: tpstress).

3. If you would like to generate separate histograms for different groups

(e.g. male/female), you could put an additional variable (e.g. sex) in the

Panel by: section. Choose Rows if you would like the two graphs on top

of one another, or Column if you want them side by side. In this example,

I will put the sex variable in the Column box.

4. Click on OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

GRAPH

/HISTOGRAM=tpstress

/PANEL COLVAR=sex COLOP=CROSS .

The output generated from this procedure is shown below.

68 Preliminary analyses

Interpretation of output from Histogram

Inspection of the shape of the histogram provides information about the distribution

of scores on the continuous variable. Many of the statistics discussed in this manual

assume that the scores on each of the variables are normally distributed (i.e. follow the

shape of the normal curve). In this example the scores are reasonably normally distrib-

uted, with most scores occurring in the centre, tapering out towards the extremes. It is

quite common in the social sciences, however, to ﬁ nd that variables are not normally

distributed. Scores may be skewed to the left or right or, alternatively, arranged in a

rectangular shape. For further discussion of the assessment of the normality of vari-

ables see Chapter 6.

Using graphs to describe and explore the data 69

BAR GRAPHS

Bar graphs can be simple or very complex, depending on how many variables you

wish to include. The bar graph can show the number of cases in particular categories,

or it can show the score on some continuous variable for different categories. Basi-

cally, you need two main variables—one categorical and one continuous. You can also

break this down further with another categorical variable if you wish.

Procedure for creating a bar graph

1. From the menu at the top of the screen, click on Graphs, then select

Legacy Dialogs. Choose Bar. Click on Clustered.

2. In the Data in chart are section, click on Summaries for groups of cases.

Click on Deﬁ ne.

3. In the Bars represent box, click on Other statistic (e.g. mean).

4. Click on the continuous variable you are interested in (e.g. Total perceived

stress: tpstress). This should appear in the box listed as Mean (Total

perceived stress). This indicates that the mean on the Perceived Stress

Scale for the different groups will be displayed.

5. Click on your ﬁ rst categorical variable (e.g. agegp3). Click on the arrow

button to move it into the Category axis box. This variable will appear

across the bottom of your bar graph (X axis).

6. Click on another categorical variable (e.g. sex) and move it into the

Deﬁ ne Clusters by: box. This variable will be represented in the legend.

7. If you would like to display error bars on your graph, click on the Options

button and click on Display error bars. Choose what you want the bars to

represent (e.g. conﬁ dence intervals).

8. Click on Continue and then OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

GRAPH

/BAR(GROUPED)=MEAN(tpstress) BY agegp3 BY sex.

/INTERVAL CI(95.0).

70 Preliminary analyses

The output generated from this procedure is shown below.

Interpretation of output from Bar Graph

The output from this procedure gives you a quick summary of the distribution of

scores for the groups that you have requested (in this case, males and females from

the different age groups). The graph presented above suggests that females had higher

perceived stress scores than males, and that this difference is more pronounced among

the two older age groups. Among the 18 to 29 age group, the difference in scores

between males and females is very small.

Care should be taken when interpreting the output from Bar Graph. You should

always look at the scale used on the Y (vertical) axis. Sometimes what looks like a

dramatic difference is really only a few scale points and, therefore, probably of little

importance. This is clearly evident in the bar graph displayed above. You will see that

the difference between the groups is quite small when you consider the scale used to

display the graph. The difference between the smallest score (males aged 45 or more)

and the highest score (females aged 18 to 29) is only about three points.

To assess the signiﬁ cance of any difference you might ﬁ nd between groups, it is

necessary to conduct further statistical analyses. In this case, a two-way, between-

groups analysis of variance (see Chapter 19) would be conducted to ﬁ nd out if the

differences are statistically signiﬁ cant.

Using graphs to describe and explore the data 71

LINE GRAPHS

A line graph allows you to inspect the mean scores of a continuous variable across a

number of different values of a categorical variable (e.g. time 1, time 2, time 3). They are

also useful for graphically exploring the results of a one- or two-way analysis of variance.

Line graphs are provided as an optional extra in the output of analysis of variance (see

Chapters 18 and 19). The following procedure shows you how to generate a line graph

using the same variables as in the previous procedure for bar graphs.

Procedure for creating a line graph

1. From the menu at the top of the screen, select Graphs, then Legacy

Dialogs, then Line.

2. Click on Multiple. In the Data in Chart Are section, click on Summaries for

groups of cases. Click on Deﬁ ne.

3. In the Lines represent box, click on Other statistic. Click on the

continuous variable you are interested in (e.g. Total perceived stress:

tpstress). Click on the arrow button. The variable should appear in

the box listed as Mean (Total perceived stress). This indicates that the

mean on the Perceived Stress Scale for the different groups will be

displayed.

4. Click on your ﬁ rst categorical variable (e.g. agegp3). Click on the arrow

button to move it into the Category Axis box. This variable will appear

across the bottom of your line graph (X axis).

5. Click on another categorical variable (e.g. sex) and move it into the

Deﬁ ne Lines by: box. This variable will be represented in the legend.

6. If you would like to add error bars to your graph, you can click on the

Options button. Click on the Display error bars box and choose what you

would like the error bars to represent (e.g. conﬁ dence intervals).

7. Click on OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

GRAPH

/LINE(MULTIPLE)MEAN(tpstress) BY agegp3 BY sex.

72 Preliminary analyses

The output generated from this procedure is shown below.

18-29 30-44 45+

age 3 groups

25

26

27

28

Mean total perceived stress

sex

MALES

FEMALES

Interpretation of output from Line Chart

• First, you can look at the impact of age on perceived stress for each of the sexes

separately. Younger males appear to have higher levels of perceived stress than

either middle-aged or older males. For females, the difference across the age

groups is not quite so pronounced. The older females are only slightly less stressed

than the younger group.

• You can also consider the difference between males and females. Overall, males

appear to have lower levels of perceived stress than females. Although the differ-

ence for the younger group is only small, there appears to be a discrepancy for

the older age groups. Whether or not these differences reach statistical signiﬁ -

cance can be determined only by performing a two-way analysis of variance (see

Chapter 19).

Using graphs to describe and explore the data 73

The results presented above suggest that to understand the impact of age on

perceived stress you must consider the respondents’ gender. This sort of relationship

is referred to, when doing analysis of variance, as an interaction effect. While the use

of a line graph does not tell you whether this relationship is statistically signiﬁ cant, it

certainly gives you a lot of information and raises a lot of additional questions.

Sometimes in interpreting the output it is useful to consider other questions. In

this case, the results suggest that it may be worthwhile to explore in more depth the

relationship between age and perceived stress for the two groups (males and females).

To do this I decided to split the sample, not just into three groups for age, as in the

above graph, but into ﬁ ve groups to get more detailed information concerning

the inﬂ uence of age.

After dividing the group into ﬁ ve equal groups (by creating a new variable,

age5gp—instructions for this process are presented in Chapter 8), a new line graph

was generated. This gives us a clearer picture of the inﬂ uence of age than the previous

line graph using only three age groups.

18-24 25-32 33-40 41-49 50+

age 5 groups

24

25

26

27

28

29

Mean total perceived stress

sex

MALES

FEMALES

74 Preliminary analyses

SCATTERPLOTS

Scatterplots are typically used to explore the relationship between two continuous

variables (e.g. age and self-esteem). It is a good idea to generate a scatterplot before

calculating correlations (see Chapter 11). The scatterplot will give you an indication

of whether your variables are related in a linear (straight-line) or curvilinear fashion.

Only linear relationships are suitable for correlation analyses.

The scatterplot will also indicate whether your variables are positively related

(high scores on one variable are associated with high scores on the other) or nega-

tively related (high scores on one are associated with low scores on the other). For

positive correlations, the points form a line pointing upwards to the right (that is,

they start low on the left-hand side and move higher on the right). For negative corre-

lations, the line starts high on the left and moves down on the right (see an example

of this in the output below).

The scatterplot also provides a general indication of the strength of the relationship

between your two variables. If the relationship is weak the points will be all over the

place, in a blob-type arrangement. For a strong relationship the points will form a vague

cigar shape, with a deﬁ nite clumping of scores around an imaginary straight line.

In the example that follows, I request a scatterplot of scores on two of the scales

in the survey: the Total perceived stress and the Total Perceived Control of Internal

States Scale (PCOISS). I have asked for two groups in my sample (males and females)

to be represented separately on the one scatterplot (using different symbols). This not

only provides me with information concerning my sample as a whole but also gives

additional information on the distribution of scores for males and females.

If you would prefer to have separate scatterplots for each group, you can specify a

categorical variable in the Panel by: section instead of the Set Markers by: box shown

below. If you wish to obtain a scatterplot for the full sample (not split by group), just

ignore the instructions below in the section labelled Set Markers by:

Procedure for creating a scatterplot

1. From the menu at the top of the screen, click on Graphs, then Legacy

Dialogs and then on Scatter/Dot.

2. Click on Simple Scatter and then Deﬁ ne.

3. Click on your ﬁ rst variable, usually the one you consider is the dependent

variable (e.g. Total perceived stress: tpstress).

4. Click on the arrow to move it into the box labelled Y axis. This variable

will appear on the vertical axis.

5. Move your other variable (e.g. Total PCOISS: tpcoiss) into the box labelled

X axis. This variable will appear on the horizontal axis.

6. You can also have SPSS mark each of the points according to some other

Using graphs to describe and explore the data 75

categorical variable (e.g. sex). Move this variable into the Set Markers by:

box. This will display males and females using different markers.

7. Move the ID variable in the Label Cases by: box. This will allow you to

ﬁ nd out the ID number of a case from the graph if you ﬁ nd an outlier.

8. Click on OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

GRAPH

/SCATTERPLOT(BIVAR)=tpcoiss WITH tpstress BY sex BY id (IDENTIFY)

/MISSING=LISTWISE .

The output generated from this procedure, modiﬁ ed slightly for display

purposes, is shown below.

20 30 40 50 60 70 80 90

total PCOISS

10

20

30

40

50

total perceived stress

sex

MALES

FEMALES

76 Preliminary analyses

Interpretation of output from Scatterplot

From the output on the previous page, there appears to be a moderate, negative correla-

tion between the two variables (Perceived Stress and PCOISS) for the sample as a whole.

Respondents with high levels of perceived control (shown on the X, or horizontal, axis)

experience lower levels of perceived stress (shown on the Y, or vertical, axis). On the

other hand, people with low levels of perceived control have much greater perceived

stress.

Remember, the scatterplot does not give you deﬁ nitive answers; you need to follow

it up with the calculation of the appropriate statistic. There is no indication of a curvi-

linear relationship, so it would be appropriate to calculate a Pearson product-moment

correlation for these two variables (see Chapter 11) if the distributions are roughly

normal (check the histograms for these two variables).

In the example above, I have looked at the relationship between only two vari-

ables. It is also possible to generate a matrix of scatterplots between a whole group

of variables. This is useful as preliminary assumption testing for analyses such as

MANOVA.

Procedure to generate a matrix of scatterplots

1. From the menu at the top of the screen, click on Graphs, then Legacy

Dialogs and then on Scatter/Dot.

2. Click on Matrix Scatter. Click on the Deﬁ ne button.

3. Select all of your continuous variables (tnegaff, tposaff, tpstress) and

move them into the Matrix Variables box.

4. Select the sex variable and move it into the Rows box.

5. Click on the Options button and select Exclude cases variable by variable.

6. Click on Continue and then OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

GRAPH

/SCATTERPLOT(MATRIX)=tposaff tnegaff tpstress

/PANEL ROWVAR=sex ROWOP=CROSS

/MISSING=VARIABLEWISE .

Using graphs to describe and explore the data 77

The output generated from this procedure is shown below.

BOXPLOTS

Boxplots are useful when you wish to compare the distribution of scores on variables.

You can use them to explore the distribution of one continuous variable (e.g. positive

affect) or, alternatively, you can ask for scores to be broken down for different groups

(e.g. age groups). You can also add an extra categorical variable to compare (e.g. males

and females). In the example below, I will explore the distribution of scores on the

Positive Affect scale for males and females.

78 Preliminary analyses

Procedure for creating a boxplot

1. From the menu at the top of the screen, click on Graphs, then select

Legacy Dialogs and then Boxplot.

2. Click on Simple. In the Data in Chart Are section, click on Summaries for

groups of cases. Click on the Deﬁ ne button.

3. Click on your continuous variable (e.g. Total Positive Affect: tposaff). Click

on the arrow button to move it into the Variable box.

4. Click on your categorical variable (e.g. sex). Click on the arrow button to

move it into the Category axis box.

5. Click on ID and move it into the Label cases box. This will allow you to

identify the ID numbers of any cases with extreme values.

6. Click on OK (or on Paste to save to Syntax Editor).

The syntax generated from this procedure is:

EXAMINE

VARIABLES=tposaff BY sex

/PLOT=BOXPLOT/STATISTICS=NONE/NOTOTAL/ID=id.

The output generated from this procedure is shown as follows.

Using graphs to describe and explore the data 79

Interpretation of output from Boxplot

The output from Boxplot gives you a lot of information about the distribution of

your continuous variable and the possible inﬂ uence of your other categorical variable

(and cluster variable if used).

• Each distribution of scores is represented by a box and protruding lines (called

whiskers). The length of the box is the variable’s interquartile range and contains

50 per cent of cases. The line across the inside of the box represents the median

value. The whiskers protruding from the box go out to the variable’s smallest and

largest values.

• Any scores that SPSS considers are outliers appear as little circles with a number

attached (this is the ID number of the case). Outliers are cases with scores that are

quite different from the remainder of the sample, either much higher or much

lower. SPSS deﬁ nes points as outliers if they extend more than 1.5 box-lengths

from the edge of the box. Extreme points (indicated with an asterisk, *) are those

that extend more than three box-lengths from the edge of the box. For more

information on outliers, see Chapter 6. In the example above, there are a number

of outliers at the low values for Positive Affect for both males and females.

• In addition to providing information on outliers, a boxplot allows you to inspect

the pattern of scores for your various groups. It provides an indication of the

variability in scores within each group and allows a visual inspection of the differ-

ences between groups. In the example presented above, the distribution of scores

on Positive Affect for males and females is very similar.

EDITING A CHART OR GRAPH

Sometimes modiﬁ cations need to be made to the titles, labels, markers etc. of a graph

before you can print it or use it in your report. For example, I have edited some of the

graphs displayed in this chapter to make them clearer (e.g. changing the patterns in the

bar graph, thickening the lines used in the line graph). To edit a chart or graph, you

need to open the Chart Editor window. To do this, place your cursor on the graph that

you wish to modify. Double-click and a new window will appear showing your graph,

complete with additional menu options and icons (see Figure 7.1).

You should see a smaller Properties window pop up, which allows you to make

changes to your graphs. If this does not appear, click on the Edit menu and select

Properties.

There are a number of changes you can make while in Chart Editor:

• To change the words used in labels, click once on the label to highlight it (a gold-

coloured box should appear around the text). Click once again to edit the text (a

80 Preliminary analyses

red cursor should appear). Modify the text and then press Enter on your keyboard

when you have ﬁ nished.

• To change the position of the X and Y axis labels (e.g. to centre them), double-click

on the title you wish to change. In the Properties box, click on the Text Layout

tab. In the section labelled Justify, choose the position you want (the dot means

centred, the left arrow moves it to the left, and the right arrow moves it to the

right).

• To change the characteristics of the text, lines, markers, colours, patterns and scale

used in the chart, click once on the aspect of the graph that you wish to change.

The Properties window will adjust its options depending on the aspect you click

on. The various tabs in this box will allow you to change aspects of the graph. If

you want to change one of the lines of a multiple-line graph (or markers for a

group), you will need to highlight the speciﬁ c category in the legend (rather than

on the graph itself). This is useful for changing one of the lines to dashes so that

it is more clearly distinguishable when printed out in black and white.

The best way to learn how to use these options is to experiment—so go ahead and

play!

IMPORTING CHARTS AND GRAPHS INTO WORD

DOCUMENTS

SPSS allows you to copy charts directly into your word processor (e.g. Microsoft

Word). This is useful when you are preparing the ﬁ nal version of your report and

want to present some of your results in the form of a graph. Sometimes a graph will

present your results more simply and clearly than numbers in a box. Don’t go over-

board—use only for special effect. Make sure you modify the graph in SPSS to make

it as clear as possible before transferring it to Word.

Figure 7.1

Example of a Chart

Editor menu bar

Using graphs to describe and explore the data 81

Procedure for importing a chart into a Word document

Windows allows you to have more than one program open at a time. To

transfer between SPSS and Word, you will need to have both of these

programs open. It is possible to swap backwards and forwards between the

two just by clicking on the appropriate icon in the taskbar at the bottom of

your screen, or from the Window menu. This is like shufﬂ ing pieces of paper

around on your desk.

1. Start Word and open the ﬁ le in which you would like the graph to appear.

Click on the SPSS icon on the bottom of your screen to return to SPSS.

2. In SPSS make sure you have the Viewer window on the screen in front of

you.

3. Click once on the graph that you would like to copy. A border should

appear around the graph.

4. Click on Edit (from the menu at the top of the page) and then choose

Copy. This saves the chart to the clipboard (you won’t be able to see it,

however).

5. From the list of minimised programs at the bottom of your screen, click

on your Word document.

6. In the Word document, place your cursor where you wish to insert the

chart.

7. Click on Edit from the Word menu and choose Paste. Or just click on the

Paste icon on the top menu bar (it looks like a clipboard).

8. Click on File and then Save to save your Word document.

9. To move back to SPSS to continue with your analyses just click on the

SPSS icon, which should be listed at the bottom of your screen. With both

programs open you can just jump backwards and forwards between the

two programs, copying charts, tables etc. There is no need to close either

of the programs until you have ﬁ nished completely. Just remember to

save as you go along.

ADDITIONAL EXERCISES

Business

Data ﬁ le: staffsurvey4ED.sav. See Appendix for details of the data ﬁ le.

1. Generate a histogram to explore the distribution of scores on the Staff Satisfac-

tion Scale (totsatis).

82 Preliminary analyses

2. Generate a bar graph to assess the staff satisfaction levels for permanent versus

casual staff employed for less than or equal to 2 years, 3 to 5 years and 6 or more

years. The variables you will need are totsatis, employstatus and servicegp3.

3. Generate a scatterplot to explore the relationship between years of service and

staff satisfaction. Try ﬁ rst using the service variable (which is very skewed)

and then try again with the variable towards the bottom of the list of variables

(logservice). This new variable is a mathematical transformation (log 10) of the

original variable (service), designed to adjust for the severe skewness. This pro-

cedure is covered in Chapter 8.

4. Generate a boxplot to explore the distribution of scores on the Staff Satisfaction

Scale (totsatis) for the different age groups (age).

5. Generate a line graph to compare staff satisfaction for the different age groups

(use the agerecode variable) for permanent and casual staff.

Health

Data ﬁ le: sleep4ED.sav. See Appendix for details of the data ﬁ le.

1. Generate a histogram to explore the distribution of scores on the Epworth

Sleepiness Scale (ess).

2. Generate a bar graph to compare scores on the Sleepiness and Associated Sen-

sations Scale (totSAS) across three age groups (agegp3) for males and females

(gender).

3. Generate a scatterplot to explore the relationship between scores on the Epworth

Sleepiness Scale (ess) and the Sleepiness and Associated Sensations Scale (totSAS).

Ask for different markers for males and females (gender).

4. Generate a boxplot to explore the distribution of scores on the Sleepiness and

Associated Sensations Scale (totSAS) for people who report that they do/don’t

have a problem with their sleep (probsleeprec).

5. Generate a line graph to compare scores on the Sleepiness and Associated Sen-

sations Scale (totSAS) across the different age groups (use the agegp3 variable) for

males and females (gender).

83

8

Manipulating the data

Once you have entered the data and the data ﬁ le has been checked for accuracy, the

next step involves manipulating the raw data into a form that you can use to conduct

analyses and to test your hypotheses. Depending on the data ﬁ le, your variables of

interest and the type of research questions that you wish to address, this process may

include:

• adding up the scores from the items that make up each scale to give an overall

score for scales such as self-esteem, optimism, perceived stress etc. SPSS does this

quickly, easily and accurately—don’t even think about doing this by hand for each

separate case

• transforming skewed variables for analyses that require normally distributed

scores

• collapsing continuous variables (e.g. age) into categorical variables (e.g. young,

middle-aged and old) to do some analyses such as analysis of variance

• reducing or collapsing the number of categories of a categorical variable (e.g.

collapsing the marital status into just two categories representing people ‘in a

relationship’/‘not in a relationship’).

When you make the changes to the variables in your data ﬁ le, it is important that

you note this in your codebook. The other way that you can keep a record of all the

changes made to your data ﬁ le is to use the SPSS Syntax option that is available in all

SPSS procedures. I will describe this process ﬁ rst before demonstrating how to recode

and transform your variables.

Using Syntax to record procedures

As discussed previously in Chapter 3, SPSS has a Syntax Editor window that can

be used to record the commands generated using the Windows menus for each

84 Preliminary analyses

procedure. To access the syntax, follow the instructions shown in the procedure sections

to follow but stop before clicking the ﬁ nal OK button. Instead, click on the Paste button.

This will open a new window, the Syntax Editor, showing the commands you have

selected. Figure 8.1 shows part of the Syntax Editor window that was used to recode

items and compute the total scores used in survey4ED.sav. The complete syntax ﬁ le

(surveysyntax.sps) can be downloaded from the SPSS Survival Manual website.

The commands pasted to the Syntax Editor are not executed until you choose to

run them. To run the command, highlight the speciﬁ c command (making sure you

include the ﬁ nal full stop) and then click on the Run menu option or the arrow icon

from the menu bar. Alternatively, you can select the name of the analysis you wish to

run from the left-hand side of the screen.

Extra comments can be added to the syntax ﬁ le by starting them with an asterisk

(*). If you add comments, make sure you leave at least one line of space both before

and after syntax commands.

For each of the procedures described in the following sections, the syntax will also

be shown.

Figure 8.1

Example of a Syntax

Editor window

CALCULATING TOTAL SCALE SCORES

Before you can perform statistical analyses on your data set, you need to calculate total

scale scores for any scales used in your study. This involves two steps:

• Step 1: reverse any negatively worded items.

• Step 2: add together scores from all the items that make up the subscale or scale.

Manipulating the data 85

It is important that you understand the scales and measures that you are using for

your research. You should check with the scale’s manual or the journal article it was

published in to ﬁ nd out which items, if any, need to be reversed and how to go about

calculating a total score. Some scales consist of a number of subscales that either can,

or alternatively should not, be added together to give an overall score. It is important

that you do this correctly, and it is much easier to do it right the ﬁ rst time than to have

to repeat analyses later.

Important: you should do this only when you have a complete data ﬁ le as SPSS

does not update these commands when you add extra data.

Step 1: Reversing negatively worded items

In some scales the wording of particular items has been reversed to help prevent

response bias. This is evident in the Optimism Scale used in the survey (see Appendix).

Item 1 is worded in a positive direction (high scores indicate high optimism): ‘In

uncertain times I usually expect the best.’ Item 2, however, is negatively worded (high

scores indicate low optimism): ‘If something can go wrong for me it will.’ Items 4 and

6 are also negatively worded. The negatively worded items need to be reversed before

a total score can be calculated for this scale. We need to ensure that all items are scored

so that high scores indicate high levels of optimism.

The procedure for reversing items 2, 4 and 6 of the Optimism Scale is shown in

the table that follows. A ﬁ ve-point Likert-type scale was used for the Optimism Scale;

therefore, scores for each item can range from 1 (strongly disagree) to 5 (strongly

agree).

Although it is possible to rescore variables into the same variable name, we will

ask SPSS to create new variables rather than overwrite the existing data. This is a

much safer option, and it retains our original data unchanged.

If you wish to follow along with the instructions shown below, you should open

survey4ED.sav.

1. From the menu at the top of the screen, click on Transform, then click on

Recode Into Different Variables.

2. Select the items you want to reverse (op2, op4, op6). Move these into the

Input Variable—Output Variable box.

3. Click on the ﬁ rst variable (op2) and type a new name in the Output

Variable section on the right-hand side of the screen and then click the

Change button. I have used Rop2 in the existing data ﬁ le. If you wish to

create your own (rather than overwrite the ones already in the data ﬁ le),

use another name (e.g. revop2). Repeat for each of the other variables

you wish to reverse (op4 and op6).

4. Click on the Old and new values button.

86 Preliminary analyses

In the Old value section, type 1 in the Value box.

In the New value section, type 5 in the Value box (this will change all

scores that were originally scored as 1 to a 5).

5. Click on Add. This will place the instruction (1 → 5) in the box labelled Old

> New.

6. Repeat the same procedure for the remaining scores. For example:

Old value—type in 2 New value—type in 4 Add

Old value—type in 3 New value—type in 3 Add

Old value—type in 4 New value—type in 2 Add

Old value—type in 5 New value—type in 1 Add

Always double-check the item numbers that you specify for recoding and

the old and new values that you enter. Not all scales use a ﬁ ve-point scale;

some have four possible responses, some six and some seven. Check that

you have reversed all the possible values for your particular scale.

7. Click on Continue and then OK (or on Paste if you wish to paste this

command to the Syntax Editor window). To execute it after pasting to the

Syntax Editor, highlight the command and select Run from the menu.

The syntax generated for this command is:

RECODE

op2 op4 op6

(1=5) (2=4) (3=3) (4=2) (5=1) INTO Rop2 Rop4 Rop6 .

EXECUTE .

The new variables with reversed scores should be found at the end of the data ﬁ le.

Check this in your Data Editor window, choose the Variable View tab and go down to

the bottom of the list of variables. In the survey4ED.sav ﬁ le you will see a whole series

of variables with an R at the front of the variable name. These are the items that I have

reversed. If you follow the instructions shown above, you should see yours at the very

bottom with ‘rev’ at the start of each. It is important to check your recoded variables

to see what effect the recode had on the values. For the ﬁ rst few cases in your data set,

take note of the scores on the original variables and then check the corresponding

reversed variables to ensure that it worked properly.

Step 2: Adding up the total scores for the scale

After you have reversed the negatively worded items in the scale, you will be ready to

calculate total scores for each subject.

Important: you should do this only when you have a complete data ﬁ le as SPSS

does not update this command when you add extra data.

Manipulating the data 87

Procedure for calculating total scale scores

1. From the menu at the top of the screen, click on Transform, then click on

Compute Variable.

2. In the Target Variable box, type in the new name you wish to give to the

total scale scores. (It is useful to use a T preﬁ x to indicate total scores, as

this makes them easier to ﬁ nd in the list of variables when you are doing

your analyses.)

Important: make sure you do not accidentally use a variable name that

has already been used in the data set. If you do, you will lose all the

original data—potential disaster—so check your codebook.

3. Click on the Type and Label button. Click in the Label box and type in a

description of the scale (e.g. total optimism). Click on Continue.

4. From the list of variables on the left-hand side, click on the ﬁ rst item in

the scale (op1).

5. Click on the arrow button to move it into the Numeric Expression box.

6. Click on + on the calculator.

7. Repeat the process until all scale items appear in the box. In this example

we would select the unreversed items ﬁ rst (op3, op5) and then the

reversed items (obtained in the previous procedure), which are located at

the bottom of the list of variables (Rop2, Rop4, Rop6).

8. The complete numeric expression should read as follows:

op1+op3+op5+Rop2+Rop4+Rop6.

9. Double-check that all items are correct and that there are + signs in the

right places. Click OK (or on Paste if you wish to paste this command

to the Syntax Editor window). To execute it after pasting to the Syntax

Editor, highlight the command and select Run from the menu.

The syntax for this command is:

COMPUTE toptim = op1+op3+op5+Rop2+Rop4+Rop6 .

EXECUTE .

This will create a new variable at the end of your data set called TOPTIM. Scores

for each person will consist of the addition of scores on each of the items op1 to op6

(with recoded items where necessary). If any items had missing data, the overall score

will also be missing. This is indicated by a full stop instead of a score in the data ﬁ le.

You will notice in the literature that some researchers go a step further and divide the

total scale score by the number of items in the scale. This can make it a little easier to

88 Preliminary analyses

interpret the scores of the total scale because it is back in the original scale used for each

of the items (e.g. from 1 to 5 representing strongly disagree to strongly agree). To do this,

you also use the Transform, Compute menu of SPSS. This time you will need to specify

a new variable name and then type in a suitable formula (e.g. TOPTIM/6).

Always record details of any new variables that you create in your codebook.

Specify the new variable’s name, what it represents and full details of what was done

to calculate it. If any items were reversed, this should be speciﬁ ed along with details

of which items were added to create the score. It is also a good idea to include the

possible range of scores for the new variable in the codebook (see the Appendix). This

gives you a clear guide when checking for any out-of-range values.

After creating a new variable, it is important to run Descriptives on this new scale

to check that the values are appropriate (see Chapter 5). It also helps you get a feel for

the distribution of scores on your new variable.

• Check back with the questionnaire—what is the possible range of scores that could

be recorded? For a ten-item scale, using a response scale from 1 to 4, the minimum

value would be 10 and the maximum value would be 40. If a person answered 1 to

every item, that overall score would be 10 × 1 = 10. If a person answered 4 to each

item, that score would be 10 × 4 = 40.

• Check the output from Descriptives to ensure that there are no out-of-range

cases (see Chapter 5).

• Compare the mean score on the scale with values reported in the literature. Is

your value similar to that obtained in previous studies? If not, why not? Have you

done something wrong in the recoding? Or is your sample different from that

used in other studies?

You should also run other analyses to check the distribution of scores on your new

total scale variable:

• Check the distribution of scores using skewness and kurtosis (see Chapter 6).

• Obtain a histogram of the scores and inspect the spread of scores. Are they

normally distributed? If not, you may need to consider ‘transforming’ the scores

for some analyses (this is discussed later in this chapter).

COLLAPSING A CONTINUOUS VARIABLE INTO GROUPS

For some analyses or when you have very skewed distributions, you may wish to divide

the sample into equal groups according to respondents’ scores on some variable (e.g.

to give low, medium and high scoring groups).

To illustrate this process, I will use the survey4ED.sav ﬁ le that is included on the

Manipulating the data 89

website that accompanies this book (see p. viii and the Appendix for details). I will use

Visual Binning to identify suitable cut-off points to break the continuous variable age

into three approximately equal groups. The same technique could be used to create

a ‘median split’; that is, to divide the sample into two groups, using the median as

the cut-off point. Once the cut-off points are identiﬁ ed, Visual Binning will create

a new categorical variable that has only three values corresponding to the three age

ranges chosen. This technique leaves the original variable age, measured as a continu-

ous variable, intact so that you can use it for other analyses.

Procedure for collapsing a continuous variable into groups

1. From the menu at the top of the screen, click on Transform and choose

Visual Binning.

2. Select the continuous variable that you want to use (e.g. age). Transfer it

into the Variables to Bin box. Click on the Continue button.

3. In the Visual Binning screen, a histogram showing the distribution of age

scores should appear.

4. In the section at the top labelled Binned Variable, type the name for the

new categorical variable that you will create (e.g. Agegp3). You can also

change the suggested label that is shown (e.g. age in 3 groups).

5. Click on the button labelled Make Cutpoints. In the dialogue box that

appears, click on the option Equal Percentiles Based on Scanned Cases.

In the box Number of Cutpoints, specify a number one less than the

number of groups that you want (e.g. if you want three groups, type in

2 for cutpoints). In the Width (%) section below, you will then see 33.33

appear. This means that SPSS will try to put 33.3 per cent of the sample in

each group. Click on the Apply button.

6. Click on the Make Labels button back in the main dialogue box. This will

automatically generate value labels for each of the new groups created.

7. Click on OK (or on Paste if you wish to paste this command to the Syntax

Editor window). To execute it after pasting to the Syntax Editor, highlight

the command and select Run from the menu.

The syntax generated by this command is:

RECODE age

( MISSING = COPY )

( LO THRU 29 =1 )

( LO THRU 44 =2 )

( LO THRU HI = 3 )

( ELSE = SYSMIS ) INTO agegp3.

90 Preliminary analyses

VARIABLE LABELS agegp3 ‘age in 3 groups’.

FORMAT agegp3 (F5.0).

VALUE LABELS agegp3

1 ‘<= 29’

2 ‘30—44’

3 ‘45+’.

MISSING VALUES agegp3 ( ).

VARIABLE LEVEL agegp3 ( ORDINAL ).

EXECUTE.

A new variable (Agegp3) should appear at the end of your data ﬁ le. Go back

to your Data Editor window, choose the Variable View tab, and it should be

at the bottom. To check the number of cases in each of the categories of your

newly created variable (Agegp3), go to Analyze and select Descriptives, then

Frequencies.

COLLAPSING THE NUMBER OF CATEGORIES OF A

CATEGORICAL VARIABLE

There are some situations where you may want to reduce or collapse the number of

categories of a categorical variable. You may want to do this for research or theoretical

reasons (e.g. collapsing the marital status into just two categories representing people

‘in a relationship’/‘not in a relationship’), or you may make the decision after looking

at the nature of the data. For example, after running Descriptive Statistics you may

ﬁ nd you have only a few people in your sample who fall into a particular category (e.g.

for our education variable, we only have two people in our ﬁ rst category, ‘primary

school’). As it stands, this variable could not appropriately be used in many of the

statistical analyses covered later in the book. We could decide just to remove these

people from the sample, or we could recode them to combine them with the next

category (some secondary school). We would have to relabel the variable so that it

represented people who did not complete secondary school.

The procedure for recoding a categorical variable is shown below. It is very impor-

tant to note that here we are creating a new additional variable (so that we keep our

original data intact).

Procedure for recoding a categorical variable

1. From the menu at the top of the screen, click on Transform, then

on Recode into Different Variables. (Make sure you select ‘different

variables’, as this retains the original variable for other analyses.)

2. Select the variable you wish to recode (e.g. educ). In the Name box, type

Manipulating the data 91

a name for the new variable that will be created (e.g. educrec). Type in

an extended label if you wish in the Label section. Click on the button

labelled Change.

3. Click on the button labelled Old and New Values.

4. In the section Old Value, you will see a box labelled Value. Type in the ﬁ rst

code or value of your current variable (e.g. 1). In the New Value section,

type in the new value that will be used (or, if the same one is to be used,

type that in). In this case I will recode to the same value, so I will type 1 in

both the Old Value and New Value sections. Click on the Add button.

5. For the second value, I would type 2 in the Old Value but in the New Value

I would type 1. This will recode all the values of both 1 and 2 from the

original coding into one group in the new variable to be created with a

value of 1.

6. For the third value of the original variable, I would type 3 in the Old

Value and 2 in the New Value. This is just to keep the values in the new

variable in sequence. Click on Add. Repeat for all the remaining values of

the original values. In the table Old > New, you should see the following

codes for this example: 1→1; 2→1; 3→2; 4→3; 5→4; 6→5.

7. Click on Continue and then on OK (or on Paste if you wish to paste this

command to the Syntax Editor window). To execute it after pasting to the

Syntax Editor, highlight the command and select Run from the menu.

8. Go to your Data Editor window and choose the Variable View tab. Type

in appropriate values labels to represent the new values (1=did not

complete high school, 2=completed high school, 3=some additional

training, 4=completed undergrad uni, 5=completed postgrad uni).

Remember, these will be different from the codes used for the original

variable, and it is important that you don’t mix them up.

The syntax generated by this command is:

RECODE

educ

(1=1) (2=1) (3=2) (4=3) (5=4) (6=5) INTO educrec .

EXECUTE .

When you recode a variable, make sure you run Frequencies on both the old

variable (educ) and the newly created variable (educrec:, which appears at the end of

your data ﬁ le). Check that the frequencies reported for the new variable are correct.

For example, for the newly created educrec variable, we should now have 2+53=55 in

92 Preliminary analyses

the ﬁ rst group. This represents the two people who ticked 1 on the original variable

(primary school) and the 53 people who ticked 2 (some secondary school).

The Recode procedure demonstrated here could be used for a variety of purposes.

You may ﬁ nd later, when you come to do your statistical analyses, that you will need

to recode the values used for a variable. For example, in Chapter 14 (Logistic regres-

sion) you may need to recode variables originally coded 1=yes, 2=no to a new coding

system 1=yes, 0=no. This can be achieved in the same way as described in the previous

procedures section. Just be very clear before you start on what your original values are,

and what you want the new values to be.

TRANSFORMING VARIABLES

Often when you check the distribution of scores on a scale or measure (e.g. self-

esteem, anxiety) you will ﬁ nd (to your dismay!) that the scores do not fall in a nice,

normally distributed curve. Sometimes scores will be positively skewed, where most

of the respondents record low scores on the scale (e.g. depression). Sometimes you

will ﬁ nd a negatively skewed distribution, where most scores are at the high end (e.g.

self-esteem). Given that many of the parametric statistical tests assume normally

distributed scores, what do you do about these skewed distributions?

One of the choices you have is to abandon the use of parametric statistics (e.g.

Pearson correlation, analysis of variance) and instead choose to use non-parametric

alternatives (e.g. Spearman’s rho, Kruskal-Wallis). SPSS includes a number of useful

non-parametric techniques in its package. These are discussed in Chapter 16.

Another alternative, when you have a non-normal distribution, is to ‘transform’ your

variables. This involves mathematically modifying the scores using various formulas until

the distribution looks more normal. There are a number of different types of transfor-

mation, depending on the shape of your distribution. There is considerable controversy

concerning this approach in the literature, with some authors strongly supporting, and

others arguing against, transforming variables to better meet the assumptions of the

various parametric techniques. For a discussion of the issues and the approaches to

transformation, you should read Chapter 4 in Tabachnick and Fidell (2007).

In Figure 8.2 some of the more common problems are represented, along with

the type of transformation recommended by Tabachnick and Fidell (2007, p. 87). You

should compare your distribution with those shown, and decide which picture it most

closely resembles. I have also given a nasty-looking formula beside each of the suggested

transformations. Don’t let this throw you—these are just formulas that SPSS will use

on your data, giving you a new, hopefully normally distributed variable to use in your

analyses. In the procedures section to follow, you will be shown the SPSS procedure for

this. Before attempting any of these transformations, however, it is important that you

read Tabachnick and Fidell (2007, Chapter 4), or a similar text, thoroughly.

Manipulating the data 93

Figure 8.2

Distribution

of scores and

suggested

transformations

Square root

Formula: new variable = SQRT (old variable)

Logarithm

Formula: new variable = LG10 (old variable)

Inverse

Formula: new variable = 1 / (old variable)

Reﬂ ect and square root

Formula: new variable = SQRT (K – old variable) where

K = largest possible value +1

Reﬂ ect and logarithm

Formula: new variable = LG10 (K – old variable) where

K = largest possible value +1

Reﬂ ect and inverse

Formula: new variable = 1 / (K – old variable) where

K = largest possible value +1

94 Preliminary analyses

Procedure for transforming variables

1. From the menu at the top of the screen, click on Transform, then click on

Compute Variable.

2. Target Variable. In this box, type in a new name for the variable. Try to

include an indication of the type of transformation and the original name

of the variable. For example, for a variable called tnegaff I would make this

new variable sqtnegaff, if I had performed a square root. Be consistent in

the abbreviations that you use for each of your transformations.

3. Functions. Listed are a wide range of possible actions you can use. You

need to choose the most appropriate transformation for your variable.

Look at the shape of your distribution; compare it with those in

Figure 8.2. Take note of the formula listed next to the picture that

matches your distribution. This is the one that you will use.

4. Transformations involving square root or logarithm. In the Function

group box, click on Arithmetic, and scan down the list that shows up in

the bottom box until you ﬁ nd the formula you need (e.g. Sqrt or Lg10).

Highlight the one you want and click on the up arrow. This moves the

formula into the Numeric Expression box. You will need to tell it which

variable you want to recalculate. Find it in the list of variables and click

on the arrow to move it into the Numeric Expression box. If you prefer,

you can just type the formula in yourself without using the Functions or

Variables list. Just make sure you spell everything correctly.

5. Transformations involving Reﬂ ect. You need to ﬁ nd the value K for

your variable. This is the largest value that your variable can have (see

your codebook) + 1. Type this number in the Numeric Expression box.

Complete the remainder of the formula using the Functions box, or

alternatively type it in yourself.

6. Transformations involving Inverse. To calculate the inverse, you need to

divide your scores into 1. So, in the Numeric Expression box type in 1, then

type / and then your variable or the rest of your formula (e.g. 1/tslfest).

7. Check the ﬁ nal formula in the Numeric Expression box. Write this down

in your codebook next to the name of the new variable you created.

8. Click on the button Type and Label. Under Label, type in a brief description

of the new variable (or you may choose to use the actual formula you used).

9. Check in the Target Variable box that you have given your new variable

a new name, not the original one. If you accidentally put the old variable

name, you will lose all your original scores. So, always double-check.

10. Click on OK (or on Paste if you wish to paste this command to the Syntax

Editor window). To execute it after pasting to the Syntax Editor, highlight

Manipulating the data 95

the command and select Run from the menu. A new variable will be

created and will appear at the end of your data ﬁ le.

11. Run Analyze, Frequencies to check the skewness and kurtosis values for

your old and new variables. Have they improved?

12. Under Frequencies, click on the Charts button and select Histogram

to inspect the distribution of scores on your new variable. Has the

distribution improved? If not, you may need to consider a different type

of transformation.

If none of the transformations work, you may need to consider using non-para-

metric techniques to analyse your data (see Chapter 16). Alternatively, for very skewed

variables you may wish to divide your continuous variable into a number of discrete

groups. Instructions for doing this are presented earlier in this chapter.

ADDITIONAL EXERCISES

Business

Data ﬁ le: staffsurvey4ED.sav. See Appendix for details of the data ﬁ le.

1. Practise the procedures described in this chapter to add up the total scores for

a scale using the items that make up the Staff Satisfaction Survey. You will need

to add together the items that assess agreement with each item in the scale (i.e.

Q1a+Q2a+Q3a … to Q10a). Name your new variable staffsatis.

2. Check the descriptive statistics for your new total score (staffsatis) and compare

this with the descriptives for the variable totsatis, which is already in your data ﬁ le.

This is the total score that I have already calculated for you.

3. What are the minimum possible and maximum possible scores for this new

variable? Tip: check the number of items in the scale and the number of response

points on each item (see Appendix).

4. Check the distribution of the variable service by generating a histogram. You will

see that it is very skewed, with most people clustered down the low end (with less

than 2 years’ service) and a few people stretched up at the very high end (with

more than 30 years’ service). Check the shape of the distribution against those

displayed in Figure 8.2 and try a few different transformations. Remember to

check the distribution of the new transformed variables you create. Are any of the

new variables more ‘normally’ distributed?

5. Collapse the years of service variable (service) into three groups using the Visual

Binning procedure from the Transform menu. Use the Make Cutpoints button

and ask for Equal Percentiles. In the section labelled Number of Cutpoints,

specify 2. Call your new variable gp3service to distinguish it from the variable

96 Preliminary analyses

I have already created in the data ﬁ le using this procedure (service3gp). Run

Frequencies on your newly created variable to check how many cases are in each

group.

Health

Data ﬁ le: sleep4ED.sav. See Appendix for details of the data ﬁ le.

1. Practise the procedures described in this chapter to add up the total scores for a

scale using the items that make up the Sleepiness and Associated Sensations Scale.

You will need to add together the items fatigue, lethargy, tired, sleepy, energy. Call

your new variable sleeptot. Please note: none of these items needs to be reversed

before being added.

2. Check the descriptive statistics for your new total score (sleeptot) and compare

them with the descriptives for the variable totSAS, which is already in your data

ﬁ le. This is the total score that I have already calculated for you.

3. What are the minimum possible and maximum possible scores for this new

variable? Tip: check the number of items in the scale and the number of response

points on each item (see Appendix).

4. Check the distribution (using a histogram) of the variable that measures the

number of cigarettes smoked per day by the smokers in the sample (smokenum).

You will see that it is very skewed, with most people clustered down the low end

(with less than 10 per day) and a few people stretched up at the very high end

(with more than 70 per day). Check the shape of the distribution against those

displayed in Figure 8.2 and try a few different transformations. Remember to

check the distribution of the new transformed variables you create. Are any of the

new transformed variables more ‘normally’ distributed?

5. Collapse the age variable (age) into three groups using the Visual Binning pro-

cedure from the Transform menu. Use the Make Cutpoints button and ask for

Equal Percentiles. In the section labelled Number of Cutpoints, specify 2. Call

your new variable gp3age to distinguish it from the variable I have already created

in the data ﬁ le using this procedure (age3gp). Run Frequencies on your newly

created variable to check how many cases are in each group.

97

9

Checking the reliability

of a scale

When you are selecting scales to include in your study, it is important to ﬁ nd scales

that are reliable. There are a number of different aspects to reliability (see discussion

of this in Chapter 1). One of the main issues concerns the scale’s internal consistency.

This refers to the degree to which the items that make up the scale ‘hang together’.

Are they all measuring the same underlying construct? One of the most commonly

used indicators of internal consistency is Cronbach’s alpha coefﬁ cient. Ideally, the

Cronbach alpha coefﬁ cient of a scale should be above .7 (DeVellis 2003). Cronbach

alpha values are, however, quite sensitive to the number of items in the scale. With

short scales (e.g. scales with fewer than ten items) it is common to ﬁ nd quite low

Cronbach values (e.g. .5). In this case, it may be more appropriate to report the mean

inter-item correlation for the items. Briggs and Cheek (1986) recommend an optimal

range for the inter-item correlation of .2 to .4.

The reliability of a scale can vary depending on the sample. It is therefore necessary

to check that each of your scales is reliable with your particular sample. This infor-

mation is usually reported in the Method section of your research paper or thesis. If

your scale contains some items that are negatively worded (common in psychological

measures), these need to be ‘reversed’ before checking reliability. Instructions on how

to do this are provided in Chapter 8.

Make sure that you check with the scale’s manual (or the journal article in which it

is reported) for instructions concerning the need to reverse items and for information

on any subscales. Sometimes scales contain a number of subscales that may or may

not be combined to form a total scale score. If necessary, the reliability of each of the

subscales and the total scale will need to be calculated.

If you are developing your own scale for use in your study, make sure you read

widely on the principles and procedures of scale development. There are some good

easy-to-read books on the topic, including Streiner & Norman (2008), DeVellis (2003)

and Kline (2005).

98 Preliminary analyses

DETAILS OF EXAMPLE

To demonstrate this technique, I will be using the survey4ED.sav data ﬁ le included on

the website accompanying this book. Full details of the study, the questionnaire and

scales used are provided in the Appendix. If you wish to follow along with the steps

described in this chapter, you should start SPSS and open the ﬁ le survey4ED.sav. In

the procedure described below, I will explore the internal consistency of one of the

scales from the questionnaire. This is the Satisfaction with Life Scale (Pavot, Diener,

Colvin & Sandvik 1991), which is made up of ﬁ ve items. In the data ﬁ le these items are

labelled as lifsat1, lifsat2, lifsat3, lifsat4, lifsat5.

Procedure for checking the reliability of a scale

Important: before starting, you should check that all negatively worded items

in your scale have been reversed (see Chapter 8). If you don’t do this, you will

ﬁ nd that you have very low (and incorrect) Cronbach alpha values. In this

case, none of the items needs to be rescored.

1. From the menu at the top of the screen, click on Analyze, select Scale,

then Reliability Analysis.

2. Click on all of the individual items that make up the scale (e.g. lifsat1,

lifsat2, lifsat3, lifsat4, lifsat5). Move these into the box marked Items.

3. In the Model section, make sure Alpha is selected.

4. In the Scale label box, type in the name of the scale or subscale (Life

Satisfaction).

5. Click on the Statistics button. In the Descriptives for section, select

Item, Scale, and Scale if item deleted. In the Inter-Item section, click on

Correlations. In the Summaries section, click on Correlations.

6. Click on Continue and then OK (or on Paste to save to Syntax Editor).

The syntax from this procedure is:

RELIABILITY

/VARIABLES=lifsat1 lifsat2 lifsat3 lifsat4 lifsat5

/SCALE(‘Life Satisfaction’) ALL/MODEL=ALPHA

/STATISTICS=DESCRIPTIVE SCALE CORR

/SUMMARY=TOTAL CORR .

The output generated from this procedure is shown below.

Checking the reliability of a scale 99

100 Preliminary analyses

INTERPRETING THE OUTPUT FROM RELIABILITY

• Check that the number of cases is correct (in the Case Processing Summary table)

and that the number of items is correct (in the Reliability Statistics table).

• Check the Inter-Item Correlation Matrix for negative values. All values should

be positive, indicating that the items are measuring the same underlying charac-

teristic. The presence of negative values could indicate that some of the items

have not been correctly reverse scored. Incorrect scoring would also show up in

the Item-Total Statistics table with negative values for the Corrected-Item Total

Correlation values. These should be checked carefully if you obtain a lower than

expected Cronbach alpha value. (Check what other researchers report for the

scale.)

• Check the Cronbach’s Alpha value shown in the Reliability Statistics table. In

this example the value is .89, suggesting very good internal consistency reliability

for the scale with this sample. Values above .7 are considered acceptable; however,

values above .8 are preferable.

• The Corrected Item-Total Correlation values shown in the Item-Total Statistics

table give you an indication of the degree to which each item correlates with the

total score. Low values (less than .3) here indicate that the item is measuring some-

thing different from the scale as a whole. If your scale’s overall Cronbach alpha is

too low (e.g. less than .7) and you have checked for incorrectly scored items, you

may need to consider removing items with low item-total correlations.

• In the column headed Alpha if Item Deleted, the impact of removing each item

from the scale is given. Compare these values with the ﬁ nal alpha value obtained. If

any of the values in this column are higher than the ﬁ nal alpha value, you may want

to consider removing this item from the scale. This is useful if you are developing a

scale, but if you are using established, validated scales, removal of items means that

you could not compare your results with other studies using the scale.

• For scales with a small number of items (e.g. less than 10), it is sometimes difﬁ cult

to get a decent Cronbach alpha value and you may wish to consider reporting the

mean inter-item correlation value, which is shown in the Summary Item Statis-

tics table. In this case the mean inter-item correlation is .63, with values ranging

from .48 to .76. This suggests quite a strong relationship among the items. For

many scales, this is not the case.

PRESENTING THE RESULTS FROM RELIABILITY

You would normally report the internal consistency of the scales that you are using in

your research in the Method section of your report, under the heading Measures, or

Materials. After describing the scale (number of items, response scale used, history of

Checking the reliability of a scale 101

use), you should include a summary of reliability information reported by the scale

developer and other researchers, and then a sentence to indicate the results for your

sample. For example:

According to Pavot, Diener, Colvin and Sandvik (1991), the Satisfaction with Life

Scale has good internal consistency, with a Cronbach alpha coefﬁ cient reported

of .85. In the current study, the Cronbach alpha coefﬁ cient was .89.

ADDITIONAL EXERCISES

Business

Data ﬁ le: staffsurvey4ED.sav. See Appendix for details of the data ﬁ le.

1. Check the reliability of the Staff Satisfaction Survey, which is made up of the

agreement items in the data ﬁ le: Q1a to Q10a. None of the items of this scale

needs to be reversed.

Health

Data ﬁ le: sleep4ED.sav. See Appendix for details of the data ﬁ le.

1. Check the reliability of the Sleepiness and Associated Sensations Scale, which is

made up of items fatigue, lethargy, tired, sleepy, energy. None of the items of this

scale needs to be reversed.

102

10

Choosing the

right statistic

One of the most difﬁ cult (and potentially fear-inducing) parts of the research process

for most research students is choosing the correct statistical technique to analyse their

data. Although most statistics courses teach you how to calculate a correlation coef-

ﬁ cient or perform a t-test, they typically do not spend much time helping students learn

how to choose which approach is appropriate to address particular research questions.

In most research projects it is likely that you will use quite a variety of different types of

statistics, depending on the question you are addressing and the nature of the data that

you have. It is therefore important that you have at least a basic understanding of the

different statistics, the type of questions they address and their underlying assumptions

and requirements.

So, dig out your statistics texts and review the basic techniques and the principles

underlying them. You should also look through journal articles on your topic and

identify the statistical techniques used in these studies. Different topic areas may make

use of different statistical approaches, so it is important that you ﬁ nd out what other

researchers have done in terms of data analysis. Look for long, detailed journal articles

that clearly and simply spell out the statistics that were used. Collect these together in

a folder for handy reference. You might also ﬁ nd them useful later when considering

how to present the results of your analyses.

In this chapter we will look at the various statistical techniques that are available,

and I will then take you step by step through the decision-making process. If the whole

statistical process sends you into a panic, just think of it as choosing which recipe you

will use to cook dinner tonight. What ingredients do you have in the refrigerator,

what type of meal do you feel like (soup, roast, stir-fry, stew), and what steps do you

have to follow? In statistical terms, we will look at the type of research questions you

have, which variables you want to analyse, and the nature of the data itself. If you take

this process step by step, you will ﬁ nd the ﬁ nal decision is often surprisingly simple.

Once you have determined what you have and what you want to do, there often is only

Choosing the right statistic 103

one choice. The most important part of this whole process is clearly spelling out what

you have and what you want to do with it.

OVERVIEW OF THE DIFFERENT STATISTICAL

TECHNIQUES

This section is broken into two main parts. First, we will look at the techniques used

to explore the relationship among variables (e.g. between age and optimism), followed

by techniques you can use when you want to explore the differences between groups

(e.g. sex differences in optimism scores). I have separated the techniques into these

two sections, as this is consistent with the way in which most basic statistics texts are

structured and how the majority of students will have been taught basic statistics. This

tends to somewhat artiﬁ cially emphasise the difference between these two groups of

techniques. There are, in fact, many underlying similarities between the various statis-

tical techniques, which is perhaps not evident on initial inspection. A full discussion

of this point is beyond the scope of this book. If you would like to know more, I

would suggest you start by reading Chapter 17 of Tabachnick and Fidell (2007). That

chapter provides an overview of the General Linear Model, under which many of the

statistical techniques can be considered.

I have deliberately kept the summaries of the different techniques brief and simple,

to aid initial understanding. This chapter certainly does not cover all the different

techniques available, but it does give you the basics to get you started and to build

your conﬁ dence.

Exploring relationships

Often in survey research you will not be interested in differences between groups, but

instead in the strength of the relationship between variables. There are a number of

different techniques that you can use.

Correlation

Pearson correlation or Spearman correlation is used when you want to explore the

strength of the relationship between two continuous variables. This gives you an indi-

cation of both the direction (positive or negative) and the strength of the relationship.

A positive correlation indicates that as one variable increases, so does the other. A

negative correlation indicates that as one variable increases, the other decreases. This

topic is covered in Chapter 11.

Partial correlation

Partial correlation is an extension of Pearson correlation—it allows you to control for

the possible effects of another confounding variable. Partial correlation ‘removes’ the

104 Preliminary analyses

effect of the confounding variable (e.g. socially desirable responding), allowing you to

get a more accurate picture of the relationship between your two variables of interest.

Partial correlation is covered in Chapter 12.

Multiple regression

Multiple regression is a more sophisticated extension of correlation and is used when

you want to explore the predictive ability of a set of independent variables on one

continuous dependent measure. Different types of multiple regression allow you to

compare the predictive ability of particular independent variables and to ﬁ nd the best

set of variables to predict a dependent variable. See Chapter 13.

Factor analysis

Factor analysis allows you to condense a large set of variables or scale items down to a

smaller, more manageable number of dimensions or factors. It does this by summaris-

ing the underlying patterns of correlation and looking for ‘clumps’ or groups of closely

related items. This technique is often used when developing scales and measures, to

identify the underlying structure. See Chapter 15.

Summary

All of the analyses described above involve exploration of the relationship between

continuous variables. If you have only categorical variables, you can use the Chi Square

Test for Relatedness or Independence to explore their relationship (e.g. if you wanted

to see whether gender inﬂ uenced clients’ dropout rates from a treatment program). In

this situation, you are interested in the number of people in each category (males and

females who drop out of/complete the program) rather than their score on a scale.

Some additional techniques you should know about, but which are not covered in this

text, are described below. For more information on these, see Tabachnick and Fidell

(2007). These techniques are as follows:

• Discriminant function analysis is used when you want to explore the predictive ability

of a set of independent variables, on one categorical dependent measure. That is,

you want to know which variables best predict group membership. The dependent

variable in this case is usually some clear criterion (passed/failed, dropped out of/

continued with treatment). See Chapter 9 in Tabachnick and Fidell (2007).

• Canonical correlation is used when you wish to analyse the relationship between

two sets of variables. For example, a researcher might be interested in how a

variety of demographic variables relate to measures of wellbeing and adjustment.

See Chapter 12 in Tabachnick and Fidell (2007).

• Structural equation modelling is a relatively new, and quite sophisticated, technique

that allows you to test various models concerning the interrelationships among a set

Choosing the right statistic 105

of variables. Based on multiple regression and factor analytic techniques, it allows you

to evaluate the importance of each of the independent variables in the model and to

test the overall ﬁ t of the model to your data. It also allows you to compare alterna-

tive models. SPSS does not have a structural equation modelling module, but it does

support an ‘add on’ called AMOS. See Chapter 14 in Tabachnick and Fidell (2007).

Exploring differences between groups

There is another family of statistics that can be used when you want to ﬁ nd out

whether there is a statistically signiﬁ cant difference among a number of groups. The

parametric versions of these tests, which are suitable when you have interval-scaled

data with normal distribution of scores, are presented below, along with the non-

parametric alternative.

T-tests

T-tests are used when you have two groups (e.g. males and females) or two sets of

data (before and after), and you wish to compare the mean score on some continuous

variable. There are two main types of t-tests. Paired sample t-tests (also called repeated

measures) are used when you are interested in changes in scores for participants tested

at Time 1, and then again at Time 2 (often after some intervention or event). The

samples are ‘related’ because they are the same people tested each time. Independent

sample t-tests are used when you have two different (independent) groups of people

(males and females), and you are interested in comparing their scores. In this case,

you collect information on only one occasion but from two different sets of people.

T-tests are covered in Chapter 17. The non-parametric alternatives, Mann-Whitney

U Test and Wilcoxon Signed Rank Test, are presented in Chapter 16.

One-way analysis of variance

One-way analysis of variance is similar to a t-test, but is used when you have two or more

groups and you wish to compare their mean scores on a continuous variable. It is called

one-way because you are looking at the impact of only one independent variable on your

dependent variable. A one-way analysis of variance (ANOVA) will let you know whether

your groups differ, but it won’t tell you where the signiﬁ cant difference is (gp1/gp3,

gp2/gp3 etc.). You can conduct post-hoc comparisons to ﬁ nd out which groups are

signiﬁ cantly different from one another. You could also choose to test differences between

speciﬁ c groups, rather than comparing all the groups, by using planned comparisons.

Similar to t-tests, there are two types of one-way ANOVAs: repeated measures ANOVA

(same people on more than two occasions), and between-groups (or independent

samples) ANOVA, where you are comparing the mean scores of two or more different

groups of people. One-way ANOVA is covered in Chapter 18, while the non-parametric

alternatives (Kruskal-Wallis Test and Friedman Test) are presented in Chapter 16.

106 Preliminary analyses

Two-way analysis of variance

Two-way analysis of variance allows you to test the impact of two independent vari-

ables on one dependent variable. The advantage of using a two-way ANOVA is that it

allows you to test for an interaction effect—that is, when the effect of one indepen-

dent variable is inﬂ uenced by another; for example, when you suspect that optimism

increases with age, but only for males.

It also tests for ‘main effects’—that is, the overall effect of each independent

variable (e.g. sex, age). There are two different two-way ANOVAs: between-groups

ANOVA (when the groups are different) and repeated measures ANOVA (when the

same people are tested on more than one occasion). Some research designs combine

both between-groups and repeated measures in the one study. These are referred to

as ‘Mixed Between-Within Designs’, or ‘Split Plot’. Two-way ANOVA is covered in

Chapter 19. Mixed designs are covered in Chapter 20.

Multivariate analysis of variance

Multivariate analysis of variance (MANOVA) is used when you want to compare

your groups on a number of different, but related, dependent variables; for example,

comparing the effects of different treatments on a variety of outcome measures (e.g.

anxiety, depression). Multivariate ANOVA can be used with one-way, two-way and

higher factorial designs involving one, two or more independent variables. MANOVA

is covered in Chapter 21.

Analysis of covariance

Analysis of covariance (ANCOVA) is used when you want to statistically control for

the possible effects of an additional confounding variable (covariate). This is useful

when you suspect that your groups differ on some variable that may inﬂ uence the

effect that your independent variables have on your dependent variable. To be sure

that it is the independent variable that is doing the inﬂ uencing, ANCOVA statistically

removes the effect of the covariate. Analysis of covariance can be used as part of a

one-way, two-way or multivariate design. ANCOVA is covered in Chapter 22.

THE DECISION-MAKING PROCESS

Having had a look at the variety of choices available, it is time to choose which tech-

niques are suitable for your needs. In choosing the right statistic, you will need to

consider a number of different factors. These include consideration of the type of

question you wish to address, the type of items and scales that were included in your

questionnaire, the nature of the data you have available for each of your variables and

the assumptions that must be met for each of the different statistical techniques. I

have set out below a number of steps that you can use to navigate your way through

the decision-making process.

Choosing the right statistic 107

Step 1: What questions do you want to address?

Write yourself a full list of all the questions you would like to answer from your

research. You might ﬁ nd that some questions could be asked a number of different

ways. For each of your areas of interest, see if you can present your question in a

number of different ways. You will use these alternatives when considering the differ-

ent statistical approaches you might use. For example, you might be interested in the

effect of age on optimism. There are a number of ways you could ask the question:

• Is there a relationship between age and level of optimism?

• Are older people more optimistic than younger people?

These two different questions require different statistical techniques. The question of

which is more suitable may depend on the nature of the data you have collected. So,

for each area of interest, detail a number of different questions.

Step 2: Find the questionnaire items and scales that you will

use to address these questions

The type of items and scales that were included in your study will play a large part

in determining which statistical techniques are suitable to address your research

questions. That is why it is so important to consider the analyses that you intend to

use when ﬁ rst designing your study. For example, the way in which you collected

information about respondents’ age (see example in Step 1) will determine which

statistics are available for you to use. If you asked people to tick one of two options

(under 35/over 35), your choice of statistics would be very limited because there

are only two possible values for your variable age. If, on the other hand, you asked

people to give their age in years, your choices are broadened because you can have

scores varying across a wide range of values, from 18 to 80+. In this situation, you

may choose to collapse the range of ages down into a smaller number of categories

for some analyses (ANOVA), but the full range of scores is also available for other

analyses (e.g. correlation).

If you administered a questionnaire or survey for your study, go back to the

speciﬁ c questionnaire items and your codebook and ﬁ nd each of the individual ques-

tions (e.g. age) and total scale scores (e.g. optimism) that you will use in your analyses.

Identify each variable, how it was measured, how many response options there were

and the possible range of scores.

If your study involved an experiment, check how each of your dependent and

independent variables was measured. Did the scores on the variable consist of the

number of correct responses, an observer’s rating of a speciﬁ c behaviour, or the length

of time a subject spent on a speciﬁ c activity? Whatever the nature of the study, just be

clear that you know how each of your variables was measured.

108 Preliminary analyses

Step 3: Identify the nature of each of your variables

The next step is to identify the nature of each of your variables. In particular, you need

to determine whether each of your variables is an independent variable or a dependent

variable. This information comes not from your data but from your understanding

of the topic area, relevant theories and previous research. It is essential that you are

clear in your own mind (and in your research questions) concerning the relationship

between your variables—which ones are doing the inﬂ uencing (independent) and

which ones are being affected (dependent). There are some analyses (e.g. correlation)

where it is not necessary to specify which variables are independent and dependent.

For other analyses, such as ANOVA, it is important that you have this clear. Drawing

a model of how you see your variables relating is often useful here (see Step 4,

discussed next).

It is also important that you know the level of measurement for each of your vari-

ables. Different statistics are required for variables that are categorical and continuous,

so it is important to know what you are working with. Are your variables:

• categorical (also referred to as nominal level data, e.g. sex: male/females)?

• ordinal (rankings: 1st, 2nd, 3rd)?

• continuous (also referred to as interval level data, e.g. age in years, or scores on the

Optimism Scale)?

There are some occasions when you might want to change the level of measurement

for particular variables. You can ‘collapse’ continuous variable responses down into a

smaller number of categories (see Chapter 8). For example, age can be broken down

into different categories (e.g. under 35/over 35). This can be useful if you want to

conduct an ANOVA. It can also be used if your continuous variables do not meet

some of the assumptions for particular analyses (e.g. very skewed distributions).

Summarising the data does have some disadvantages, however, as you lose infor-

mation. By ‘lumping’ people together, you can sometimes miss important differences.

So you need to weigh up the beneﬁ ts and disadvantages carefully.

Additional information required for continuous and categorical

variables

For continuous variables, you should collect information on the distribution of scores

(e.g. are they normally distributed or are they badly skewed?). What is the range of

scores? (See Chapter 6 for the procedures to do this.) If your variable involves cat-

egories (e.g. group 1/group 2, males/females), ﬁ nd out how many people fall into each

category (are the groups equal or very unbalanced?). Are some of the possible cat-

egories empty? (See Chapter 6.) All of this information that you gather about your

variables here will be used later to narrow down the choice of statistics to use.

Choosing the right statistic 109

Step 4: Draw a diagram for each of your research questions

I often ﬁ nd that students are at a loss for words when trying to explain what they

are researching. Sometimes it is easier, and clearer, to summarise the key points in a

diagram. The idea is to pull together some of the information you have collected in

Steps 1 and 2 above in a simple format that will help you choose the correct statistical

technique to use, or to choose among a number of different options.

One of the key issues you should be considering is: am I interested in the relation-

ship between two variables, or am I interested in comparing two groups of participants?

Summarising the information that you have, and drawing a diagram for each question,

may help clarify this for you. I will demonstrate by setting out the information and

drawing diagrams for a number of different research questions.

Question 1: Is there a relationship between age and level of optimism?

Variables:

• Age—continuous: age in years from 18 to 80.

• Optimism—continuous: scores on the Optimism Scale, ranging from 6 to 30.

From your literature review you hypothesise that older people are more optimistic

than younger people. This relationship between two continuous variables could be

illustrated as follows:

**

*

****

*

*****

*

***

*

Optimism

Age

If you expected optimism scores to increase with age, you would place the points

starting low on the left and moving up towards the right. If you predicted that

optimism would decrease with age, then your points would start high on the left-hand

side and would fall as you moved towards the right.

Question 2: Are males more optimistic than females?

Variables:

• Sex—independent, categorical (two groups): males/females.

• Optimism—dependent, continuous: scores on the Optimism Scale, ranging from

6 to 30.

110 Preliminary analyses

The results from this question, with one categorical variable (with only two groups)

and one continuous variable, could be summarised as follows:

Males Females

Mean optimism score

Question 3: Is the effect of age on optimism different for males and

females?

If you wished to investigate the joint effects of age and gender on optimism scores,

you might decide to break your sample into three age groups (under 30, 31–49 years

and 50+).

Variables:

• Sex—independent, categorical: males/females.

• Age—independent, categorical: participants divided into three equal groups.

• Optimism—dependent, continuous: scores on the Optimism Scale, ranging from

6 to 30.

The diagram might look like this:

Age

Under 30 31–49 50 years and over

Mean optimism

score

Males

Females

Question 4: How much of the variance in life satisfaction can be

explained by a set of personality factors (self-esteem, optimism,

perceived control)?

Perhaps you are interested in comparing the predictive ability of a number of differ-

ent independent variables on a dependent measure. You are also interested in how

much variance in your dependent variable is explained by the set of independent

variables.

Variables:

• Self-esteem—independent, continuous.

• Optimism—independent, continuous.

• Perceived control—independent, continuous.

• Life satisfaction—dependent, continuous.

Choosing the right statistic 111

Your diagram might look like this:

Self-esteem

Optimism

Perceived control

Life satisfaction

Step 5: Decide whether a parametric or a non-parametric

statistical technique is appropriate

Just to confuse research students even further, the wide variety of statistical techniques

that are available are classiﬁ ed into two main groups: parametric and non-parametric.

Parametric statistics are more powerful, but they do have more ‘strings attached’; that

is, they make assumptions about the data that are more stringent. For example, they

assume that the underlying distribution of scores in the population from which you

have drawn your sample is normal.

Each of the different parametric techniques (such as t-tests, ANOVA, Pearson

correlation) has other additional assumptions. It is important that you check these

before you conduct your analyses. The speciﬁ c assumptions are listed for each of the

techniques covered in the remaining chapters of this book.

What if you don’t meet the assumptions for the statistical technique that you want to

use? Unfortunately, in social science research this is a common situation. Many of the

attributes we want to measure are in fact not normally distributed. Some are strongly

skewed, with most scores falling at the low end (e.g. depression); others are skewed so

that most of the scores fall at the high end of the scale (e.g. self-esteem).

If you don’t meet the assumptions of the statistic you wish to use you have a

number of choices, and these are detailed below.

Option 1

You can use the parametric technique anyway and hope that it does not seriously

invalidate your ﬁ ndings. Some statistics writers argue that most of the approaches are

fairly ‘robust’; that is, they will tolerate minor violations of assumptions, particularly

if you have a good size sample. If you decide to go ahead with the analysis anyway you

will need to justify this in your write-up, so collect together useful quotes from statis-

tics writers, previous researchers etc. to support your decision. Check journal articles

on your topic area, particularly those that have used the same scales. Do they mention

similar problems? If so, what have these other authors done? For a simple, easy-

to-follow review of the robustness of different tests, see Cone and Foster (2006).

Option 2

You may be able to manipulate your data so that the assumptions of the statistical

test (e.g. normal distribution) are met. Some authors suggest ‘transforming’ your

112 Preliminary analyses

variables if their distribution is not normal (see Chapter 8). There is some controversy

concerning this approach, so make sure you read up on this so that you can justify

what you have done (see Tabachnick & Fidell 2007).

Option 3

The other alternative when you don’t meet parametric assumptions is to use a

non-parametric technique instead. For many of the commonly used parametric

techniques, there is a corresponding non-parametric alternative. These still come

with some assumptions, but less stringent ones. These non-parametric alternatives

(e.g. Kruskal-Wallis, Mann-Whitney U, Chi-square) tend to be not as powerful;

that is, they may be less sensitive in detecting a relationship or a difference among

groups. Some of the more commonly used non-parametric techniques are covered

in Chapter 16.

Step 6: Making the ﬁ nal decision

Once you have collected the necessary information concerning your research ques-

tions, the level of measurement for each of your variables and the characteristics

of the data you have available, you are ﬁ nally in a position to consider your options.

In the text below, I have summarised the key elements of some of the major statisti-

cal approaches you are likely to encounter. Scan down the list, ﬁ nd an example of the

type of research question you want to address and check that you have all the neces-

sary ingredients. Also consider whether there might be other ways you could ask your

question and use a different statistical approach. I have included a summary table at

the end of this chapter to help with the decision-making process.

Seek out additional information on the techniques you choose to use to ensure

that you have a good understanding of their underlying principles and their assump-

tions. It is a good idea to use a number of different sources for this process: different

authors have different opinions. You should have an understanding of the contro-

versial issues—you may even need to justify the use of a particular statistic in your

situation—so make sure you have read widely.

KEY FEATURES OF THE MAJOR STATISTICAL

TECHNIQUES

This section is divided into two sections:

1. techniques used to explore relationships among variables (covered in Part Four of

this book)

2. techniques used to explore differences among groups (covered in Part Five of this

book).

Choosing the right statistic 113

Exploring relationships among variables

Chi-square for independence

Example of research question: What is the relationship between gender and dropout

rates from therapy?

What you need:

• one categorical independent variable (e.g. sex: males/females)

• one categorical dependent variable (e.g. dropout: Yes/No).

You are interested in the number of people in each category (not scores on a scale).

Diagram:

Males Females

Dropout Yes

No

Correlation

Example of research question: Is there a relationship between age and optimism

scores? Does optimism increase with age?

What you need: two continuous variables (e.g. age, optimism scores).

Diagram:

**

*

*****

****

****

*

Optimism

Age

Non-parametric alternative: Spearman’s Rank Order Correlation.

Partial correlation

Example of research question: After controlling for the effects of socially desirable

responding, is there still a signiﬁ cant relationship between optimism and life satisfac-

tion scores?

What you need: Three continuous variables (e.g. optimism, life satisfaction, socially

desirable responding).

Non-parametric alternative: None.

114 Preliminary analyses

Multiple regression

Example of research question: How much of the variance in life satisfaction scores

can be explained by the following set of variables: self-esteem, optimism and perceived

control? Which of these variables is a better predictor of life satisfaction?

What you need:

• one continuous dependent variable (e.g. life satisfaction)

• two or more continuous independent variables (e.g. self-esteem, optimism,

perceived control).

Diagram:

Self-esteem

Optimism

Perceived control

Life satisfaction

Non-parametric alternative: None.

Exploring differences between groups

Independent-samples t-test

Example of research question: Are males more optimistic than females?

What you need:

• one categorical independent variable with only two groups (e.g. sex: males/

females)

• one continuous dependent variable (e.g. optimism score).

Participants can belong to only one group.

Diagram:

Males Females

Mean optimism score

Non-parametric alternative: Mann-Whitney U Test.

Paired-samples t-test (repeated measures)

Example of research question: Does ten weeks of meditation training result in a

decrease in participants’ level of anxiety? Is there a change in anxiety levels from Time 1

(pre-intervention) to Time 2 (post-intervention)?

What you need:

• one categorical independent variable (e.g. Time 1/Time 2)

• one continuous dependent variable (e.g. anxiety score).

Choosing the right statistic 115

Same participants tested on two separate occasions: Time 1 (before intervention) and

Time 2 (after intervention).

Diagram:

Time 1 Time 2

Mean anxiety score

Non-parametric alternative: Wilcoxon Signed Rank Test.

One-way between-groups analysis of variance

Example of research question: Is there a difference in optimism scores for people

who are under 30, between 31–49 and 50 years and over?

What you need:

• one categorical independent variable with two or more groups (e.g. age: under

30/31–49/50+)

• one continuous dependent variable (e.g. optimism score).

Diagram:

Age

Under 30 31–49 50 years and over

Mean optimism score

Non-parametric alternative: Kruskal-Wallis Test.

Two-way between-groups analysis of variance

Example of research question: What is the effect of age on optimism scores for males

and females?

What do you need:

• two categorical independent variables (e.g. sex: males/females; age group: under

30/31–49/50+)

• one continuous dependent variable (e.g. optimism score).

Diagram:

Age

Under 30 31–49 50 years and over

Mean optimism

score

Males

Females

Non-parametric alternative: None.

116 Preliminary analyses

Note: analysis of variance can also be extended to include three or more independent

variables (usually referred to as Factorial Analysis of Variance).

Mixed between-within analysis of variance

Example of research question: Which intervention (maths skills/conﬁ dence building)

is more effective in reducing participants’ fear of statistics, measured across three

periods (pre-intervention, post-intervention, three-month follow-up)?

What you need:

• one between-groups independent variable (e.g. type of intervention)

• one within-groups independent variable (e.g. time 1, time 2, time 3)

• one continuous dependent variable (e.g. scores on Fear of Statistics Test).

Diagram:

Time

Time 1 Time 2 Time 3

Mean score on Fear of

Statistics Test

Maths skills intervention

Conﬁ dence-building intervention

Non-parametric alternative: None.

Multivariate analysis of variance

Example of research question: Are males better adjusted than females in terms of

their general physical and psychological health (in terms of anxiety and depression

levels and perceived stress)?

What you need:

• one categorical independent variable (e.g. sex: males/females)

• two or more continuous dependent variables (e.g. anxiety, depression, perceived

stress).

Diagram:

Males Females

Anxiety

Depression

Perceived stress

Non-parametric alternative: None.

Note: multivariate analysis of variance can be used with one-way (one independent

variable), two-way (two independent variables) and higher-order factorial designs.

Covariates can also be included.

Choosing the right statistic 117

Analysis of covariance

Example of research question: Is there a signiﬁ cant difference in the Fear of Statis-

tics Test scores for participants in the maths skills group and the conﬁ dence-building

group, while controlling for their pre-test scores on this test?

What you need:

• one categorical independent variable (e.g. type of intervention)

• one continuous dependent variable (e.g. Fear of Statistics Test scores at Time 2)

• one or more continuous covariates (e.g. Fear of Statistics Test scores at Time 1).

Non-parametric alternative: None.

Note: analysis of covariance can be used with one-way (one independent variable),

two-way (two independent variables) and higher-order factorial designs, and with

multivariate designs (two or more dependent variables).

FURTHER READINGS

The statistical techniques discussed in this chapter are only a small sample of all the

different approaches that you can take to data analysis. It is important that you are

aware of the existence, and potential uses, of a wide variety of techniques in order to

choose the most suitable one for your situation. Read as widely as you can.

For a coverage of the basic techniques (t-test, analysis of variance, correlation) go

back to your basic statistics texts, for example Cooper and Schindler (2003); Gravetter

and Wallnau (2004); Peat (2001); Runyon, Coleman and Pittenger (2000); Norman

and Streiner (2000). If you would like more detailed information, particularly on

multivariate statistics, see Hair, Black, Babin, Anderson and Tatham (2006) or Tabach-

nick and Fidell (2007).

118 Preliminary analyses

Summary table of the characteristics of the main statistical techniques

Purpose

Example of

question

Parametric

statistic

Non-parametric

alternative

Independent

variable

Dependent

variable

Essential

features

Exploring

relationships

What is the relationship

between gender and

dropout rates from

therapy

None Chi-square

Chapter 16

One categorical

variable

Sex: M/F

One categorical

variable

Dropout/complete

therapy: Yes/No

The number of cases

in each category is

considered, not scores

Is there a relationship

between age and

optimism scores?

Pearson product-

moment correlation

coefﬁ cient (r)

Chapter 11

Spearman’s

Rank Order

Correlation (rho)

Chapter 11

Two continuous

variables

Age, Optimism scores

One sample with

scores on two

different measures,

or same measure at

Time 1 and Time 2

After controlling for

the effects of socially

desirable responding

bias, is there still a

relationship between

optimism and life

satisfaction?

Partial correlation

Chapter 12

None Two continuous

variables and one

continuous variable

for which you wish to

control Optimism, life

satisfaction, scores on

a social desirability

scale

One sample with

scores on two

different measures,

or same measure at

Time 1 and Time 2

How much of the

variance in life

satisfaction scores can be

explained by self-esteem,

perceived control and

optimism?

Which of these variables

is the best predictor?

Multiple regression

Chapter 13

None Set of two or

more continuous

independent

variables

Self-esteem,

perceived control,

optimism

One continuous

dependent variable

Life satisfaction

One sample with

scores on all measures

What is the underlying

structure of the items

that make up the

Positive and Negative

Affect Scale? How many

factors are involved?

Factor analysis

Chapter 15

None Set of related

continuous variables

Items of the Positive

and Negative Affect

Scale

One sample, multiple

measures

Comparing

groups

Are males more likely to

drop out of therapy than

females?

None Chi-square

Chapter 16

One categorical

independent variable

Sex

One categorical

dependent variable

Dropout/complete

therapy

You are interested in

the number of people

in each catgegory,

not scores on a scale

Is there a change in

participants’ anxiety

scores from Time 1 to

Time 2?

Paired samples t-test

Chapter 17

Wilcoxon Signed

Rank Test

Chapter 16

One categorical

independent variable

(two levels)

Time 1/Time 2

One continuous

dependent variable

Anxiety score

Same people on two

different occasions

Choosing the right statistic 119

Purpose

Example of

question

Parametric

statistic

Non-parametric

alternative

Independent

variable

Dependent

variable

Essential

features

Is there a difference

in optimism scores for

people who are under 35

yrs, 36–49 yrs and

50+ yrs?

One-way between

groups ANOVA

Chapter 18

Kruskal-Wallis

Test

Chapter 16

One categorical

independent variable

(three or more levels)

Age group

One continuous

dependent variable

Anxiety score

Three or more

groups: different

people in each group

Is there a change in

participants’ anxiety

scores from Time 1,

Time 2 and Time 3?

Two-way repeated

measures ANOVA

Chapter 18

Friedman Test

Chapter 16

One categorical

independent variable

(three or more levels)

Time 1/ Time 2/Time 3

One continuous

dependent variable

Anxiety score

Three or more

groups: same people

on two different

occasions

Is there a difference in

the optimism scores for

males and females, who

are under 35 yrs,

36–49 yrs and 50+ yrs?

Two-way between

groups ANOVA

Chapter 19

None Two categorical

independent

variables (two or

more levels)

Age group, Sex

One continuous

dependent variable

Optimism score

Two or more groups

for each independent

variable: different

people in each group

Which intervention

(maths skills/conﬁ dence

building) is more

effective in reducing

participants’ fear of

statistics, measured

across three time

periods?

Mixed between-

within ANOVA

Chapter 20

None One between-groups

independent variable

(two or more levels),

one within-groups

independent variable

(two or more levels)

Type of intervention,

Time

One continuous

dependent variable

Fear of Statistics Test

scores

Two or more groups

with different people

in each group, each

measured on two or

more occasions

Is there a difference

between males and

females, across three

different age groups, in

terms of their scores on

a variety of adjustment

measures (anxiety,

depression and perceived

stress)?

Multivariate

ANOVA (MANOVA)

Chapter 21

None One or more

categorical

independent

variables (two or

more levels)

Age group, Sex

Two or more

related continuous

dependent variables

Anxiety, depression

and perceived stress

scores

Is there a signiﬁ cant

difference in the Fear

of Stats Test scores

for participants in the

maths skills group and

the conﬁ dence building

group, while controlling

for their scores on this

test at Time 1?

Analysis of

covariance

(ANCOVA)

Chapter 22

None One or more

categorical

independent

variables (two or

more levels), one

continuous covariate

variable Type of

intervention, Fear of

Stats Test scores at

Time 1

One continuous

dependent variable

Fear of Stats Test

scores at Time 2

This page intentionally left blank

PART FOUR

Statistical

techniques

to explore

relationships

among variables

In the chapters included in this section, we will be looking at some of the techniques

available in SPSS for exploring relationships among variables. In this section, our

focus is on detecting and describing relationships among variables. All of the tech-

niques covered here are based on correlation. Correlational techniques are often used

by researchers engaged in non-experimental research designs. Unlike experimen-

tal designs, variables are not deliberately manipulated or controlled—variables are

described as they exist naturally. These techniques can be used to:

• explore the association between pairs of variables (correlation)

• predict scores on one variable from scores on another variable (bivariate

regression)

121

122 Statistical techniques to explore relationships among variables

• predict scores on a dependent variable from scores of a number of independent

variables (multiple regression)

• identify the structure underlying a group of related variables (factor analysis).

This family of techniques is used to test models and theories, predict outcomes and

assess reliability and validity of scales.

TECHNIQUES COVERED IN PART FOUR

There is a range of techniques available in SPSS to explore relationships. These vary

according to the type of research question that needs to be addressed and the type of

data available. In this book, however, only the most commonly used techniques are

covered.

Correlation (Chapter 11) is used when you wish to describe the strength and

direction of the relationship between two variables (usually continuous). It can also

be used when one of the variables is dichotomous—that is, it has only two values (e.g.

sex: males/females). The statistic obtained is Pearson’s product-moment correlation

(r). The statistical signiﬁ cance of r is also provided.

Partial correlation (Chapter 12) is used when you wish to explore the relationship

between two variables while statistically controlling for a third variable. This is useful

when you suspect that the relationship between your two variables of interest may

be inﬂ uenced, or confounded, by the impact of a third variable. Partial correlation

statistically removes the inﬂ uence of the third variable, giving a cleaner picture of the

actual relationship between your two variables.

Multiple regression (Chapter 13) allows prediction of a single dependent continu-

ous variable from a group of independent variables. It can be used to test the predictive

power of a set of variables and to assess the relative contribution of each individual

variable.

Logistic regression (Chapter 14) is used instead of multiple regression when your

dependent variable is categorical. It can be used to test the predictive power of a set of

variables and to assess the relative contribution of each individual variable.

Factor analysis (Chapter 15) is used when you have a large number of related

variables (e.g. the items that make up a scale) and you wish to explore the underly-

ing structure of this set of variables. It is useful in reducing a large number of related

variables to a smaller, more manageable, number of dimensions or components. In

the remainder of this introduction to Part Four I will review some of the basic prin-

ciples of correlation that are common to all the techniques covered in Part Four. This

material should be reviewed before you attempt to use any of the procedures covered

in this section.

Statistical techniques to explore relationships among variables 123

REVISION OF THE BASICS

Correlation coefﬁ cients (e.g. Pearson product-moment correlation) provide a nu-

merical summary of the direction and the strength of the linear relationship between

two variables. Pearson correlation coefﬁ cients (r) can range from –1 to +1. The sign

in front indicates whether there is a positive correlation (as one variable increases,

so too does the other) or a negative correlation (as one variable increases, the other

decreases). The size of the absolute value (ignoring the sign) provides information

on the strength of the relationship. A perfect correlation of 1 or –1 indicates that the

value of one variable can be determined exactly by knowing the value on the other

variable. On the other hand, a correlation of 0 indicates no relationship between the

two variables. Knowing the value of one of the variables provides no assistance in

predicting the value of the second variable.

The relationship between variables can be inspected visually by generating

a scatterplot. This is a plot of each pair of scores obtained from the participants

in the sample. Scores on the ﬁ rst variable are plotted along the X (horizontal) axis and

the corresponding scores on the second variable are plotted on the Y (vertical) axis.

An inspection of the scatterplot provides information on both the direction of the

relationship (positive or negative) and the strength of the relationship (this is demon-

strated in more detail in Chapter 11). A scatterplot of a perfect correlation (r=1 or –1)

would show a straight line. A scatterplot when r=0, however, would show a circle or

blob of points, with no pattern evident.

Factors to consider when interpreting a correlation

coefﬁ cient

There are a number of things you need to be careful of when interpreting the results

of a correlation analysis, or other techniques based on correlation. Some of the key

issues are outlined below, but I would suggest you go back to your statistics books and

review this material (see, for example, Gravetter & Wallnau 2004, pp. 520–76).

Non-linear relationship

The correlation coefﬁ cient (e.g. Pearson r) provides an indication of the linear

(straight-line) relationship between variables. In situations where the two variables

are related in non-linear fashion (e.g. curvilinear), Pearson r will seriously underesti-

mate the strength of the relationship. Always check the scatterplot, particularly if you

obtain low values of r.

Outliers

Outliers (values that are substantially lower or higher than the other values in the data set)

can have a dramatic effect on the correlation coefﬁ cient, particularly in small samples.

124 Statistical techniques to explore relationships among variables

In some circumstances outliers can make the r value much higher than it should be, and

in other circumstances they can result in an underestimate of the true relationship. A

scatterplot can be used to check for outliers—just look for values that are sitting out on

their own. These could be due to a data entry error (typing 11, instead of 1), a careless

answer from a respondent, or it could be a true value from a rather strange individual!

If you ﬁ nd an outlier, you should check for errors and correct if appropriate. You may

also need to consider removing or recoding the offending value to reduce the effect it is

having on the r value (see Chapter 6 for a discussion on outliers).

Restricted range of scores

You should always be careful interpreting correlation coefﬁ cients when they come

from only a small subsection of the possible range of scores (e.g. using university

students to study IQ). Correlation coefﬁ cients from studies using a restricted range of

cases are often different from studies where the full range of possible scores is sampled.

In order to provide an accurate and reliable indicator of the strength of the relation-

ship between two variables, there should be as wide a range of scores on each of the

two variables as possible. If you are involved in studying extreme groups (e.g. clients

with high levels of anxiety), you should not try to generalise any correlation beyond

the range of the variable used in the sample.

Correlation versus causality

Correlation provides an indication that there is a relationship between two vari-

ables; it does not, however, indicate that one variable causes the other. The correlation

between two variables (A and B) could be due to the fact that A causes B, that B causes

A, or (just to complicate matters) that an additional variable (C) causes both A and

B. The possibility of a third variable that inﬂ uences both of your observed variables

should always be considered. To illustrate this point, there is the famous story of the

strong correlation that one researcher found between ice-cream consumption and the

number of homicides reported in New York City. Does eating ice-cream cause people

to become violent? No. Both variables (ice-cream consumption and crime rate) were

inﬂ uenced by the weather. During the very hot spells, both the ice-cream consump-

tion and the crime rate increased. Despite the positive correlation obtained, this did

not prove that eating ice-cream causes homicidal behaviour. Just as well—the ice-

cream manufacturers would very quickly be out of business!

The warning here is clear—watch out for the possible inﬂ uence of a third,

confounding variable when designing your own study. If you suspect the possibility

of other variables that might inﬂ uence your result, see if you can measure these at the

same time. By using partial correlation (described in Chapter 12) you can statistically

control for these additional variables, and therefore gain a clearer, and less contami-

nated, indication of the relationship between your two variables of interest.

Statistical techniques to explore relationships among variables 125

Statistical versus practical signiﬁ cance

Don’t get too excited if your correlation coefﬁ cients are ‘signiﬁ cant’. With large samples,

even quite small correlation coefﬁ cients (e.g. r=.2) can reach statistical signiﬁ cance.

Although statistically signiﬁ cant, the practical signiﬁ cance of a correlation of .2 is very

limited. You should focus on the actual size of Pearson’s r and the amount of shared

variance between the two variables. The amount of shared variance can be calculated by

squaring the value of the correlation coefﬁ cient (e.g. .2 X .2 =.04 = 4% shared variance).

To interpret the strength of your correlation coefﬁ cient, you should also take into

account other research that has been conducted in your particular topic area. If other

researchers in your area have been able to predict only 9 per cent of the variance

(r=.3) in a particular outcome (e.g. anxiety), then your study that explains 25 per cent

(r=.5) would be impressive in comparison. In other topic areas, 25 per cent of the

variance explained may seem small and irrelevant.

Assumptions

There are a number of assumptions common to all the techniques covered in

Part Four. These are discussed below. You will need to refer back to these assumptions

when performing any of the analyses covered in Chapters 11, 12, 13, 14 and 15.

Level of measurement

The scale of measurement for the variables for most of the techniques covered in

Part Four should be interval or ratio (continuous). One exception to this is if you

have one dichotomous independent variable (with only two values e.g. sex) and one

continuous dependent variable. You should, however, have roughly the same number

of people or cases in each category of the dichotomous variable.

Spearman’s rho, which is a correlation coefﬁ cient suitable for ordinal or ranked

data, is included in Chapter 11, along with the parametric alternative Pearson corre-

lation coefﬁ cient. Rho is commonly used in the health and medical literature, and is

also increasingly being used in psychology research as researchers become more aware

of the potential problems of assuming that ordinal level ratings (e.g. Likert scales)

approximate interval level scaling.

Related pairs

Each subject must provide a score on both variable X and variable Y (related pairs).

Both pieces of information must be from the same subject.

Independence of observations

The observations that make up your data must be independent of one another. That

is, each observation or measurement must not be inﬂ uenced by any other observation

or measurement. Violation of this assumption, according to Stevens (1996, p. 238), is

126 Statistical techniques to explore relationships among variables

very serious. There are a number of research situations that may violate this assump-

tion of independence. Examples of some such studies are described below (these are

drawn from Stevens 1996, p. 239; and Gravetter & Wallnau 2004, p. 251):

• Studying the performance of students working in pairs or small groups. The

behaviour of each member of the group inﬂ uences all other group members,

thereby violating the assumption of independence.

• Studying the TV-watching habits and preferences of children drawn from the same

family. The behaviour of one child in the family (e.g. watching Program A) is likely

to affect all children in that family; therefore the observations are not independent.

• Studying teaching methods within a classroom and examining the impact

on students’ behaviour and performance. In this situation, all students could

be inﬂ uenced by the presence of a small number of trouble-makers; therefore

individual behavioural or performance measurements are not independent.

Any situation where the observations or measurements are collected in a group

setting, or participants are involved in some form of interaction with one another,

should be considered suspect. In designing your study, you should try to ensure that all

observations are independent. If you suspect some violation of this assumption, Stevens

(1996, p. 241) recommends that you set a more stringent alpha value (e.g. p<.01).

There are more complex statistical techniques that can be used for data that

involve non-independent samples (e.g. children within different classrooms, within

different schools). This approach involves multilevel modelling, which is beyond the

scope of this book.

Normality

Scores on each variable should be normally distributed. This can be checked by

inspecting the histograms of scores on each variable (see Chapter 6 for instructions).

Linearity

The relationship between the two variables should be linear. This means that when you

look at a scatterplot of scores you should see a straight line (roughly), not a curve.

Homoscedasticity

The variability in scores for variable X should be similar at all values of variable Y.

Check the scatterplot (see Chapter 6 for instructions). It should show a fairly even

cigar shape along its length.

Missing data

When you are doing research, particularly with human beings, it is very rare that you

will obtain complete data from every case. It is thus important that you inspect your

Statistical techniques to explore relationships among variables 127

data ﬁ le for missing data. Run Descriptives and ﬁ nd out what percentage of values is

missing for each of your variables. If you ﬁ nd a variable with a lot of unexpected missing

data, you need to ask yourself why. You should also consider whether your missing

values are happening randomly, or whether there is some systematic pattern (e.g. lots

of women failing to answer the question about their age). SPSS has a Missing Value

Analysis procedure that may help ﬁ nd patterns in your missing values.

You also need to consider how you will deal with missing values when you come

to do your statistical analyses. The Options button in many of the SPSS statistical

procedures offers you choices for how you want SPSS to deal with missing data. It is

important that you choose carefully, as it can have dramatic effects on your results.

This is particularly important if you are including a list of variables and repeating the

same analysis for all variables (e.g. correlations among a group of variables, t-tests for

a series of dependent variables).

• The Exclude cases listwise option will include cases in the analysis only if it has

full data on all of the variables listed in your variables box for that case. A case will

be totally excluded from all the analyses if it is missing even one piece of infor-

mation. This can severely, and unnecessarily, limit your sample size.

• The Exclude cases pairwise option, however, excludes the cases (persons) only

if they are missing the data required for the speciﬁ c analysis. They will still be

included in any of the analyses for which they have the necessary information.

• The Replace with mean option, which is available in some SPSS statistical proce-

dures (e.g. multiple regression), calculates the mean value for the variable and

gives every missing case this value. This option should never be used as it can

severely distort the results of your analysis, particularly if you have a lot of missing

values.

Always press the Options button for any statistical procedure you conduct and

check which of these options is ticked (the default option varies across procedures).

I would strongly recommend that you use pairwise exclusion of missing data, unless

you have a pressing reason to do otherwise. The only situation where you might

need to use listwise exclusion is when you want to refer only to a subset of cases that

provided a full set of results.

Strange-looking numbers

In your output, you may come across some strange-looking numbers that take the

form 1.24E-02. These small values are presented in scientiﬁ c notation. To prevent

this happening, choose Edit from the main menu bar, select Options, and make sure

there is a tick in the box No scientiﬁ c notation for small numbers in tables on the

General tab.

128

11

Correlation

Correlation analysis is used to describe the strength and direction of the linear re-

lationship between two variables. There are a number of different statistics available

from SPSS, depending on the level of measurement and the nature of your data. In

this chapter, the procedure for obtaining and interpreting a Pearson product-moment

correlation coefﬁ cient (r) is presented along with Spearman Rank Order Correlation

(rho). Pearson r is designed for interval level (continuous) variables. It can also be

used if you have one continuous variable (e.g. scores on a measure of self-esteem)

and one dichotomous variable (e.g. sex: M/F). Spearman rho is designed for use with

ordinal level or ranked data and is particularly useful when your data does not meet

the criteria for Pearson correlation.

SPSS will calculate two types of correlation for you. First, it will give you a simple

bivariate correlation (which just means between two variables), also known as zero-

order correlation. SPSS will also allow you to explore the relationship between two

variables while controlling for another variable. This is known as partial correlation.

In this chapter, the procedure to obtain a bivariate Pearson r and non-parametric

Spearman rho is presented. Partial correlation is covered in Chapter 12.

Pearson correlation coefﬁ cients (r) can only take on values from –1 to +1. The

sign out the front indicates whether there is a positive correlation (as one variable

increases, so too does the other) or a negative correlation (as one variable increases,

the other decreases). The size of the absolute value (ignoring the sign) provides an

indication of the strength of the relationship. A perfect correlation of 1 or –1 indi-

cates that the value of one variable can be determined exactly by knowing the value

on the other variable. A scatterplot of this relationship would show a straight line. On

the other hand, a correlation of 0 indicates no relationship between the two variables.

Knowing the value on one of the variables provides no assistance in predicting the

value on the second variable. A scatterplot would show a circle of points, with no

pattern evident.

Correlation 129

There are a number of issues associated with the use of correlation. These include

the effect of non-linear relationships, outliers, restriction of range, correlation versus

causality and statistical versus practical signiﬁ cance. These topics are discussed in the

introduction to Part Four of this book. I would strongly recommend that you read

through