Experimental Design for the Life Sciences

Citation preview

second edition

Preface

HOW

This

to read this book

book i

an introduction to exp rimental design. We mean it to be a

good starter if you have never thought about experimental de ign before, and a good tune-up if you feel the need to take design more eriou Iy. It does not come

c l o se

to being the la t word on experimental design. We

cover few areas of design exhaustively, and

orne areas not at all. We use

the Bibliography to recommend some good books that would facilitate

a

deeper consideration of des ig n issues. That said it is also i m portan t to

realize that the basic ideas of de sig n covered in thi book are enough to carry out a wi d e range of scientific investigations. Many scientists forge very successful career using experiments that never take t hem outside the conlines of the material covered in this book. Tllis book will also help you

tackle more advanced tex ts , but if you a bsorb all that we di scuss here, then you may find that you know all the experimental design that you feel you need to know.

This book is about how to design experiments so as to collect good quality data. Of course, that data will almost certainly need statistical analy is in order to answer the research questions that you are interested

in. In order to keep the size of this book manageable, we do not enter into the details of stati tical analysis. Fortunately, th ere are a huge number of books on

tatistical analysis available: we even like some of them ! We

recommend some of our favourites in the Bibliography. We also provide

some po inters in the text to the types of stati tical test

t h a t different

designs are li k ely to lead to. We often illustrate our points through use of examples. Indeed, in some cases we have found that the onl y way to discuss some issues was through examples. In other cases, our point can be stated in general terms. Here we still use examples to

a m pli

fy and

illustrate. Although we think t h at you

should aim to read the book right throu gh like a novel at least once, we have tried to organize the text to make dipping into the book easy too. Our first job will be to remove any doubt from your nlind that experi­ mental des ign is important, and so we tackle this in Chapter 1. Chapter 2 di sc u sse s how good designs flow naturally from clearly stated scientific que rions. Almost all experiments involve studying a sample and extrapol­ ating conelu ion

about that sample more widely; how to select a good

viii

Preface

sample is the key theme of Chapter 3. The nitty-gritty of orne simple designs forms the basis of Chapte r 4. There i no point i n formally designing an elegant experiment if you then make a poor j b of actually collecting

C1ata from your sample individuals,

0

Chapter 5 contain some tips on

taking effective measurements. Chapter 6 is a compendium of slightly more specialized points, which do

but at least some of which

n ot fit naturally into [he other chapters

g ht to be useful to your branch of the life

ou

sciences. This is not a maths book. Nowhere in this book will you find the equation for the normal distribution, or any other equation for that matter. Experimental design is not a

ubsection of maths-you don't need

to he a mathematician to understand simple but effective designs and you certainly don't need to know maths to understand this bo k. Good experimental design is vital to good ctence. It i generally nothing like as difficult as some would have you believe: you can go with just the few simple guidelines

a

long way

overed in this book. Perhaps most

amazingly of all, it is possible to derive

enjoyment from thinking about

experimental design: why else would we have wanted to write thi book!

On the second edition The coverage has not changed dramatically from the first edition, although all sections have been rewritten-sometimes extensively

o-for greater

clarity. We have also attended to filling in a few gap, most nota bly in sub­ stantially increa i ng our consideration of the spe ial challenge associated with using human subjects. However:, the big change in thj edition is the way in which the material is presented. As well as extending the use of some of the features of the previou edition, we also in c l ude a number of new features that we hope will substantia Uy increase the ea book. We introduce these new features

e

of use of this

on the following two pages.

Learning features

SeJf-te t questions The more you think about experimental

Key definitions There is a lor of

jargon a

ociared with

experimental design and statistical analysis.

We have not tried to avoid thi . Indeed, we have deliberately tried to introduce you to as

mllch of

the jargon as we can. By being

e posed to this jargon, we h o pe that it wi ll

become second nature to you to use-and under tand-it. Thi hould also make read ­ ing more advanced texts and the primary lit­

design, the easier de meors

straightforward, we have increa ed the num­

ber of definitions of key term provided and

becomes. To get you thinking while

you are reading thi book, we now include a number of self-te t questions in every chap­ ter. Often there

w

wrong answer to a que tion, but sugge ted back of the book. hMpit3.1 ward respond

e

mort'

effectively

ocner �d of rho ""'rd. Cand,d,,,. hypo,h

02.1 Su1fiGSl:some bn>Othest!lll 'hilt could

1. PaDcDts .ue nor d.slllbu[�d random1

fJCt plDln

having.o1 rr:ndency

peopl� ami faster on the

m

pLl(:� mor� snCf

m 3way frarn lhr. enfrancc.

Jo� to woMi: l!\in on the­ wa.".�me

2. Panenrs nearer [he door arc: seen tirst 0 gemng more of tbe d()(1or'ts time and

increa ed the level of detail in all definjrjons. Each key word or phra

i l l not be a clear right or

answer to all questions can be found at the

erature less daunting. However, to make negotiating the minefield of jargon more

igning robust experi­

iii

3. Patients �pond po::sirivrly to iDCle�

is emboldened at

contnct :\oio"j;oci8�d with being ncan'r

�

dle point where we first u e it, and is given a clear definition in the margin nearby. ,.-----------;:;;;;::;::: ;;;;;; :;;; ;;;;;;;:;;;: � � � 3.4 Randomization An easy wily m avoid manr SOUrce!. of p� R,l1domfndon 5frnply meil1l$ c1rirMni F"ijIndtim s.amplM fQ( ,rudy from the wider pcpulauc," 01 I rn. pos5lblelneMduaf.5 that cooltJ be in yOUr $'amplE!

mal your �xpenmcnr is properly randomiz

,l:implest techniqut:&

ro usc In experiI:neo

the most misunderstood Dond abl.l..Sed. Pro

end mo r section

with a take-horne mes­

sage. By including the e throughout the text we hope to give you an immediate opportu­

a't'Qid many :5t1un;e!l of

nity to evaluate your under tanding of a sec­

subfect

""/iI..d"al u{ fi"dJng Itself In LAn

To help to con olidate your thinking, we

has

any ind'tl'fdwd trP�fm,mtJ11 randomiutlon

Take-home messages

.a­ u c CII :s I:T

!!!

IL

x 2Y Figure

x +

500