Experimental Design for the Life Sciences

by Ruxton, Graeme D.; Colegrave, Nick

• ISBN13: 9780199252329
• ISBN10: 0199252327
• Format: Paperback
• Copyright: 2003-04-10
• Publisher: Oxford University Press

Summary

This book aims to teach the reader how to design effective experiments. The overwhelming majority of life scientists design experiments. However they tend to approach design in an informal ad hoc way, improving their techniques in the light of experience. This book aims to provide the junior scientist a short-cut way to learn how to design effective experiments without going through a painful trial and error process. For more experienced scientists, the text should also function to stimulate them to think about the way they design experiments, and perhaps lead them to design more effective experiments in future. Concepts, such as power analysis and pseudoreplication, that many experienced scientists consider to be mystifying or difficult, are explained in clear and practical terms. The emphasis throughout is to demonstrate that good experimental design is about clear thinking and biological understanding, not mathematical or statistical complexity. Companion Web Site All the figures from the book will be available to download free from the companion web site at www.oup.com/uk/best.textbooks/biology/ruxtoncolegrave

Author Biography

Graeme D. Ruxton is a Reader in Behavioural and Physiological Ecology at the University of Glasgow Nick Colegrave is a Lecturer in Invertebrate Biology at the University of Edinburgh

Table of Contents

Flow chart on designing an experiment	xiv
Why you need to care about design
Why experiments need to be designed	1	(2)
The costs of poor design	3	(1)
Time and money	3	(1)
Ethical issues	3	(1)
The relationship between experimental design and statistics	4	(1)
Why good experimental design is particularly important to life scientists	5	(2)
Summary	7	(1)
Starting with a well-defined hypothesis
Why your experiment should be focused: questions, hypotheses and predictions	8	(4)
Producing the strongest evidence with which to challenge a hypothesis	12	(2)
Satisfying sceptics	14	(1)
The importance of a pilot study and preliminary data	15	(4)
Making sure that you are asking a sensible question	16	(1)
Making sure that your techniques work	17	(1)
Obtaining basic data to fine-tune design and statistics	18	(1)
Experimental manipulation versus natural variation	19	(6)
Arguments for doing a correlational study	20	(1)
Arguments for doing a manipulative experiment	21	(3)
Situations where manipulation is impossible	24	(1)
Deciding whether to work in the field or the laboratory	25	(2)
There is no perfect study	27	(1)
Summary	28	(2)
Between-individual variation, replication and sampling
Between-individual variation	30	(2)
Replication	32	(3)
Pseudoreplication	35	(10)
Explaining what pseudoreplication is	35	(3)
Common sources of pseudoreplication	38	(3)
Dealing with pseudoreplication	41	(2)
Accepting that sometimes pseudoreplication is unavoidable	43	(1)
Pseudoreplication, third variables and confounding variables	44	(1)
Randomisation	45	(5)
Why you often want a random sample	45	(1)
Haphazard sampling	46	(1)
Self-selection	47	(1)
Some pitfalls associated with randomisation procedures	48	(1)
Randomising the order in which you treat replicates	48	(1)
Random samples and representative samples	49	(1)
Selecting an appropriate number of replicates	50	(6)
Educated guesswork	50	(1)
Formal power analysis	50	(1)
Factors affecting the power of an experiment	51	(5)
Summary	56	(2)
Different experimental designs
Controls	58	(6)
Blind procedures	61	(1)
Making sure the control is as reliable as it can be	62	(1)
The ethics of controlling	63	(1)
Situations where a control is not required	63	(1)
Completely randomised and factorial experiments	64	(6)
Pros and cons of complete randomisation	68	(2)
Blocking	70	(6)
Blocking on individual characters, space and time	72	(1)
The pros and cons of blocking	73	(1)
Paired designs	73	(1)
How to select blocks	74	(1)
Covariates	74	(2)
Cross-over designs	76	(2)
Split-plot designs	78	(2)
Summary	80	(2)
Taking measurements
Calibration	82	(1)
Inaccuracy and imprecision	83	(3)
Intra-observer variability	86	(2)
Inter-observer variability	88	(1)
Defining categories	89	(1)
Observer effects	90	(1)
Recording data	91	(2)
Computers and automated data collection	93	(1)
Summary	94	(1)
Final thoughts
How to select the levels for a treatment	95	(2)
Subsampling: more woods or more trees?	97	(2)
Using unbalanced groups for ethical reasons	99	(3)
Other sampling schemes	102	(2)
Sequential sampling	102	(1)
Stratified sampling	102	(2)
Systematic sampling	104	(1)
Latin square designs	104	(1)
More on interactions	105	(3)
Covariates can interact too	105	(2)
The importance of interactions (and the interaction fallacy)	107	(1)
Summary	108	(2)
Bibliography	110	(3)
Index	113