Bootstrap methods are less than 20 years old, having been introduced by Bradley Efron in 1979, but they are now a familiar component of the statistical armoury in use by ecologists. The essence of the approach is simple: an observational or experimental data set is resampled many times and the new data sets so created are used to build confidence intervals for parameters, or tests of hypotheses, that do not depend on knowledge of the underlying distribution from which the data were sampled. This simplicity is both powerful and seductive. In the words of the introduction to this book (p. 4), ‘Bootstrap methods are intended to help avoid tedious calculations based on questionable assumptions, and this they do. But they cannot replace clear critical thought about the problem, appropriate design of the investigation and data analysis, and incisive presentation of conclusions.’
Davidson and Hinkley claim an ‘emphasis on practicalities’ across a broad range of uses for bootstrap methods. To this end, they include many theoretical problems and practical exercises (the book evolved from a lecture course) and a set of routines in S-Plus on an accompanying disk (and also available on their World Wide Web site). However, the authors are statisticians, and they are speaking primarily to statisticians. I count myself as a reasonably numerate biologist, and yet, I confess, I found this volume daunting. It is doubtless a goldmine for the mathematical biologist, but it is not a guide to the use of bootstrap approaches for the average ecologist. Such a thing is certainly needed because, if nothing else, it is clear from this book that the field has come a long way in its short life and is now full of potential, but remains a dangerous one for the naïve user.