Pp. xvii + 315. Princeton University Press, Princeton, New Jersey.
£16.95 (paperback), ISBN 0-691-03497-4. £30.00 (cloth), ISBN 0-691-03496-6.
It is the nature of ecological systems that experimentation is frequently impossible, ecological dynamics operate over very long time-scales, and the questions ecologists ask, particularly in applied ecology, often cannot be answered by simple experiments. It is against this background that Hilborn and Mangel present a series of tools for ecological modelling set within the philosophical framework of likelihood and Bayesian statistics. Likelihood allows us to attach support or a degree of belief in a model or several competing models. The notion of the ‘ecological detective’ neatly summarises this. The work of an applied ecologist is akin to that of a detective: we are frequently constrained in terms of the amount of evidence that is available, and ecology generally proceeds by contrasting the compatibility of a number of competing hypotheses with the data rather than a ‘tree’ of successive hypotheses.
The 11 chapters of the book break down into three introductory chapters, four case studies and four ‘confrontations’. The confrontations are techniques for analysing and fitting models to data. Specifically, they deal with sums of squares, likelihood, Bayesian statistics and model-fitting algorithms. These areas generally come under an ‘out of scope’ category for most bio-statistical text books, and so this book therefore nicely complements existing statistics texts.
This book is not explicitly written as an ‘applied’ ecological text. The material of the four case studies make it clear, however, that the techniques and approaches are entirely relevant to applied ecology. Four out of the five extended case studies are clearly applied problems; namely, an analysis of incidental catches in trawlers, the population dynamics of wildebeest, and an analysis of the effects of environmental change on Namibian hake fisheries. The case studies illustrate well how the techniques that are described in the book may be put into direct practice.
The style of the book is very engaging, and whilst there is some maths the extent and level of this should not be off-putting. The third chapter presents all the material on probability theory that is required to work through the rest of the book. My main gripe would be that the body of opinion that objects to the Bayesian inverse probability approach is not addressed very extensively and it would be necessary to look elsewhere for a critical justification of using this. What I particularly like about this book, though, is that whilst at first sight it may appear to be a statistical recipe book, it achieves more than this: rather, taken as a whole, the philosophical basis for ecological data analysis and modelling is addressed. I would urge anyone involved in ecological data analysis to read this book as much for the general open-minded approach to ecology as for the techniques themselves. Hilborn & Mangel’s approach is neatly summarised on page 11: ‘…if the techniques do not exist, then we must invent them.’