2
PDFlib PLOP: PDF Linearization, Optimization, Protection Page inserted by evaluation version www.pdflib.com – [email protected]

Models for Ecological Data: An Introduction

Embed Size (px)

Citation preview

Page 1: Models for Ecological Data: An Introduction

PDFlib PLOP: PDF Linearization, Optimization, Protection

Page inserted by evaluation versionwww.pdflib.com – [email protected]

Page 2: Models for Ecological Data: An Introduction

include estimation of parameters, calculation of confi-dence intervals and analogous measures of precision,an introduction to the Bayesian approach to estima-tion, model assessment and selection, and computa-tional aspects of Bayesian analyses. In some ways,these first seven chapters act as introductory material.Chapter 8 covers hierarchical models, illustrating howthey can be applicable to many ecological datasetsand permit integration of information from multiplesources.

With a focus on population ecology, the ninth andtenth chapters include some of the most detailedmodels in the book, moving beyond the simplest illus-trative examples. The ninth chapter covers analyses oftemporal datasets, introducing a variety of modelsfor population dynamics. Topics include time seriesanalysis, detection of density dependence and the as-sociated estimation of parameters, models for casesin which the population size is observed with error,mark-recapture models, and stage-structuredmodels.

The tenth chapter investigates models that integratespatial and temporal data. The first topic is not sur-prising given Clark’s own work: the integration ofpopulation growth and models of random walks,followed by the importance of considering rarelong-distance dispersal and its ability to accelerate therate of spread of invading organisms. Subsequenttopics include island biogeography and metapopula-tion models, other models of patch occupancy,and estimation of dispersal rates as a function ofdistance.

This book is an excellent introduction to ecologicalmodelling. In the preface of Hilborn and Mangel’s(1997) The Ecological Detective, they state unapologeti-cally, ‘This book has equations in it.’ Clark has taken asimilar approach. Mathematics is the basis of ecologi-cal modelling, so if you are serious about doing eco-logical modelling, you will have to understand themathematical details. After the first chapter, readersshould expect multiple equations per page, and insome cases almost entire pages of equations.There areappendices on a variety of mathematical topics to help:a very nice and succinct one on Taylor series, differen-tial and difference equations, matrix algebra, proba-bility models, life history calculations, probability dis-tributions and conjugate pairs for likelihoods andBayesian priors. This material is used throughout thebook, and it indicates the kind of mathematical topicsthat readers will encounter in the book.

Clark is pragmatic in his choice of statistical para-digm, using the approach that gets the job done mosteasily. He uses Bayesian methods extensively forthe more complicated examples and conventional

methods for the simplest cases because they requireless effort. Perhaps this overstates the burden imposedby using Bayesian analysis, which is minimal in myopinion, and undersells the benefits of Bayesianmethods even for simple cases (e.g. it is easier to cal-culate confidence intervals for regression models byusing Bayesian methods, and the logical consistency ofBayes is appealing). Nevertheless, readers will find acomprehensive introduction to Bayesian data analysis,including the tools for programming their ownanalyses.

On reading the book in isolation, one might get theimpression that the development and analysis of eco-logical models is achieved by doing algebra and solvingequations by hand.This might put many ecologists off.Clark’s view is that ecological modelling requires anability to write one’s own computer code for conduct-ing analyses, and understanding the mathematics isessential. Allowing non-statisticians to use softwaresuch as WinBUGS, which would make many of theexamples in the book more accessible to most ecolo-gists, is viewed as ‘like giving guns to children’. Whilean exaggeration, I see Clark’s point. However, ecolo-gists misuse statistical models routinely, without evertouching software like WinBUGS (Fidler et al. 2006).Clark’s approach means that most ecologists will notgain as much from his book as they might because theywill not have the time or motivation to learn computerprogramming. But it makes an excellent book forserious students of ecological modelling.The book hasan accompanying lab manual (a copy of which I wasnot able to obtain before completing the review) thatdescribes how the various analyses can be undertakenin the free statistical software R. This lab manual mayturn out to be the most important element for intro-ducing readers to ecological modelling.

MICHAEL McCARTHYSchool of Botany

The University of MelbourneParkville,Victoria, Australia

Email: [email protected]

REFERENCES

Fidler F., Burgman M. A., Cumming G., Buttrose R. &Thomason N. (2006) Impact of criticism of null-hypothesissignificance testing on statistical reporting practices in con-servation biology. Conserv. Biol. 20, 1539–44.

Hilborn R. & Mangel M. (1997) The Ecological Detective. Con-fronting Models with Data. Princeton University Press,Princeton.

242 BOOK REVIEWS

© 2008 Ecological Society of Australiadoi:10.1111/j.1442-9993.2007.01838.x