Categorial data analysis for geographers and environmental scientists

Categorial data analysis for geographers and environmental scientists

360 Book Reviews you as regards your children’s inheritance: to the male a proportion equal to that of two females’ (p. 175). Like the Dutch elm, th...

111KB Sizes 0 Downloads 46 Views

360

Book Reviews

you as regards your children’s inheritance: to the male a proportion equal to that of two females’ (p. 175). Like the Dutch elm, the ‘Dutch disease’ is simply that export booms actually retard growth in other parts of the economy analogous to the observed impact of North Sea gas on the Dutch economy (p. 236). The socialist experience section notes that gossip is the kibbutz police force (p. 347). Socialist calls for land to the tiller and where the landlord will never eat of your poverty were broadly discussed (p. 304). Technically the book has been well put together with consistently numbered end-notes to each chapter followed by a bibliography. However, the references to journal articles are without page numbers. In one chapter extensive use was made of formulae, but happily this was well typeset making reading and cross-referencing easy. The ten diagrams and forty statistical tables were neat and well set out and there is a comprehensive ten-page index. The four ‘typos’ of Perak (p. 172). Research School of Social Science (p. 252) Ohlin (p. 402) and Voon (p. 408) do not however blemish what is otherwise a valuable book on the primary sector. George

Cho

Department of Applied Geography, Canberra College of Advanced Education

Wrigley. N. Categorial data analysis for geographers and environmental scientists. London: Longman, 1985. 392 pp. f45 hardback. Categorical data arise from classification procedures where the number of observations falling into particular categories (socioeconomic class, political affiliation, vegetation type etc.) are recorded. It thus refers to data measured at the nominal or ordinal level. Wrigley notes that whilst much social science data is of the categorical type the best known and frequently taught analytical techniques (notably linear regression) are primarily for data recorded at interval or ratio scales of measurement. The social scientist’s armoury for dealing with categorical data is all too often restricted to analysis of variance or non-parametric techniques such as chi-squared tests, or rather ad hoc alterations to the linear regression model. Wrigley’s text is concerned exclusively with the presentation of a family of statistical models appropriate to dealing with nominal and ordinallevel data. The book assumes a basic knowledge of descriptive statistics, the principles of classical statistical inference, multivariate methods and also matrix manipulation. After an introductory chapter, Part Two of the book deals with regression models and how to extend them to handle data where the response variable (‘dependent’ variable) is categorical. A series of chapters then deal with

situations where the explanatory variables are either continuous or categorical or both. Chaprer 4 introduces the linear logit model and Chapter 6 the log linear model for categorical explanatory variables. Chapter 7 contains brief references to available computer packages. Part Three extends the models of the previous section while Part Four discusses discrete choice modelling, which is perhaps the major area of use of categorical data analytic methods. A great deal of effort has gone into the presentation of a complicated literature and the true value of this book will only emerge for those who seek to use it as a guide book into this area of statistical analysis. It is less a book to sit down and read from cover to cover, more a text to use in detail in order to understand particular areas of this field or to carry out a ‘hands-on’ data analysis. Does it present material in an easy-to-read way. is it easy to access information from? These are the prime considerations for such users. It is good to see a statistical textbook for geographers written by someone who has devoted considerable time and effort to mastering and applying the techniques. It is an impressive book but it is not an easy book; moreover, despite being presented as a book for ‘geographers and environmental scientists’ it could more accurately have been labelled for ‘social and environmental scientists’ in terms of the sorts of examples presented. Wrigley feels that this class of techniques will become more important over the nest ten years in geography. This reviewer concurs, though like a number of geographical statistical texts that have appeared over the last few years its encyclopaedic, atheoretical approach may do less to encourage this adoption than would something rather more limited in scope, directed at a narrow but coherent body of geographical problems. There is one obvious lacuna in this area of technical development that is particularly important for geographers: namely, the recognition of the spatial nature of much geographical (and environmental) data. The problems raised, particularly in the w-variable models, are mentioned from time to time, but as the author admits, there is still a long way to go. The linking together of this area of statistics with developments in the area of spatial analysis are, as he points out in an important research area. the conclusions, Indeed, I would suggest that for the application of these models in the wider geographical and environmental field, such further research developments are a precondition. It is to be hoped that effort will be concentrated here. This is an important book and is certainly essential reading for anyone considering working with this class of techniques. I have no doubt it will be much referenced in years to come. R. P. Haining

Department of Geography, University of Sheffield