No label defined (Q4254891)

The class of Generalized Linear Models, introduced by \textit{J.A. Nelder} and {R.W.M. Wedderburn} [J. R. Stat. Soc., Ser. A 135, 370-384 (1972)], unifies logistic regression and loglinear models for categorical data with the traditional linear regression and ANOVA methods. Adopting the generalized linear model framework, the authors of this book explain the common features and the relationships between methods for the analysis of categorical and continuous data. Their primary audience is the applied social scientist. The book focuses on the explanation of concepts and the interpretation of computer output -- theoretical derivations are largely omitted. Chapter 1 presents a brief introduction to Generalized Linear Models (GLIM) and their basic building blocks. Chapter 2 (Data screening) focuses on initial variable screening techniques that can be used prior to the construction of formal models. Chapter 3 (Ordinary least-squares regression) discusses simple and multiple linear regression models, emphasizing their specification and validation. Chapter 4 (Logistic regression) considers simple and multiple logistic regression, with a careful discussion of the logit link. The logit is the logarithm of the odds ratio, which is the ratio of the probability of an event happening and the probability of the event not happening. Especially useful are the detailed explanations of how OLS and logistic regression can be carried out within the SPSS and the GLIM software. Chapter 5 (Loglinear analysis) discusses the GLIM model building approach for the analysis of categorical frequency data. Data for categorical variables can be described in terms of a count of the number of instances that occurred for each combination of variables. It is assumed that counts follow a Poisson distribution, and a log link is used to relate the systematic components to the explanatory variables. Generalized linear models with a log link are known as loglinear models. Chapter 6 on Factor Analysis discusses techniques for representing several correlated variables in terms of a smaller number of factors. Generalized Linear Modeling is enhanced by using a small number of factors instead of many multicollinear variables. Chapter 7 (Conclusions) reviews extensions and advanced areas worthy of further study. The book gives a good introduction to ordinary regression, logistic regression, and loglinear models for the analysis of categorical frequency data. It is useful for practitioners who are interested in learning how to apply these methods, but are less interested in the theoretical aspects of statistical inference. Readers who want to understand how parameter estimates and their properties are derived must look for other books. A nice feature of the book and one that social scientists will find very useful is the detailed listing of computer codes for the implementation of these models within the SPSS and GLIM software. It is unfortunate, however, that virtually all data examples have been generated. One would have hoped for ''real'' data sets, especially in a book that is written for practitioners.