BOOK REVIEWS: 2

This book gives a fresh approach to the topic of categorical data analysis. The presentation of the statistical methods exploits the connection to regression modeling with a focus on practical features rather than formal theory. However, I am in doubt whether the regression approach of categorical data analysis is more intuitive than the approach that starts with contingency tables (see, e.g., Agresti, 1996). Surely, this may depend on the readers themselves. Epidemiologists thinking in fourfold tables may find the regression approach harder to understand than econometricians. The author discusses the various aspects of statistical modeling, i.e., model selection, checking the modeling assumptions, and points the reader to applied problems such as outliers and sparse cell counts in contingency tables. Finally, the reader is given an interpretation of analytical results along with real-world examples. And all this is done in a readable way, especially as the references are avoided in the main text body and discussed at the end of each chapter. The text is broken down into three main parts. The first part gives a prerequisite for the following by reviewing Gaussian-based data analysis and model building. The second part examines the modeling of count data. This includes count regression models such as Poisson regression and negative binomial models and log-linear models for contingency tables. Lastly, the third part of the book discusses logistic regression and alternative models for binary data, as well as extensions for multinomial and ordinal response variables. As emphasized by the title “Analyzing Categorical Data,” this is not a reference, but a textbook. The author makes use of many worked-out real data examples. The data and the computer code for analysis of the examples presented throughout the book are available to the reader via the Internet. The target audience is anyone faced with categorical data, ranging from undergraduate to Ph.D. students to professional data analysts in diverse fields including econometrics, sociology, and management, as well as biometrics and epidemiology. There is much to learn from this book. Aside from the ordinary materials such as association diagrams, Mantel–Haenszel estimators, or overdispersion, the reader will also find some less-often presented but interesting and stimulating topics. Among these is the “mosaic plot” for graphically representing association in contingency tables or “Benford's law” for anomalous numbers. However, from a personal point of view the book is missing here and there a few more pieces of information. For example, rules of thumb for the validity of asymptotic inference, e.g., when can we approximate the binomial by the normal distribution, and when do we need Fisher's exact test, and cannot use the Pearson χ2-statistic. Furthermore, the text does not address the analysis of dependent or clustered data. The presentation is restricted to fixed effects models, e.g., random coefficient and mixed models approaches such as GLMM (generalized linear mixed models) and GEE (generalized estimation equations) are not dealt with. For an introductory text, one might think that these topics can be excluded, but I wished that at least a discussion of these important aspects of analyzing categorical data would have been given in the form of an outlook chapter at the end. The interested reader may then consult a book with a broader scope such as Agresti (2002), or more specialized books such as Hosmer and Lemeshow (2000) or Diggle et al. (2002). Despite this, I think this is an excellent book, giving an up-to-date introduction to the wide field of analyzing categorical data.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.