Negative Binomial Regression

Negative Binomial Regression is aimed at ‘practicing researchers and statisticians’ and describes count response models, in particular the negative binomial distribution. The author discusses the problem of overdispersion in great detail and ways of accounting for it in statistical models. The book gives an overview of count response models, starting with basic Poisson regression. Negative binomial regression is then studied followed by its many extensions, e.g. zero-inflation, censoring and truncation, sample selection and multilevel models. The book focuses on procedures that are currently available in commercial statistical software packages and pays particular attention to model selection and correct estimation and interpretation of the models’ parameters. Derivations of models, estimation methods and algorithms, and techniques for goodness of fit are also considered. There are several useful appendices including a list of log-likelihood functions for the negative binomial distribution and its variations. Very good examples are given of the applications of models that are presented, including Stata code and output. An on-line collection of materials (data sets and modelling code) is also provided on the publisher’s Web site, all of which are helpful for those wanting to put techniques into practice. Summaries at the end of each chapter enable the reader to consolidate the topics that are covered. Figures are clearly produced but are often difficult to find owing to a high volume of output in the main body of text. This is a good starting point for those who are not familiar with this topic although not all aspects of the negative binomial regression models are considered. For instance little is said about residuals and goodness-of-fit measures. I would recommend this book to researchers and students who would like to gain an overview of the negative binomial distribution and its extensions.