Book Reviews 855 De Muth’s first text-book on statistics in pharma- ceutical sciences. Although it is produced mainly for pharmacy students and individuals working in the pharmaceutical industry, it is also helpful for prac- tising pharmacists who need to know that, to pro- vide a high level of service to their customers, they themselves must be educated consumers of academic research. This book not only educates the reader about different statistical tests but also trains the reader to be a wise user of research, especially where there might be errors in the data analysis process. By keeping the mathematics to a minimum, the author will certainly attract users from other dis- ciplines in the medical field. All that is needed to read and benefit from it are numeracy up to General Certificate of Secondary Education ordinary level, and an open mind. There are 24 chapters each hav- ing a brief introduction, followed by many worked examples. They end with a short reference section, some practice problems and complete solutions to help the reader to assimilate the material. The text covers more than a usual basic statistics text and in- cludes chapters on survival analysis, tests of equiv- alence, evidence-based practice and outlier tests. This handy reference text empowers medical or pharmaceutical personnel with the statistical tools that are necessary to make an impact in both academia and industry. It can be used as the re- quired course text-book or as a reference text. Li- braries are strongly encouraged to purchase copies for research students. Isaac Dialsingh University of Trinidad and Tobago Generalized Linear Models with Random Effects: Unified Analysis via H-likelihood Y. Lee, J. A. Nelder and Y. Pawitan, 2007 Boca Raton, Chapman and Hall–CRC 396 pp., $89.95 ISBN 1-584-88631-5 This book both demands and rewards close atten- tion. The authors set out to describe, deploy and defend the hierarchical likelihood. They provide de- tailed reviews of background material, an extremely generous palette of motivating examples and suffi- cient software to realize their models—on the right platform. The book might be more accurately subtitled ‘Unified Analysis via H-likelihood in Genstat’. A 12- month access code to the Windows version of the Genstat software is bundled with the book, as are the necessary data sets and scripts to repeat many of the analyses. Genstat is required to run the scripts and, indeed, to read the data sets, which are provided, disap- pointingly, as binary Genstat data files. Unfortu- nately the code to run the models has been wrapped in higher level functions, which are presumably op- timized for speed and stability. These functions act as black boxes, which conceal the underlying al- gorithms. Therefore reconstructing the underlying algorithms in other languages is not straightfor- ward. Also, running the models on other data, or with different options, requires care. I have coded and checked some of the basic algorithms in R, and those algorithms that I could construct worked beautifully (R code is available from the reviewer). However, it would have been illuminating to be able to run the raw algorithms, as it were, even if they were slower. The book is sensibly structured. The authors review likelihood theory, generalized linear mod- els and quasi-likelihood compactly yet thoroughly in the first three chapters. The reviews of existing material, mainly in Chapters 1–3, are first rate. They devote Chapter 4 to introducing the extended likelihood, and Chapter 5 to the development of the normal mixed effects model in the context of extended likelihood. Chapter 6 generalizes the principle of mixed effects models into hierarchical generalized linear models. Chapters 7 and 8 add structured dispersion and correlated random effects respectively. Chapters 9 and 10 provide specific applications of the techniques, using smoothing, and survival analysis as motivation. The authors then present double-hierarchical generalized linear models and finish with a pot-luck chapter that covers more complicated situations. The material is arranged carefully and augmented with connecting tissue, including text to help the reader to position the chapter in the larger struc- ture of the book. This linking material proves very helpful, as it assists the reader in cross-referencing the ideas in the book. The important chapters each conclude with bullet points that detail the major issues that are covered by the chapter, providing a quick review and a valuable refresher. Pains have clearly been taken over the structure of the book. The authors have gathered a broad array of prac- tical examples, and provide detailed analyses, up to a reasonable point. The data for the examples and the scripts to run appropriate models are included on a supplied compact disc. As noted above, Gen- stat is required to read the data sets, which limits the usefulness of the data sets somewhat. Exporting the data sets to a more general for- mat is relatively easy, if your computer will run the promotional version of Genstat provided, but, then again, providing them in a more general format