APPLIED MIXED MODELS IN MEDICINE, 2ND ED.
Authors: Helen Brown; Robin Prescott
Bibliographic Data: John Wiley & Sons, Inc., 2006. ISBN: 978-0-470-02356-3, Series Title: Statistics in Practice, 473 pages, hard cover, $130.00.
Reviewer's Expert Opinion
Description: Mixed models, also known as multilevel models in the social sciences, allow both fixed and random variables within a statistical analysis. The general linear model assumption that the error terms are independent and identically distributed is relaxed in mixed models so that observations can be correlated (e.g., repeated measures, cross-over trial, etc.). This second edition describes current methods and advanced SAS techniques for applying mixed models. The first edition was published in 1999. Purpose: The authors present the theory and application of mixed models in medical research, including the latest developments in bioequivalence, cross-over trials, and cluster randomized trials. Their purpose is to make mixed modeling easily accessible to practitioners such as medical statisticians and economists. The book is well written, thorough, and highly applicable. The examples, complete with SAS code and output, are outstanding. The authors meet their objectives. Audience: This is written for applied statisticians working in medical research and the pharmaceutical industry, as well as teachers and students of statistic courses in mixed models. The authors are experienced statisticians and highly credible scholars from the U.K. The book is part of the Statistics in Practice international series of books that provide statistical support for professionals and researchers. Features: Beginning by describing the capabilities of mixed models, the authors introduce readers to the general linear model for fitting normally distributed data and then extend the general linear model to general linear mixed models. The authors then examine how mixed models can be applied with categorical outcome variables. Chapters 5 to 7 are devoted to practical application of mixed models using particular designs. Chapter 8 includes a new section on bioequivalence studies and cluster randomized trials. Chapter 9 concludes by describing software options, including SAS and the PROC GLIMIX and PROC GENMOD procedures. The reference pages on mixed model notation, the glossary of terms, and the detailed SAS programming code and annotation greatly enhance the book's usefulness. Assessment: This is an excellent resource for biostatisticians and medical researchers. It provides the reader with a thorough understanding of the concepts of mixed models. There are many social science texts on mixed modeling (also called multilevel modeling), but few that clearly link mixed models to clinical research designs. The second edition uses the 9th version of SAS and expands the coverage of categorical outcomes.
Sharon M. Homan
(St. Louis University School of Public Health, St. Louis, MO)