Biostatistics: a computing approach, by Stewart J. Anderson, Boca Raton, FL, CRC Press, 2012, xx + 326 pp., £39.99 or US$79.95 (hardback), ISBN 978-1-58488-834-5
Biostatistics: A computing approach is an intense read. While assuming that readers have at least a basic understanding of statistical concepts, the author emphasises substance over style and develops many tests and equations from first principles. Although this choice does make the text significantly more difficult to read, it allows for the precision and analytical clarity that make this volume a deep and authoritative biostatistical resource. Many statistical textbooks introduce concepts and apply them to simple examples, but the approach taken in this text is that complete, painstaking development of statistical principles, replete with equations, is a better method.
The book builds slowly from first principles, deriving each equation along the way. It is divided into two roughly equal sections (Chapters 1–5 and 6–11), the first introducing a series of basic statistical principles and the second addressing a variety of specific statistical techniques. In the first section, the author introduces five core areas of biostatistical analysis: the principles of probability (Chapter 1), the use of computer simulation (Chapter 2), the central limit theorem (Chapter 3), correlation and regression (Chapter 4) and ANOVA (Chapter 5). In the second section, he writes on risk (Chapter 6), multivariate statistics (Chapter 7), repeated measures (Chapter 8), nonparametric statistical methods (Chapter 9), time to event data (Chapter 10) and power analysis (Chapter 11). The chapters in the first half of the book build directly on each other in sequence, although those in the second are more independent.
The author incorporates statistical computation using R and SAS throughout the book, notably devoting the last two chapters to offer tips while using these languages. The author attends to the basics well, and the material offered may be found to be very useful to the intermediate user of SAS or R: readers unfamiliar with these languages may find the author's introduction too brief, whereas advanced users may be left wanting more. The computational tutorials found throughout this book are most useful when addressing a particular test or statistical principle, although many of these will take time to set up. Curiously, the data sets referred to in many of the tutorials are either held in the program library (the stock files that come with R or SAS) or a paper the reader would presumably have to look up, which is annoying. The patient reader might gain some computational knowledge from mimicking the code provided by the author, but is unlikely to become well-versed enough to perform their own analyses based solely on the information provided in this book.
The strength of this text relies on the expositional clarity of the author's explanation of statistical principles more than its potential to be a complete DIY course in SAS or R. It is a comprehensive book that lets the reader see how statistical arguments are developed and would make an excellent textbook for a statistics course at the graduate level.