Statistical Computing: An Introduction to Data Analysis Using S-PLUS

First, a clariŽ cation: This book is not about traditional statistical computing topics, such as random number generation, approximation of probability distribution functions, least squares computations, and so forth. The book’s subtitle reveals what it is about: “An Introduction to Data Analysis using S-PLUS.” Crawley’s intention, as stated in the Preface, is to provide “both an introduction to and a reference manual for statistics and computing.” (The commercial package S-PLUS appears in the subtitle, but the Preface notes the availability of the freeware R system, which can be obtained at www.r-project.org. The author states that all the examples in the book will also work in R. Based on my experience with the two packages, some code tweaking may be required when using R, but you cannot beat the price.) My Ž rst reaction when I began to peruse this book was “wow”: More than 750 pages, covering a vast range of topics in statistical modeling. Roughly speaking, the Ž rst third of the book covers material suitable for a Ž rst course in statistics, the middle third covers basic applied linear models (regression and analysis of variance), and the Ž nal third covers a number of advanced statistical modeling methods, including generalized linear models, tree-based models, nonparametric smoothing, survival analysis, time series, mixed-effects models, and spatial statistics. There is a certain logic to the sequencing of topics in the text, but Crawley has made some nonstandard (one might even say quirky) decisions. For instance, power calculations are discussed three chapters before the classical hypothesis testing procedures are introduced. The early chapter on basic graphical techniques includes displays of tree-based models but excludes probability plotting. The formal introduction of probability distributions is quite thorough, but does not appear until Chapter 26, two-thirds of the way through the book. A number of examples involve the use of procedures that are not formally introduced until much later in the book. (Including more cross-referencing when this happens would make the book more effective as a self-study guide.) On balance, I can recommend this book for the intended audience, as long as they understand what they are in for. It contains a wealth of sage advice about the process of designing studies, analyzing data, and Ž tting statistical models and could become a treasured reference. However, the style is fairly terse and fast-paced, and neophytes may Ž nd it somewhat overwhelming. In the Preface, Crawley says “Learning S-PLUS will not be easy, but you won’t regret making the effort.” The same can be said for reading this book.