Handbook of Statistics 13: Design and Analysis of Experiments

point out that both volumes are available for $307, which saves $167 over the prices of the individual volumes purchased separately. Ziegel (1998) was generally positive about Part A, other than some criticism of the audacity of a group of chemists who were essentially writing a statistics book. Unlike Part A, which emphasized data collection and focused mostly on standard univariate statistics topics, Part B is mostly concerned with multivariate statistics and their unique applications that arise in chemistry. In these areas, the chemists have made some significant contributions-most of them, in fact. So here in Part B one encounters QSAR (quantitative structure activity relationships), multivariate calibration, sensory data, pharmokinetic models, and many unique chemistry applications of mostly multivariate statistical procedures. These authors’ expertise for this material is unquestionable. Though I credit the chemometricians with some imaginative use of mathematics and statistics, they do not need to present a large body of material on matrix algebra too. So the first chapter, here actually Chapter 29, serves up 52 superfluous pages. Thereafter I cannot fathom any logic in the presentation order of the topics. Chapter 30 is “Cluster Analysis,” offered as straightforward statistics-book stuff. Chapter 31 is “Analysis of Measurement Tables.” Because rows are observations and columns are variables, for these authors in this context this is just typical multivariate data. Mostly their data analysis is ordinary principal components, but nonlinear analysis and three-way tables are discussed too. Chapter 32 deals with a mostly mathematical matrix approach to the analysis of contingency tables, which ends with correspondence factor analysis and log-linear models. Shifting gears, the book next considers “Supervised Pattern Recognition” as Chapter 33. A wide-ranging set of topics begins with discriminant analysis and ends with neural networks. A fairly classical chemistry problem, resolution of curves and mixtures, provides the focus for Chapter 34. First standard factor-analysis methods are presented, followed by a host of variations and related techniques. Continuing to weave its way among topics, the book offers methods for relating many dependent variables to many independent variables as Chapter 35. These methods include Procrustes analysis, canonical correlation, multivariate regression, reduced rank regression, principal-components regression, partial least squares, and continuum regression. Chapter 36, “Multivariate Calibration,” follows naturally as a principal application. Shifting gears again, Chapter 37 is “QSAR,” which relies on methods discussed previously. Totally unrelated by application or techniques is “Analysis of Sensory Data,” Chapter 38. Somewhat more related are the “Pharmokinetic Models” of Chapter 39, a huge chapter mostly devoted to compartmental models. The next direction is “Signal Processing” in Chapter 40. Another lengthy chapter, it gives everything one could want to know about Fourier transforms, ending with a discussion of other deconvolution and transform methodologies. “Kalman Filtering” follows as Chapter 41. The chapter begins with general material on recursive methods. The last part of the book starts with a rather useless chapter on operations research. The last two chapters, which cover artificial intelligence and neural networks, are a fitting conclusion to this modern update of chemometrics. These chapters, like all the others, have great sets of references, mostly to the chemical literature. There is a 12.page index. Part A of the set was a nice addition to my library, but, as a statistician, I did not need it. Topically, Part B, however, is really outstanding and seemingly very complete. Disregarding the few previous negative comments about topics inclusion and chapter organization, this book probably has everything I will ever want to know about chemometrics. Much as I hate to recommend that anyone plunk down $201 for a book, if one is involved in any way with chemometrics applications, its purchase is a must. It is as big as two books, which makes the price seem a little less exorbitant. Thanks to the six authors and Elsevier for undertaking this project and doing it very thoroughly.