Handbook of regression and modeling: applications for the clinical and pharmaceutical industries Paulson DS (2007) ISBN 157444610X; 520 pages; £57.99, $104.95 Chapman & Hall/CRC; http://www.crcpress.com/


The intended audience of this book is clinical and pharmaceutical researchers and scientists. Specifically, it is written for those researchers that need more than just simple linear regression methods to meet their research needs. To fully benefit from the book, the reader needs to have at least a basic knowledge of matrix algebra, although this is not a prerequisite, since the book has an appendix that gives the basics in matrix manipulation. It should also be noted that even though it is not clear from the book's title, the regression methods discussed in the book are limited to basic linear models. Thus, other regression methods, such as logistic, survival or nonlinear models, are not discussed.

The book consists of 11 chapters and two appendices. The first chapter gives a general overview of basic statistical concepts, such as hypothesis testing and confidence intervals, as well as a general discussion of the design and nature of experiments and empirical research. The second chapter introduces simple linear regression, discusses the basic concepts of regression analysis, including estimation, inference, diagnostics and model evaluation, and also more special subjects such as power analysis, piecewise regression and comparisons of multiple simple linear regression functions. Problems with serial correlation in simple linear regression, and how this is remedied, are discussed in Chapter 3. The fourth chapter gives an introduction to multiple linear regression analysis by showing how it is a direct extension of the simple linear regression case. While the previous three chapters have used standard algebra, this chapter and most of the following ones are relying on matrix algebra. The fifth chapter gives a short discussion about correlation analysis in multiple regression models, while the sixth chapter is devoted to the problem of collinearity, discussing how it is measured and remedied. Polynomial regression, including piecewise polynomial regression, is the topic of the seventh chapter, Chapter 8 is concerned with some special topics in multiple regression analysis, including interaction, confounding, heteroskedasticity, outliers, leverage and influence, while Chapter 9 gives a lengthy discussion about indicator variable regression. Chapter 10 provides an overview of model building and model selection methods, including stepwise regression, forward selection, backward selection and best subset procedures, while analysis of covariance is the topic of the last chapter. Finally, there is an appendix containing tables of critical values and another appendix about matrix algebra.

A distinctive feature of this book is the systematic layout of hypothesis testing, where the author uses the same algorithmic six-step procedure throughout the whole book. The discussed methods are richly illustrated with applications using biological, medical and pharmaceutical data. The author uses Minitab for the examples, but all necessary formulas are given for the user who wants to calculate these by hand or code it into other programs. The book is also richly illustrated with useful diagrams, graphs and tables. All data sets used in the book are presented in their entirety in the book, and one thus does not have to rely on a CD or the internet to get access to these. Although this in general is a good feature, it should also be supplemented by having the data files available from a website. Having the data only in printed form makes it very impractical for the reader to try the analyses by him/herself. Apart from this drawback, the book is well suited for medical and pharmaceutical researchers and scientists with only a basic knowledge of statistics who want to learn about linear regression analysis and modeling.