The Statistical Evaluation of Medical Tests for Classification and Prediction
暂无分享,去创建一个
the text. Briefly, ordinary least squares regression and maximum likelihood estimation in regression, along with matrix algebra (with a tutorial in the Appendix), are reviewed. Techniques are discussed for managing unequal variance around a regression line and for the analysis of data collected via probability sampling. Methods for nonresponse and missing data are touched on. The second half of the book concentrates on the analysis of binomial and Poisson outcomes, as well as on the modeling of correlated data with random (mixed)effects models and generalized estimating equations. In general, this is a well-written, clear, and concise text with interesting and elaborate examples illustrating several of the “messy” situations often encountered in practice, including unbalanced and missing data. SAS code and output are provided for every example, and the author goes to some length to compare and contrast several SAS procedures. She provides some useful insight into the procedures, going beyond what is typically presented in SAS manuals. Although the author does not mention any other statistical package, the descriptions of the analysis techniques contain sufficient detail to allow the reader to apply the methods using some other package. To non-SAS users, the meticulous coverage of the various SAS procedures may turn out to be tiresome; examples from another statistical package would make the book more directly appealing to a wider audience. Palta uses basic statistical theory to motivate and connect the various topics. The theory provides a framework for appreciating why certain techniques have been developed, and gives the reader tools for understanding the methods used. The author succeeds in making the text accessible to students who have had several courses in applied statistical methods, by making her explanations descriptive and conversational. However, the book might be better suited to readers with some background in theory, in that concepts such as exponential families, consistency, and link functions would not be new to them. Although the book does not provide exercises, it could be used as a text for an upper-level applied methods course. The chapters follow a natural progression, with a few shorter ones reviewing relevant topics like transformations and likelihood analysis. This text would be useful to many applied statisticians looking for an applied, yet thorough reference on the analysis of correlated data. The connections made between the analysis of survey sample data and the analysis of other correlated data, often omitted in similar texts, help to provide a well-rounded picture of the analysis of dependent data. Because this text is an overview of many regression topics, a more extensive bibliography would be useful to the reader. For example, references following each major chapter could be given to emphasize the important seminal works in each area and to suggest further reading for those interested. The articles listed as suggested reading at the beginning of the book do provide some additional examples of the techniques to follow, but are limited to those by the author. Overall, Quantitative Methods in Population Health: Extensions of Ordinary Regression would be a worthwhile addition to the libraries of applied researchers in the health sciences. It is priced similarly to other books in the field, and is one reference that covers numerous topics. This well-organized book can serve as a refresher of basic topics for more statistically sophisticated readers and as a tutorial for those without the background. The types of analyses detailed in the text are often called for in practical situations.
[1] Chris Lloyd,et al. Using Smoothed Receiver Operating Characteristic Curves to Summarize and Compare Diagnostic Systems , 1998 .
[2] C. Le,et al. ROC curve estimation based on local smoothing , 2001 .