‘Quintessentially, statistics is about solving problems: data are collected and statistical analyses are used to provide answers ’ (p. 41); with SAS, I would add. I believe this phrase catches the meaning of the book by Der and Everitt, throughout which the authors have chosen to depict an amalgamation of the applied facet of statistics with SAS along with theory, creating an enjoyable read. Overall, I think the authors have fulfilled their aim of giving a disquisition on medical statistics with SAS.
This book has 18 chapters which cover a full range of topics in medical statistics. It starts with introducing basic commands in SAS and providing examples (DATA and PROC), moves on to the issue of measurement in medicine, gives an elaboration on clinical trials and epidemiology (essential in medical statistics) and then moves on to statistical methods. I enjoyed these initial chapters, because the authors provide a taste of history and epistemology in this field which can become too theoretical at some points.
Interestingly, the methods section of the book starts with meta-analysis which is not the first statistical method a researcher in medicine or health sciences would use. Possibly this choice is justified because meta-analysis is an increasingly popular manuscript type for journals and many researchers are attempting this. Then it moves on to describe the analysis of variance and covariance, which is basically the statistical methodology to explore differences between groups with respect to certain variables. It gives certain types of analysis of variance depending on the design of an experiment and it explores the non parametric case as well, then moves on to simple linear regression and ways of plotting data.
Before moving on to more complicated techniques, the authors have not described thoroughly t-tests or cross tabulations, which are simple univariate tests used to describe differences or associations. This might prove a downfall for the authors and their audience, since these are simple tests yet frequently used by applied researchers. It is not to say that they are not mentioned at all, but the discussion is more shallow than one would expect.
Next the authors move on to describe more complex statistical analyses in medical statistics, including multiple linear regression, logistic regression and the generalized linear model. This is followed by a short chapter on generalized additive models. The authors devote three chapters (80 pages) to the analysis of longitudinal data with a thorough statistical basis in each chapter as well as examples and plots. They touch difficult topics, such as linear mixed models and non normal responses. SAS can be highly efficient in performing such analyses for clinical studies adjusting for different covariance structures; and it is not without surprise that biostatisticians have become accustomed to this software.
Final topics in the book include survival analysis, Bayesian methods and dealing with missing values. Analysis of survival data enjoys a niche throughout the book along with previously mentioned longitudinal data. Unlike Dorothy Alisson's quote ‘ \ldots survival is the least of my desires ’, in medical statistics survival analysis becomes an essential part of a researcher's endeavours with questions always arising concerning time to death or an event.
The book ends with a pleasing index which proves quite useful and extensive (23 pages) making it a good reference both for SAS commands and statistical topics. Finally, as the authors state, ‘ complete data sets, all the SAS code, and complete outputs can be found on an associated website: http://support.sas.com/amsus ’.
I think the audience this book addresses is a statistically knowledgeable researcher with solid base in software that uses commands, most likely a medical statistician and not a clinician. However, a positive aspect of the book that could broaden its audience is that it serves as a reference source for medical statistics, due to its solid grounding in theory which is present nicely in the chapters of logistic regression, ANOVA and meta-analysis.
Overall, I would recommend academic libraries to purchase this book, should they wish their research community to have access to this effective resource.