is now in its fourth edition. It is "designed as a resource for practicing and student toxicologists", and should give the reader the "tools necessary to perform rigorous and critical analysis of experimental data and the insight when to use them."
The book discusses basic principles, actual techniques, and claims to review all major applications in toxicology.
The book can be roughly divided into three sections: a general introduction (four chapters), an overview of basic statistical techniques (nine chapters), and finally a review of the use of statistics in specific areas of toxicology (five chapters). Furthermore, there are chapters on epidemiology, structure activity relationships, good laboratory practice, and areas of controversy in statistics as used for toxicology. The book contains a wealth of examples. Where relevant, each chapter contains worked examples and basic SAS code to carry out the analyses.
The text is not always easy to follow. For example, the notion of sample size estimation is introduced under the heading "censoring". Here, sample size calculations for a two-sample t-test are introduced. It is stated that a good approximation can be generated by using z-values instead of t-values in the respective formula. In the case of toxicology this seems to be a questionable advice. The chapter includes an example of a sample size calculation (difference between two groups of 0.7, common standard deviation of 0.825, a ΒΌ 0.05, power ΒΌ 90%, 2-sided). Using the normal approximation, a sample size of 15 animals per group was derived. Using the appropriate t-values yields a sample size of 31 animals per group! Further annoying errors remain present. For example, in the very first table in the book (Table 1.1: Some Frequently Used Terms and Their General Meaning), the meaning of a p-value is given as "Another name for significance level; usually 0.05."
There is a new chapter about Bayesian Analysis. There is an obvious typographical problem, which results in the Bayes formula and several other terms not being represented correctly. This obscures the readability of the chapter, making it difficult for someone without experience to follow the examples.
The chapters on the specific toxicological applications may be read as "stand-alone". This causes, however, that parts of the texts are repeated. For example, the text of pages 248 and 266 on one-sided tests are almost identical.
The chapter I turned to first was the chapter on the statistical controversy. There are discussions on censoring (meaning all data not included in the analysis), the direction of hypothesis testing, unbalanced designs, and the use of computerized statistical packages. In the latter, S. Gad warns, rightly so, against "fishing expeditions", and stresses the importance of understanding the limitations and proper use of statistical methods that are automatically employed. This choice of controversial issues seems to be somewhat outdated. I have missed, for example, a discussion on the general use of statistical tests in toxicology where a "proof of safety" is intended.
Although certainly a matter of opinion, I wonder why any statistics book would have 58 pages of tables with information like logarithms, t-test critical values, and z-scores: all information that researchers and students have at hand even in Excel. The tables lack any explanation on how to use them, and the reader needs to work through the book.
In conclusion, the examples that are provided are a great asset, and provide a nice introduction to some basic ideas about statistics in toxicology. More recent statistical techniques (for example the use of mixed models) are completely ignored. I would wish that the basic principles of experiment design and statistical analyses be handled with a bit more care.