Gerald van Belle’s Statistical Rules of Thumb has a free chapter on sample size available online, which is a great way to get a feel for the content of the book. It was how I found out about it. The title was alluring, though I was skeptical at first if this is going to be one of those books that, like the much reviled p<0.05, suggests a set of rules without a good explanation. On the contrary, the book does a great job of explaining most of the rules spending, on average, 1.9 pages per rule (for the numerically inclined). It is a good sweep through the statistical space with an emphasis on practical application of the rules, and ample references for further study. A few of the Amazon reviews have called out certain domain specific chapters, such as the ones on Epidemiology, Environmental Studies, and Evidence-Based Medicine to be non-relevant. I thought they provided a good insight into statistical analysis in fields outside of one’s own. Given the widespread use of statistics for online and mobile experiments, I wish there was a chapter that covered this space as well.
I do have a couple of critiques. For one, there are more typos than I have encountered in a technical book. There is at least one instance of a reference in the text not listed in the references section at the end. There are a few rules where one does feel like a more detailed explanation would have been helpful. The last two chapters - on Words, Tables, and Graphs and another on Consulting - are fairly generic, somewhat outmoded, and don’t do justice to the title of the book.
The wisdom in the book outweighs its minor flaws, and I would highly recommend this as a beginner to intermediate level book on familiarizing oneself with the statistical space. This is by no means a textbook and doesn’t take you through the slow gradual process of learning. It is rather like an anthology that gives you a flavor of the key aspects of the field, and then lets you work your way through topics of interest using the cited references.
I have documented a few notes for personal reference here. They may not make much sense without the context provided in the book.