Statistical questions in evidence-based medicine was written long ago by Martin Bland and Janet Peacock (Oxford University Press, 2000), but is is still full of good recommendations. It is a series of case-based questions and answers.
It is supposed to serve as a companion to An introduction to Medical Statistics. I personally bought the third edition of Indrayan’s Medical Biostatistics, which I find very good, and I just ordered Medical Statistics in Clinical and Epidemiological Research, by Marit B. Veierød, Stian Lydersen and Petter Laake.
A related book is: Peacock, J and Kerry, S (2007). Presenting medical statistics. From proposal to publication. Oxford University Press.
Here is a brief sketch of the book. The second chapter asks basic questions about study design, emphasizing the role of randomized controlled trials as a gold standard in evidence-based medicine. The third chapter is about observational studies when no intervention or treatment takes place (cross-sectional, cohort and case-control) and various sources of bias that can arise as a result of the way subjects are included in the study. See this old post of mine to get an overview of this kind of studies and their biases. Chapters 4 and 5 are concerned with summarizing and presenting data. There’s no mention of ICMJE guidelines (more recently, the SAMPL guidelines).
Other articles of interest are listed below:
Here are some things I don’t like in Statistical questions in evidence-based medicine:
In chapter 5 “Presenting Data”, the use of bar charts rather than Cleveland’s dotplot to display counts or frequencies, especially in the working example on the 8 domains of the SF-36 questionnaire for which scores were available in two groups. In fact, there’s no mention of Tufte or Cleveland in the References section, which I find disappointing since bar charts do not provide more information than dot charts and have a poor data-ink ratio. Oftentimes, it also nice to display all individual information as well as summary measures; see, e.g., Beware of dynamite (PDF).
In many Tables, they display redundant information by reporting the statistics used to summarize each variable (in the case of continuous outcomes, mean ± SD). In this respect, I prefer a single legend or annotation indicating what is used to summarize numerical and/or categorical variables, as implemented in the R
summary.formula() function from the Hmisc package. (Stata’s estout command is much more complicated, in my opinion, tabout; see also, Making Working & Publication-style Tables in Stata)