Are there any BMJ readers who have not encountered statistics and statisticians at some point in their career? Most medical students, and many other health professionals, have learnt some statistics as undergraduates, and maybe more as part of their postgraduate training. If you have obtained a grant or published a paper, the chances are that some statistician will have acted as statistical consultant or referee, if not as collaborator and coauthor. If you prescribe drugs, you are relying on the many statisticians employed within the pharmaceutical industry, and those involved in the regulation of drugs. Have you ever wondered why statistics and statisticians are now such an integral part of medicine? What do statisticians do, when they are not performing emergency χ2 tests and last minute power calculations, transforming messy data into elegant analyses, or training the next generation of doctors? In fact, why did they get into that line of work in the first place? How did statistics ever develop as a subject, and when did people start using it to address health and medical problems?
Figure 1.
Stephen Senn Cambridge University Press, £14.99, pp 264 ISBN 0 521 54023 2
Rating: ★★★★
If you need to calculate a confidence interval, interpret a logistic regression, or defend your choice of a fixed or random effects meta-analysis to a referee, this is not the book for you. There are many excellent textbooks about medical statistics on the market; my personal favourite is Altman's Practical Statistics for Medical Research (Chapman & Hall, 1991). But Dicing with Death is not just another statistics textbook. This is a book that explores the history and the intellectual challenge of the subject, the personalities who shaped it, and the fundamental questions with which they grappled. In this volume you will meet familiar statistical names such as Fisher, Pearson, Poisson, and Student, and others you may not so readily associate with statistics, such as Edmund Halley. You will grapple with Simpson's paradox, the Bayesian-frequentist debate, baptism data from the 17th century, and the earliest randomised trials. Moving to current developments, you will see meta-analysis and the MMR (measles, mumps and rubella) debate in their statistical-historical context. The remarkable achievement is that all of this, and much more, are in a slim paperback that is best read cover to cover.
Inevitably, given the scope of the book, there is some unevenness. I wanted more on Florence Nightingale, whose brief mention underplayed her contributions to statistics. A few sections are harder going than others, although asterisks are used to indicate these. And some of the jokes are truly dreadful, although most are apposite.
Stephen Senn has attempted to do for medical statistics what Stephen Hawking did for physics in A Brief History of Time (Bantam Press, 1988) and Simon Singh did for pure mathematics in Fermat's Last Theorem (Fourth Estate, 1997). I think he has succeeded.