Statistics from Altmetric.com
Edited by Antony Stewart. Radcliffe Medical Press Ltd, 2002, £19.95, pp 151, softcover. ISBN 1857755898
In the preface, the author writes “Indeed, the avoidance of too much detail and too many theories is a prime objective”. I feel that this aim prevents the book from becoming a required text for either students or practitioners of epidemiological research, as the author’s writing style is engaging and very easy to follow. As a result, I struggle to imagine to whom this text is targeted. A student of epidemiology requires far more detail and, most likely, would find this text superfluous. At the other extreme, the experienced and practising epidemiologist would be knowledgeable of (almost) all the topics covered and to a greater depth than that provided.
There were a number of items that I did, however, enjoy seeing presented to the epidemiological community. Discouraging the use of special effects in graphics, the opening paragraph explaining the concept of standard error, extolling common sense when interpreting p values greater than (but close to) the traditional significance level, and the flow chart approach to explaining the difference between cohort and case-control studies are some of these. As has already been alluded to, however, these were let down by the negligible focus on any of these topics.
Another reason for my reluctance to recommend this text is the treatment of topics that seem to be placed in the “too hard basket”. These include two sample t tests (unequal variances), correlation, and the entire topic of non-parametric techniques. However, these omissions paled in comparison when I realised that the reader is referred to other texts for discussions on sample size and power (the cornerstone to any sound epidemiological design) and also for regression. Regression, arguably the single biggest contribution that the field of statistics has made to epidemiological analysis, is given one line of recognition! I have kept in mind that this is intended to be a “basic” text and a “practical” guide, but my belief is that these topics should always be included in such texts.
As in (almost) all textbooks, especially first editions, there are a small number of errors. For example, on page 40 it states that: “…the sample mean is 82.696, there is a 95% probability that the population mean lies between 80.509 and 84.883.”. In fact, the probability that the population mean lies between these two limits is 0 (zero) or 1 (one). To my mind, this is the only error worth noting, and I am impressed that the author has touched on such a large range of topics, albeit briefly, and been accurate in all the information provided.
In conclusion, I leave you with a summary that is possibly difficult to understand. I find it difficult to define an audience to whom I would recommend this text while, at the same time, I look forward to seeing a more thorough treatment of the topic by this author.
Evidence basis 15/20
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.