Medicine And Books

Statistics in Clinical Practice

BMJ 1995; 311 doi: https://doi.org/10.1136/bmj.311.7015.1311 (Published 11 November 1995) Cite this as: BMJ 1995;311:1311
  1. Tony Johnson
  1. MRC biostatistics unit, Institute of Public Health, Cambridge

    David Coggon BMJ Publishing Group, £10.95, pp 116 ISBN 0 7279 0907 X

    Over the past 20 years the BMJ has established a fine reputation for publishing papers and books that present basic statistical techniques to a clinical readership. The first book, T D V Swinscow's Statistics at Square One (now in its 8th edition), teaches simple data presentation, the basis of probability and inference, and applications of the more popular statistical tests. Statistics with Confidence (edited by Gardner and Altman, 1989) emphasises the importance of estimation, with details of the calculation of confidence intervals as well as simple statistical guidelines and checklists to aid authors and readers alike; both books were based on a series of papers initially published in the BMJ.

    They are now joined by Statistics in Clinical Practice, an excellent elementary companion text for doctors and medical students (not previously published in the BMJ), which aims “to explain the principles of statistics that must be understood in order to read journals and practise clinical medicine competently,” and with no mathematics beyond that required of school leavers. This is not a textbook of statistical methods, though the first part, which consists of three chapters devoted to types of data and how to summarise univariate, bivariate, and multivariate data, is a lucid account of the basics of data presentation. A remarkable amount of detail--including the calculation of summary statistics, construction of tables and figures, and use of logarithmic scales--is packed into fewer than 50 pages.

    The second part, a single chapter which introduces chance and probability, cleverly interweaves elections and coin-tossing with quantification of clinical uncertainty and explains sensitivity, specificity, and predictive value. The final chapters cover hypothesis testing, as well as the more contemporary techniques of estimation including confidence intervals, statistical power and sampling, and statistical modelling; the last chapter, on the interpretation of statistical analysis, includes assessment of bias, chance error, and cumulation of evidence in overviews (meta-analysis), with some brief advice on working with statisticians. The coverage is sufficient to provide some basic understanding of many of the statistical ideas and techniques reported in research papers in general medical journals; the list of eight books for further reading guides the interested reader to greater depth.

    The presentation is good, with an excellent balance between the text and the tables, figures, and boxes which liberally illustrate it. The book offers many examples drawn from a broad medical background; any general physician and some specialists will find common ground with these. Coggon skilfully uses footnotes to both tables and figures to draw attention to important points of detail; these enhance the text and do not repeat it. Each chapter ends with a selection of questions which test the reader's understanding of the ideas that have been presented; explanatory answers are included along with a comprehensive index.

    I found only three details with which to take issue. Firstly, the suggestion that three-dimensional bar charts--so popular in graphics packages--are satisfactory (the distortion of scale surely renders them useless, and in any case they are, I think, unnecessary); secondly, when illustrating the influence of an outlying observation on the mean but not the median, initial data checking should be advocated; and, thirdly, the choice of a ratio as dependent variable to illustrate multiple linear regression (without warning that analysis of ratio variables requires caution). These quibbles should not deter any doctor from using this excellent book. TONY JOHNSON, MRC biostatistics unit, Institute of Public Health, Cambridge

    View Abstract

    Sign in

    Log in through your institution

    Free trial

    Register for a free trial to thebmj.com to receive unlimited access to all content on thebmj.com for 14 days.
    Sign up for a free trial

    Subscribe