medical statistics. a commonsense approach. m. j. campbell and d. machin, wiley, chichester, 1999....
TRANSCRIPT
STATISTICS IN MEDICINEStatist. Med. 2001; 20:1887–1889
BOOKREVIEWSEditor: Neils Keiding
1. M. J. Campbell and D. Machin, Medical Statistics. A Commonsense Approach.
2. M. Elizabeth Halloran and Seymour Geisser (eds), Statistics in Genetics.
3. Paul D. Allison, Multiple Regression: A Primer.
1. MEDICAL STATISTICS. A COMMONSENSE APPROACH.M. J. Campbell and D. Machin, Wiley, Chichester,1999. No. of pages: viii + 203. Price: $ 16.99.ISBN 0-471-98731-2
This is the third edition of the book, suggestingthat it has been a successful publication. Whileneither all successful books are good nor all goodbooks are successful, this book is both. Why? Itis written for clinicians, and if a book can be rec-ommended for (i) taking away a fear of statistics,(ii) emphasizing convincingly the importance ofcareful planning and analysis of a study and (iii)demonstrating that central principles of statisticscan be understood without having a mathematicalbackground, then this book should be on this list.In only 200 pages, Campbell and Machin manageto cover topics which, as they write in their pref-ace, ‘tells one which statistics to use to summa-rize data and the assumptions underlying statisticalmethods, but also, and perhaps more importantly,discusses what restrictions there are to using themethods and when not to use them’.
The ten chapters follow a similar structure: asummary is followed by the description of the re-spective methods. These are always accompaniedby examples from the literature. The last sectionof a chapter is ‘points when reading the literature’which gives very helpful hints on assessing thequality of the design or method of analysis used.Formulae are given only where necessary. Never-theless, the authors manage to explain even morecomplicated topics, such as logistic regression orbasic elements of survival analysis, in a clear andillustrative way. A very nice chapter is the lastone: ‘Common pitfalls in medical statistics’. Herethe authors present invalid and valid approachesto some problems, for example on the use of the t-test. The reader should take time to read this care-fully, as understanding why a certain procedure is
wrong is the best check as to whether one has reallyunderstood a method.
Any more technical derivations are put into Ap-pendix I. Here the reader will Dnd a somewhatmore rigorous treatment of major statistical tests,conDdence interval construction etc. Appendix IIcontains sets of not too simple multiple choicequestions and answers. These are useful althougharguably some cannot simply be answered by ‘true’or ‘false’.
Any text with this book’s readership in mind,and with limited length, must necessarily lack fullmathematical precision. It follows that some state-ments are not completely correct. For example, itis stated that the mean can be used for summariz-ing nominal scaled variables – this however onlyholds for the example given if there are only twocategories which are scored 0=1. A somewhat moregeneral point concerns Chapter 9, ‘Designed ob-servational studies’. Here the authors manage tointroduce rates, cohort studies, case-control studiesand standardization without mentioning or deDn-ing epidemiology as the main Deld of applicationfor these study types. There are several good text-books on advanced statistical methods in epidemi-ology which could have been cited.
All these, however, are only minor critiques.In general this book can be highly recommended.The layout of it is good, the selection of topicscovered is appropriate and readers will get a verygood Drst overview and a sound basis before usingmore comprehensive textbooks for a more thor-ough treatment.
HEIKO BECHERRuprecht-Karls Universit$at Heidelberg
Abteilung Tropenhygiene and $o'.Gesundheitswesen Im Neuenheimer
Feld 324 D-69120 HeidelbergGermany
Copyright ? 2001 John Wiley & Sons, Ltd.