Introduction to the Design and Analysis of Experiments. G. M. Clarke and R. E. Kempson, Arnold, London, 1997. No. of pages. vii+334. Price: £19.99. ISBN 0-340-64555-5
✍ Scribed by Richard Jarrett
- Book ID
- 101239397
- Publisher
- John Wiley and Sons
- Year
- 2000
- Tongue
- English
- Weight
- 49 KB
- Volume
- 19
- Category
- Article
- ISSN
- 0277-6715
No coin nor oath required. For personal study only.
✦ Synopsis
move further away from the null when adjusting for too may confounders (p. 258). It is also somewhat surprising that the weighted average of stratum-speci"c estimates obtained by (direct) standardization should be easier to interpret than a pooled estimate when, in fact, there is heterogeneity between strata (p. 266), and that in a "ve-year age-matched study one may remove &any residual age confounding' by adjusting for age in one-year intervals (p. 282). Finally, the discussion on competing risks (p. 288) is rather super"cial and the discussion on &statistical interaction' versus &biological interaction' leaves the feeling that the latter concept, which again builds on counterfactuals, is of more theoretical than pratical interest.
Part III ends with two chapters on &Introduction to regression models' and &Introduction to regression modeling' which again contain several interesting and sensible points of view. However, the discussion in Chapter 20 is slightly limited by considering regression exclusively as &conditional expectations' and not as &conditional distributions'. This has the consequence that the discussion on regression analysis based on rates (&Poisson' and &Cox regression') is not as elegant as in the book by Clayton and Hills. A distinguishing feature of Chapter 21 is the inclusion of discussions of recent work by Robins and others on topics like timedependent confounding and structural nested failure time models and also on hierarhical or multi-level models. These are all concepts and methods which are likely to be highly in#uential in the future.
In conclusion, the whole book is likely to be strongly in#uential. It represents an impressive piece of work and provides an authoritative and well-written source for epidemiologists. A minor criticism is that the typography does not quite live up to today's standards. Some more severe criticisms have been mentioned above.