LETTER TO THE EDITOR: How independent are multiple ‘independent’ diagnostic classifications? by H. Brenner, Statistics in Medicine, 15, 1377–1386 (1996)
✍ Scribed by Roger Marshall
- Publisher
- John Wiley and Sons
- Year
- 1997
- Tongue
- English
- Weight
- 49 KB
- Volume
- 16
- Category
- Article
- ISSN
- 0277-6715
No coin nor oath required. For personal study only.
✦ Synopsis
It will be seen from the table, that when performed in a way which gives a fair comparison with the CROS procedure, the two-stage procedure is in no way superior to it. This justifies, if further justification were needed, the claim in my own book that the two-stage analysis should never be used. I hope that Jones and Lewis can agree with me that is the case and that we can finally dump this illogical approach to analysing cross-over trials. This does not mean, of course, that we should dispense with the AB/BA design. I agree with Jones and Lewis that this is extremely useful. Indeed in Cross-over ¹rials in Clinical Research (although unfortunately one reviewer appears not to have got this message) I stated 'I am firmly convinced that contrary to what has been claimed by many statisticians the AB/BA design is as reasonable, if not more reasonable, than any other' (p. 228). It is time, in my opinion, that we restored this design to the trialist and stopped looking for statistical solutions to what is a pharmacological problem.