On Wed, 10 Mar 1999, Stephen Black went:
> I counted three separate statistical tests on the data. Especially
> given that the result (smaller cell bodies and nuclei) was unexpected,
> a correction for multiple comparisons might have been prudent. Using
> the Bonferroni, the critical test would then be at .05/3 = 0.017.
>
> None of the values he reported would have been significant at that
> level.
But I would hate to see something go unreported due to a Bonferroni
correction. Better just to say "Here's what we found in exploratory
analyses."
Actually, if you want to see how conentious this issue can be, and if
you enjoy watching statisticians mud-wrestle one another (who
doesn't?), read this article and the disdainful replies it got, all on
the Web in full text:
www.bmj.com/cgi/content/full/316/7139/1236
Thomas V Perneger.
What's wrong with Bonferroni adjustments.
British Medical Journal 1998;316:1236-1238 (18 April)
(Oh, and while you're at it, check out the public humiliation that
befell the first group of authors to submit to "on-line peer review":
www.bmj.com/misc/peer/index.shtml)
--David Epstein
[EMAIL PROTECTED]
