On Tue, 3 Apr 2007, Stuart McKelvie went:
Dear Rick,
Because of the protection for Type I error, the largest and smallest
mean may indeed not differ with the post hoc test.
But I would argue that they MUST differ on purely logical grounds.
If that's the case, then Rick should be able to re-run the omnibus
ANOVA with just those two groups--the groups with the largest and
smallest means--and still get a significant F. I'm not suggesting
that as an actual statistical practice, but only as a test of Stuart's
statement. I have no idea whether it's true.
--David Epstein
[EMAIL PROTECTED]
---
To make changes to your subscription go to:
http://acsun.frostburg.edu/cgi-bin/lyris.pl?enter=tips&text_mode=0&lang=english