by naught@[EMAIL PROTECTED]
Apr 21, 2008 at 11:26 PM
If the null hypothesis is not true, then the error is not preserved for
the individual ANOVAs. I'm not a fan of univariate ANOVAs as follow-ups
to MANOVA--the two approaches are almost entirely different and have
different aims. My preference for follow-up of MANOVA (if I have to use
MANOVA) is to describe the discriminant weights.
If I'm just interested in controlling the error rate without regard to
some underlying discriminant structure, I'll just use ANOVAs with some
correction for multiplicity.
Mike Babyak
