Date:         Thu, 16 Apr 1998 23:15:58 GMT
Reply-To:     Big Kahuna <FakeName@FAKEADDRESS.FAK>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         Big Kahuna <FakeName@FAKEADDRESS.FAK>
Organization: EarthLink Network, Inc.
Subject:      Repeated measures design: Homogeneity Assumptions

Hello.  This is less of an SPSS question than a statistical question, but
since I used SPSS to anlayze my thesis I thought it would be ok to ask.
 
I did a Brown-Forsythe test to determine if my independent variables in my
thesis evidenced heterogeneity of error variances between my IV levels, and
also did Mauchly tests of sphericity to check the constant correlation of my
var-covar matrices which is required for the univariate approach to repeated
measures  (I have a three-way fully repeated measures design).  The outcome of
the tests indicated that for certain levels of the IV's I had both a
heterogeneity of variance problem (sig. Brown-Forsythe) and a problem with my
var-cov matrices (sig. Mauchly).  I used the Greenhosue-Geisser epsilon
correction to my degrees of freedom (as printed by SPSS MANOVA) to determine
the statistical significance of my results.  My question is:
 
1) Knowing that ANOVA is robust to minor violations of the homogeneity of
variance assumption when n is large (I have 80 subjects in each cell -
although they are the same subjects since it was a repeated measures design),
and
 
2) Knowing that Greenhouse-Geisser tends to be more conservative than might be
necessary, although it theoretically allows me to still interpret my effects
when my var-covar matrices do not have a constant correlation and do not have
circularity,
 
Would I need to also transform my DV's to take care of the heterogeneity of
variance problem, or can I assume that the Greenhouse-Geisser correction
is sufficient to have taken care of any positive bias related to both
problems?
 
Obviously I want to do proper research, but I also want to some day get out of
my university with my masters before the millenium.  Is the above adequate?
 
Thank you in advance,
 
Scott Masten
 
scottmasten@earthlink.net