Hi Gang,
I won't worry about the spurious correlations at the edges of the brain.
But, I ran 3dttest for one of the first order t-tests in my experiment
(SvNS: switch vs. no-switch) and compared the map to the t map from
running GroupAna. And, now I am really puzzled. The t values and probablilities are not that close. If I threshold the maps at p=.005,
in the t map from 3dttest, I have lots of significant clusters. In the
t map from GroupAna, I have very few. I'm not sure what could account
for this difference, unless the way I have set up the analysis is not
equivalent.
What got me to this point was that I was trying to define ROIs from
my random effects ANOVA and t-tests (run with GroupAna), but found
that I had very few clusters surviving at p<.005 or p<.001, even for
regions that are known to be associated with the contrast of interest.
--Stacia