Gang,
I am in the middle of investigating the same problem with the
higher order contrasts from GroupAna. The F tests and first
order t-tests look reasonable, but the seond order t tests have
many significant voxels outside the brain. This has been puzzling me for
the last week because the data was masked using automask during
3Ddeconvolve and the F tests and first order t-tests look similarly
confined to voxels in the brain. I had just decided last night to
try 3dttest to run a comparison of my t-test results from GroupAna.
I wasn't going to post until I was sure that something was wrong,
but then I saw this post this morning...
--Stacia