I'm looking at an fMRI experiment with two groups of subjects and many
conditions. However, for the time being I'm interested in seeing if there
are any differences between only three conditions, A, B, and C (out of a
total of more than three). More particularly, I'm interested in seeing
whether there are any interactions between these three conditions, and the
factor of "group".
At a single subject level, I'd do an F-test in SPM with the contrast matrix
1 -1 0
1 0 -1
the third (0 1 -1) comes for free of course.
To do the intended group level analysis, is it valid to take the two
contrast images (formed using the weights in the rows of that contrast
matrix) to the second level and then do a 2 X 2 ANOVA? And if so, would
one have to allow for inhomogenenity of variances? (Don't need
nonsphericity correction since in this case no factor has more than two
levels.)
TIA,
sjfromm
|