Dear Juraj,
maybe there's an option to leave out one of the three levels of the corresponding factor? With a 2x2 within design you could build three T contrasts "corresponding" to the two main effects and the interaction on single-subject level at least. You could forward these con images into three separate group analyses (two-sample t-test or Full factorial) then, including GM values as a voxel-by-voxel covariate.
Otherwise you will have to go with a ROI approach probably.
Best,
Helmut
|