Thanks again Donald for taking the time to reply to my questions. If you (or someone else) could help again with some other uncertainties that would be great!
1) I tried to mask those interaction voxels in the analysis, but there's no option in feat to supply a mask (I think that's different to SPM), so I was wondering whether you meant to simply mask the final zstats images as opposed to run the analysis again excluding those voxels. Alternatively there's probably a way to manually change a fsl script so that it takes a predefined mask, or actually remove the voxels from the 1st level copes. which (both supplying a predefined mask or removing the voxels the hard way) I feel seem the more appropriate ways of dealing with this because of less voxels that go into the analysis ...
2) A more general question but quite important for my understanding of the whole issue. I fairly often read people reporting that they adjusted for age, IQ whatever looking only at group effects... However that's where it usually ends, no further testing for interactions. From my understanding now, this seems wrong though as an interaction in the same regions would make an interpretation of group effects difficult if not impossible. So I guess what I want to ask is, as soon as you add a covariate, you have to do the full monty (1st model for main effect of covariate including group effects, then 2nd model testing for interactions) and not just looking at the 1st to check for group effects.
Thanks again for your thoughts on this,
Best wishes,
Torsten
|