Dear Indira,
GLM Flex has some option for group models to take into account voxels that contain values for x % of the subjects. Personally, I don't really like this option though, as it is another option to trick with the data. Certainly some people are going to set the threshold to just a few subjects. Now use a larger smoothing kernel, normalise onto the EPI template (which is going to warp the covered regions into the uncovered ones), change the intensity threshold, then it looks as if you've covered the whole brain, although you didn't do so in the majority of subjects maybe.
So if you're really interested in that regions, you should take care during slice positioning at the scanner and/or remove the affected subjects. I guess there are various studies out there claiming they didn't find any effects for region x just because it was not part of their group model... (that's why Poldrack et al. (2008) suggested to report the group mask). If you're not really interested in that area (and assuming only a smaller section of the brain is affected) you could also go with all the subjects, and report it appropriately (something like "the top-most x mm were not covered").
I would also make sure whether the top-most sections haven't been convered in your raw data or whether they are still present there. Sometimes the coregistration and/or normalisation fails, then the preprocessed data looks as if some part of the brain is missing.
Best,
Helmut
|