I have worked on this 1000 different ways and have been unable to find out the problem.
Here is a very simple example of what I encounter.
1. I get a significant activation at the group level A > B.
2. I create a mask with this contrast intersecting in my ROI's.
3. I take this mask for each ROI separately and apply it to A > fixation, A < fixation, B > fixation, B < fixation.
4. I then pull all the beta weights featquery calculates.
5. It doesn't matter which values I use (minz, mean, tstat, etc.....) I got bar graphs where the beta's for A & B have overlapping error bars and look very similar.
I am attempting to find out the magnitude of deactivation or activation. It doesn't make sense that if I get a significant group level contrast of A > B that I would then get non-significant BETAs.
Edward H. Patzelt
University of Minnesota