Dear Yann,
alternatively, you could run a ROI analysis with Marsbar, that is extracting the mean beta/contrast estimate for a particular cluster (averaged across all the voxels within this ROI) or reporting percent signal change values. It is common to use a threshold of .05 for this purpose, or .05/k with k being the number of ROIs investigated.
I think in most cases these ROI analyses come closer to the hypotheses, as it tells you something about the amygdala as such. Interpretation of significant results in SVC would be that a SUBSET of voxels shows some effect. However, this is often interpreted/discussed as an effect for the whole ROI (in this case, the amygdala showing some effect), which is not true.
Best,
Helmut
|