Dear Donald,
as far as I remember masking does not change the statistics / no recalculation of resel counts, p values, ... Thus, if the cluster is located at the border of the mask, it might be smaller in terms of voxel size after masking and smaller than the critical extent threshold (based on whole-brain), but as the cluster p still corresponds to whole-brain statistics it is reported as significant.
This is different with SVC, which changes the cluster statistics and the corrected peak p-values. However, it seems not all of the information at the bottom of the table are adjusted appropriately. To give an example:
Whole brain: 167260 voxels = 3313.4 resels (resel = 46.66 voxels)
After SVC: 1435 voxels = 9.9 resels (resel = 46.66 voxels)
1435 / 9.9 is 144.95 though, not 46.66.
Best,
Helmut
|