Hi, I think one way you can use is performing one-sample t-test (on the zFC*.img) within each group. Then create a mask (only including the voxels with significant positive correlations) for each group. And then make a union of these two masks for between-group analysis. One concern about this way is a multiple comparison issue. The union mask brought some information to the between-group analysis. For a voxel inside this mask, it's significant in one group, but could be either significant or insignificant in other group. E.g., for those voxels is significant (positive) in group 1 but not significant in group 2, it's more easily find significant higher activity in group 1 than group 2. By bringing such an information into between-group analysis, there should be a multiple comparison issue. I am not very sure how to address such an issue appropriately, probably other experts could provide more thoughts. Best, Chao-Gan On Fri, Aug 24, 2012 at 10:46 AM, KimMJ <[log in to unmask]> wrote: > Dear experts.. > > I'm analyzing seed-based functional connectivity in resting-state data > using SPM8 and REST software. > I've found that some published studies have compared only positive > correlation maps (and excluding negative correlations) between the groups, > because there are much controversy with regard to negative or > anti-correlated networks (? regressing out global signals). > > And my question is: > How can I make each zFC*.img (z-trasnformed FC map) that contain only > positive correlations ? > > Thank you in advance for your help. > P. Kim >