Hello,

I have a question regarding multiple comparisons at the group level. A reviewer is concerned about the way I implemented the analysis and seem to think the cluster size should be much larger than what we have.
In my experiment we have 14 subjects, whole brain 3x3x3 voxel size smoothed at 6x6x6, and the brain mask contains 60,213 voxels.
I used a p = 0.005 at voxel level and p = 0.05 at cluster level (using alphasim), which yielded cluster size of at least 25 voxels.

The reviewer says it is way too small and indicates that family-wise error corrections at cluster thresholds of <0.001 are usually in the range of 100 to 200 voxels. This seems extremely wrong!!!

I wanted to know what the afni community though about the way I did the analysis.

Thank you in advance

pat