For a functional MRI analysis, I usually create a mask of the significant clusters and then extract the mean of the voxels within that mask for each subject. However, I want to make sure that the way I calculate the standard deviation for a group is acceptable.

Let's say for instance that there are 10 subjects. I would calculate the standard deviation using the 10 subject's extracted mean cluster value. Is this appropriate? Since the analysis is being done on a voxelwise basis and each voxel passes a threshold of p>.001 would it make more sense (or be more appropriate) to take the value of each individual voxel and calculate the cluster's standard deviation in some fashion?

Thanks,
-Matt