afni.nimh.nih.gov

> But doesn't it make more sense to estimate the smoothness from the residuals of the
> group analysis? That's how the smoothness is estimated from each subject.

Ideally that should be the case. However, there are two reasons for the current situation: 1) most group analysis programs currently don't directly generate the residuals; 2) there seems no well-established approach as to how the FWHM values should be estimated at the group level.

Gang