Hi George,
In the end, the blur estimates are probably not too
far off of what they would be had the analysis been
done in standard space to begin with. But they are
probably underestimated a little.
Note that it is easy to perform the regression in
standard space, which would make such concerns moot.
Also, it should not be necessary to oversample data
when going to standard space. One can resample an
atlas to match the resolution of the group analysis.
Consider running the regression analysis in standard
space, and at a resolution that is close to that of
the original data. By default, afni_proc.py uses the
smallest voxels size, isotropic.
Yes, those 3dFWHMx and 3dClustsim values are in mm,
not voxels.
- rick