Hi Woogul,
If your mask and data are at 2x2x2, use 2 mm as your radius
(which would make your cluster volume 2x2x2x20).
However a 5mm fwhm blur seems too small, given that it is less
than one of your original voxel dimensions. Did you estimate
the blur that is in your data using 3dFWHMx on the errts output
from 3dDeconvolve (as is done by afni_proc.py)?
How are you estimating your fwhm blur? Did you apply a 5 mm
fwhm blur with 3dmerge? If so, the actual estimate is probably
higher...
- rick