Hi Alessandra,
The FWHM value is the full width at half max measure of the
estimated smoothness in your time series data. It does not
really depend on the size of a voxel.
You can run "3dFWHMx -detrend" on the input to 3dDeconvolve
to estimate those parameters (or preferably, run it on the
-errts output from 3dDeconvolve) for any given subject.
To get FWHM numbers for a group analysis, it is okay to take
the average values from all of your subjects (which should be
basically the same to begin with).
---
Yet another option is to use 3dBlurToFWHM on the time series
data, instead of 3dmerge say. Using 3dmerge, one can apply a
6mm FWHM blur to the data, for example. But that is on top of
whatever blur (smoothness) happens to be there already (right
out of the scanner).
With 3dBlurToFWHM, you don't _add_ a blur, you _set_ the blur.
So you could blur each subject's data so that the resulting
smoothness is 6x6x6, and then apply those numbers for every
AlphaSim computation.
- rick