Jim,

Please see the inline responses below.

On Wed, Aug 28, 2013 at 12:58 PM, James Lee <[log in to unmask]> wrote:

Umi,

I'm quite sure the SPM.xVol.FWHM for a group random effects result is
for the group residuals, and not an average of each individuals FWHM.

Yes. You are correct that SPM.xVol.FWHM is the result of group residual (specifically 64 of the residuals of the group analysis). It is clearly not the average of the FWHM from the residuals of the first level model.

I'm curious about the idea you read that the "right" way to do this is
to average the individuals FWHM. I may be wrong, but if the smoothness
is used to calculate a significant cluster size in the group random
effects result, it seems logical to use the residuals of the group
random effects result to calculate that smoothness, and not an average
of the individual subjects smoothness, since the random effects
analysis is not an average. I'm interested to know where you read
that.

I'm not going to take the position of the right or wrong way to do the analysis. All I will say is that SPM has one approach and AFNI recommends using the average of the first-level residuals (e.g. http://afni.nimh.nih.gov/afni/community/board/read.php?1,83561,83576#msg-83576). I use the SPM approach.

Jim

On 8/27/13, Umi V. <[log in to unmask]> wrote:

> Also just to clarify, I am considering smoothness converted to mm, not in
> voxels, so I can't attribute discrepancies to unit differences.
>