Here is a data analysis scenario and a question that follows regarding the use of 3dFWHM to estimate smoothness for use with AlphaSim:

Assume that you have scanned several subjects with a random-order block-design stimulus with two or more different stimulus conditions. You analyze each subject's data using 3dDeconvolve, generating means and t-stats for a glt for the area under the curve for each stimulus condition. You interpolate the means for each condition for each subject into talairach space and then use a 4mm FWHM blurring kernel. Then you do group analysis by running 3dANOVA2 to generate means and t-stats for each condition. You apply some threshold to the t-stats and use that as a mask to view the means for each condition in suprathreshold voxels.

Now let's say that you want to correct for multiple comparisons by excluding clusters that are smaller than some size limit. You find that limit by running AlphaSim, which simulates a noise dataset (i.e. one with no true activations). You need to supply AlphaSim with the the key parameter of FWHM smoothness, which can be estimated with 3dFWHM from the t-stats generated by 3dANOVA2.

The question is, Which, if any, of the conditions' t-stats should be used to estimate FWHM smoothness? If one of the conditions has widespread areas of activation, its FWHM measures will be substantially higher than a condition with very little activation. Because AlphaSim is supposed to be simulating noise in the absence of true activation, I would think that we should use the same cluster size limit for each condition and not penalize conditions with greater true activation.

Is there a way for 3dANOVA2 (or 3dANOVA) to output some kind of "residual error" that could be used as input to 3dFWHM? Or is there some output of 3dDeconvolve (maybe errts?) that could be interporlated to talairach, smoothed, and then averaged across subjects? Or maybe it would be best to generate a noise dataset for each subject, possibly smooth to match the average smoothness from the original raw data, interpolate to talairach, smooth, and then average?