Hi Mel,

I forget, did you not use afni_proc.py for the analysis? The
blur averages for each subject should have been computed then.

Note that afni_proc.py runs 3dFWHMx for each run separately.
Otherwise, the detrending polynomial degree becomes very high
and might lead to less accurate blur estimates.

In 3dClustSim, you do not need to provide -nxyz or -dxyz, since
the -mask dataset provides such information. But blur levels of
10..11 seem okay, assuming you added around 8 mm in your analysis.

What sort of cluster levels are you getting from this?

Note that this has nothing to do with FDR. To be specific, you
should probably use either cluster/thresholding or FDR, but not
both (though some people have tried to apply a sort of FDR to
the clusters, but that is a different story).

The only question about your mask is whether it provides an
appropriate level of coverage across your subjects. It is up
to you to choose a mask that you are happy with. Bigger masks
show more results, but will require bigger clusters.

- rick