Hi AFNI folks,

We came across some puzzling results when using 3dFWHMx on our 3dDeconvolve errs files. Namely, each subject's estimated spatial smoothness is around 2.0, and we applied a blur of 4.0 on these data in preprocessing. This seems peculiar, as I understand that the estimated values are generally larger than the blur that was applied to the data.

We tried with -automask and without. The results are the same. When we use -automask, the screen output mentions that there are only 7000-odd voxels in the mask, which seems like a red flag because using 3dinfo we should have about 60000-odd voxels to work with.

I was wondering if there is some step in preprocessing like warping or masking (before 3dDeconvolve) that would cause the errs output to misbehave like this? Is this the sort of problem the -detrend option in 3dFWHMx would fix?

Any help is appreciated.

Thank you,
Ana