Hi AFNI experts,

I write with a question regarding using 3dFWHMx to assess smoothness of data processed in FSL. I intend to use 3dClustSim with the average group-level smoothness information.

Typically when I use 3dFWHMx I run it on the preprocessed (including motion correct, registered/warped to a template) time-series data with the -detrend option, or without the detrend option on GLM residuals. In both cases, this data has been already warped to the template space where final statistics are computed. In the current FSL case, I have clean functional timecourses in native space and statistics were computed for each run and warped to the template at the second level. My question is whether it makes sense to use 3dFWHMx on the clean functional data in native space using -dtrend and an automask option, or whether it would make sense to warp the clean functional data to the template and estimate smoothness in this space. Furthermore, I do have model residuals in native space that I could use instead of the clean functional data (so I wouldn't use the -detrend option). Basically I am not sure if I should estimate smoothing in native space because that is where subject specific statistics maps were computed (which were subsequently warped to template space), or if I should estimate smoothness in group space because that is where the final statistics are computed (and warping to the template likely induces additional smoothing).

What might be the best approach to estimate smoothness at the group level so I can use 3dClustSim? I see four options, 1) Compute smoothness in native space on the 1) clean data or 2) residual data, or warp the 3) clean or 4) residual data to template space and estimate smoothness there.

Many thanks in advance, regards,

Matthew