I’ve just finished running fmriprep (
version 1.1.7) on a few subjects, and am now starting the first level GLM’s each for. When running fmriprep, I did not specify
--use-aroma, so I spatially smoothed the *preproc_nii files using a FWHM of 6mm, and then performed intensity normalization on the smoothed data.
I’m planning on running the GLMs using AFNI’s
3dDeconvolve but I’m not entirely sure what all confounds I should include in the regression. This thread suggested using 6 motion parameters, FD, and aCompCor (00-05); however, would it also be wise to regress out CSF, WM, and global signal?