3) If you are going to run group analysis on the stats, then your
method should be basically fine. Note that there would be no
reason to warp to tlrc space in align_epi_anat.py, since you
do not use that tlrc dataset. The adwarp step can be done on
the betas after 3dDeconvolve.
Keep in mind that if you use 3dREMLfit to correct the t-stats
for temporal correlation, with the intention of using 3dMEMA.R
for group analysis, then it would be better to proceed with the
tlrc datasets output from align_epi_anat.py.
- rick