Hi, Rebecca-
An example of setting up 3dMEMA with gen_group_command.py, combining masks from individual subject analysis to make a group mask, running the 3dMEMA command, extracting estimated parameters for the "mixed ACF" for 3dClustSim simulations and using 3dClustSim is provided in the "s.nimh_group_level_02_mema.tcsh" script here:
[
afni.nimh.nih.gov]
And that associated paper provides some discussion of the rationales for various pieces. I think that addresses pretty much all of the points you asked about above, but please ping back with any it doesn't and/or further ones.
NB: that paper+associated code describes ways to correct a processing stream that another group was doing, and for comparison's sake we could correct MOST of them, but not all of them-- we try to further point out in the paper preferable steps for analysis.
--pt