> At the group level, results based on the beta from the 3dREMLfit differed a lot from the
> analysis used the beta from the 3dDeconvolve.
That sounds very unusual. Because of the unbiasedness property (whether you model the temporal correlation in the residuals or not at the individual level), the difference at the group level between the two approaches should be negligible if you only take effect estimates for group level. As a first step, could you provide the scripts for one subject as well as the scripts for group analysis?
Gang
Edited 1 time(s). Last edit at 08/28/2014 09:51AM by Gang.