> I'm curious if the same logic holds if the within-subject conditions are not just contrasts of task stimuli (A-B) but different
> experimental conditions (say Treatment 1 vs Treatment 2; or just longitudinal: Day 1 vs Day 2) across days.

Yes it still holds.

> So if we wanted to say model the task using events A, B, C, D and then take the contrast of (A vs B) and compare it across
> within-subject conditions (Day 1 vs Day 2), it's easy to model each day separately and then bring A-B from each day to
> 3dTTEST, but how would you do it all at the first level in order to get a meaningful variability estimate for 3dMEMA?

Are you looking for the interaction effect between conditions and days? If so, you can obtain the interaction effect as

Day1( A - B ) - Day2 ( A - B )

at the subject level, and then feed the results (both effect estimates and their t-values) as input for 3dMEMA. If this is too much trouble, just use 3dttest++.

Gang



Edited 1 time(s). Last edit at 02/14/2018 05:51PM by Gang.