Lisa,
> I ran a RegAna where I had 4 predictors: group (2 levels), stimulus intensity
> (2 levels), age (continuous), and agexstimulus intensity (with the model 4 3 2
> 1: 0)... I ran a RegAna separately for each group with stimulus intensity and
> age as the predictors and the model statement 2 1: 0.
It's pretty tricky to code a within-subject factor such as stimulus intensity in 3dRegAna. A simple column with 0s and 1s is not good enough to account for across-subjects variability. So I would stay away from 3dRegAna when you have a within-subject factor.
I believe 3dLME is what you should try:
[
afni.nimh.nih.gov]
Gang