Hi Gang,

I’ve been doing some thinking/reading on your reply and have the following follow-up questions:

Since I have different numbers of trials for condition 1 and condition 2, how would I obtain the beta differences? Seems like simple subtraction would be a problem, since they would actually have different variances…when the value was carried forward I wouldn't be able to carry forward the variances…what are your thoughts on this? Perhaps I should actually run an analysis for each TENT on it’s own with one within subjects factor (condition) and one between subjects factor (group)? This seems problematic too…At any rate, will 3dMVM take the differences between the variances/degrees of freedom into account (between the two conditions) when it does the calculations; will it allow me to carry forward the information from the Tmaps from the Rbuck files to the group analysis? If not, am I correct in thinking that 3dMEMA will?

Second, I thought further, and my actual main test is 2X2X2 ANOVA, with two scans (pre medication and post medication), two groups (medication versus placebo) and two conditions during each scan (with the asymmetries in trial numbers as described previously) – my main hypothesis is that:
The differences in activation between condition 1 and condition 2 will diminish more in the medication group from scan 1 to scan 2 relative to the differences between the two conditions in the placebo group (which will stay the same).
Group – 2 levels, between subject, voxelwise covariate (ASL) to correct for changes in blood flow from the medication
Condition – 2 levels, within subject, nested within scan
Time (scan 1 vs scan 2) – 2 levels, within subject
(Plus we’ve got all the betas from using multiple tent functions…)

Sorry for being incomplete in my previous post! The analysis will be even more complex than I made it sound.