Dear David,
Instead of
2 -1 -1 | Control >mean(Patients1,Patients2)
-2 1 1 | Control < mean(Patients1,Patients2)
it might be more useful to run a conjunction across the first two contrasts and another across the third and forth one. Basically, if one patient group differs to a large extent from the controls, then you might detect group differences between controls and the patients as a whole even if activation level / GM density, volume / whatever in the second patient group corresponds to that in controls. But this is just a suggestion.
Best
Helmut
|