Hi Steve,
Yes, the F.A.Q. 50 explanation and approach still holds. Combining the appropriate cope images from first and second level analysis worked just fine.
A related question is whether I can examine in a single higher level analysis group mean activation (from subjects with single or multiple runs) for several EVs? I ask this because when I examined mean activation for a single EV modeled from the relevant cope images without also including other EVs that were part of the lower level analyses, I saw different effects than if that EV was evaluated in the context of the other EVs (and their corresponding copes). I think it is best to include all EVs and copes that were part of the lower level, correct?
For example if I have 10 subjects, 5 with a single run and 5 with two runs and 2 EVs, by entering the appropriate cope images, could model mean activation for EV1 and then EV2. Model set-up would look like something as follows:
Group EV1 EV2
Input 1 (subject 1 cope1) 1 1 0
Input 2 (subject 2 cope1) 1 1 0
.
.
Input 10 (subject 10 cope1) 1 1 0
Input 11 (subject 1 cope2) 1 0 1
Input 12 (subject 2 cope2) 1 0 1
.
.
Input 20 (subject 10 cope2) 1 0 1
Contrasts coded below would then provide mean activation for each EV:
EV1 EV2
C1 Group Mean EV1 1 0
C2 Group Mean EV2 0 1
I believe this would be the case. And should I be interested in comparing groups (vs. looking at a single group), I could simply add an additonal EV coding for membership and create contrasts of interest?
I think that this is straight forward at this point, but I thought I'd check.
Thanks again for the helpful response.
Jim
|