Dear Lisette,
The model does not allow to properly test for main effect group due to error term issues (see previous posts by Donald McLaren). I'd suggest to go with the simplest approach, One-way ANOVAs. You'll need two contrasts on single-subject level, for contrast 1, you average across conditions [0.5 0.5], for contrast 2, you take the difference between the two [1 -1].
Within the first One-way ANOVA based on the con_0001 images
- the F contrast [1/4 1/4 1/4 1/4] reflects average activations across groups and conditions (could be useful depending on study)
- the F contrast [1 -1 0 0; 0 1 -1 0; 0 0 1 -1] reflects main effect group
Within the second One-way ANOVA based on the con_0002 images
- the F contrast [1/4 1/4 1/4 1/4] reflects main effect condition
- the F contrast [1 -1 0 0; 0 1 -1 0; 0 0 1 -1] reflects the group x condition interaction
If necessary, you can add the covariates.
The approach should correspond to a "proper" 4x2 ANOVA with partioned error terms.
Best
Helmut
|