I fitted a model on the group level using 3dMVM/3dMEMA (I tried both ways, got almost the same results).
In the model, I have one within-subject 2-level categorical predictor which relates to the task that was performed in the scanner (words/scrambled), and three between-subject continuous variables that were measured outside the scanner (Root, Sem and Phon). I want to see how these variables are related to brain activation.
I defined the three following contrasts (in 3dMVM):
-gltlabel 1 Root_effect -gltCode 1 'Morph: 1words - 1scrambled Root : ’
-gltlabel 2 Sem_effect -gltCode 2 'Morph: 1words - 1scrambled Sem : ’
-gltlabel 3 Phon_effect -gltCode 3 'Morph: 1words - 1scrambled Phon : ’ \
Now, I want a statistical brain map that contrasts Root_effect with Sem_effect. How do I achieve that?
I want a statistical brain map that contrasts Root_effect with Sem_effect.
How do I achieve that?
That kind of comparisons is not available in 3dMVM because such a comparison is not always meaningful in general; for example, two such variables may have different metric (e.g., height and weight), therefore their comparison would be uninterpretable.
What is the underlying logic for such comparisons in your case?
I want to identify brain areas that are sensitive to the scores my subjects have on the Root variable (which is a continuous, between-subject variable). The logic is that these scores represent some cognitive capacity that might also be reflected on their brain activation during the task that they performed inside the scanner. The two other variables (Sem and Phon) are there as controls. So assuming I find a brain area whose activation during the task is modulated by the Root score, I want to show that this modulation is significantly higher than the modulation of Sem and Phon. Hence, the contrast I was thinking of. Does this make sense? All three variables are standartized, if it matters.
Such type of comparison can probably done at the region level through program “RBA”:
I’ll have to check later if this is feasible with 3dMVM.