Hello list,
I've been following the discussion on conjunction analyses for a while, but here is something I am stuck with.
We conducted a 3x2-factorial design with six conditions A1, A2, B1, B2, C1 and C2. Our hypothesis is that there are common activations in a certain anatomical ROI due to A2>A1 AND B2>B1, but NOT C2>C1. If we constructed a linear contrast with
A1 A2 B1 B2 C1 C2
-1 1 -1 1 0 0
this would simply pick up voxels responsive to any of -A1, -B1, A2 or B2, right? Plotting betas to show the desired pattern would probably qualify as "circular analysis", wouldn't it?
So, is there any way of showing (-A1 +A2) AND (-B1 +B2) NOT (-C1 +C2)? Right now, I would go for a "conjunction overlay" approach by creating a mask from (-A1 +A2) and masking the other two contrasts with it, and see if anything survives within that mask. Are there any suggestions with more elegance than that?
Best regards and thanks for any help
Cornelius Werner
|