Could a conjunction end up taking away some otherwise significant differential contrast areas though? I'm not sure if this is possible, but what if some regions fall just short of significance in the [1 0] or [0 1] contrast(s), yet appear to be significantly different in their level of activation when compared to the activation of the other group?
Just in case of this, I thought of doing the following steps in my similar analysis of group differences in activation and deactivation. I'd love to get feedback on whether or not it seems like a valid approach.
I, like Andres, also have 2 groups of participants, for which I ran contrasts:
c1: group1-activation [ 1 0 ]
c2: group1-deactivation [-1 0 ]
c3: group2-activation [ 0 1 ]
c4: group2-deactivation [ 0 -1 ]
c5: group1 > group2 [ 1 -1 ]
c6: group2 > group1 [-1 1 ]
Then I created the following masks to reflect each groups' areas of respective activation vs. deactivation:
fslmaths c1tstat -bin group1act
fslmaths c2tstat -bin group1deact
fslmaths c3tstat -bin group2act
fslmaths c4tstat -bin group2deact
After thresholding the c5 differential contrast (I used TFCE + FDR) to get areas of significant group1>2 difference, I used the following to interpret the result:
fslmaths c5sigmap -mul group1act group1greateract
fslmaths c5sigmap -mul group1deact group1lessdeact
...and similarly for the c6 differential contrast result:
fslmaths c6sigmap -mul group2act group2greateract
fslmaths c6sigmap -mul group2deact group2lessdeact
So, the separated activation/deactivation results are:
Activation = group1greateract; group2greateract
Deactivation = group1lessdeact; group2lessdeact
Anyone have any thoughts on this? Does it seem like a reasonable approach?
I think Andres and I are rowing in the same boat.
Leslie
----------------------------------------------
Leslie A. Aderhold, M.S.
Graduate Student
Department of Psychology
The Pennsylvania State University
|