Hi all,
Would anyone offer some guidance on the following.
I am doing seed-based functional connectivity analysis. I have 17 subjects who were scanned under Pre and Post conditions. I wish to look at the correlation of a clinical variable, TestScore, with the functional connectivity maps to a region, say, the left hippocampus. TestScore does *not* change between Pre and Post conditions.
I understand how to set up this Feat GLM when correlating TestScore with only the Pre condition (or only the Post condition).
But I am not sure how to set this up to correlate TestScore with the *difference* between Pre and Post conditions.
I can set up a paired t-test. But I'm not sure how to add in the EV for TestScore since it does not change between Pre and Post.
I realize I could just manually subtract each subject's Pre and Post scans. And then correlate TestScore with these difference datasets. But I have a number of questions about that route, too.
If I manually subtract:
(1) Am I subtracting the cope1.nii datasets from GLM analyses that created the functional connectivity maps? That is:
fslmaths pre.feat/stats/cope1 -sub post.feat/stats/cope1 diff.feat/cope1
(2) What do I do with the varcope1.nii datasets, if anything?
(3) To feed the manually-subtracted datasets into the higher-level analysis, I'll need to use the [Inputs are 3D cope images from FEAT directories] Data option, correct?
(4) Do I lose information in the higher-level analysis by manually subtracting the datasets? I'm concerned about the role of varcope's in this manual subtraction method and how it plays into the GLM analysis.
Finally, I will do some permutation testing using randomise. I believe I will need to do this manual subtraction anyway to use randomise. Am I correct about that? Hopefully I can find some documentation on how to use randomise in this design (manually subtracted to handle the paired-ness and including a clinical EV).
In the end, I wonder:
(A) Should I manually subtract Pre and Post?
(B) Can I set up a paired t-test using a clinical variable that does not change between the Pre and Post condition?
(C) Do I lose information when I feed the manually subtracted dataset into Feat?
(D) randomise!
Just to verify, I am demeaning my clinical variable, TestScore. This is necessary, correct?
Thanks for any guidance,
* BettyAnn
|