Hi Omar,
let's start from scratch. Say you've got n subjects in total and x contrasts.
You've got your design matrix design.mat which is right to start with.
Then for your contrast design.con (x contrasts=rows, 3*2+n EVs=columns), you need:
(EVs: CON1 CON2 ABS1 ABS2 REL1 REL2 subj1 subj2 sub3...)
-1 1 1 -1 0 0 0 0 0...
-1 1 0 0 1 -1 0 0 0...
0 0 -1 1 1 -1 0 0 0...
-1 1 0 0 0 0 0 0 0...
0 0 -1 1 0 0 0 0 0...
0 0 0 0 -1 1 0 0 0...
First row = where the changes between controls (increase of FA) are significantly different from the changes in abstinent patients
Second row = where the changes between controls (increase of FA) are significantly different from the changes in relapsing patients
etc.
Fourth row = where the changes between the two time points are significant in the controls
etc.
What you also want is to answer this question: "overall is there any significant difference between the changes across groups".
So you'll need an F-test, with a design.fts (with 1 row, x column):
1 1 0 0 0 0
That tells the model to do the F-test for the first two contrasts of the design.con (third one is redundant).
So with all that, you'll be able to answer to:
* "overall is there any significant difference between the changes across groups", that's the F-test
* "is there any significant difference between the changes in CON/ABS (increase of FA) and in ABS/REL", these are the first 3 contrasts of your model
and finally:
* "is there any significant increase between the two time points in CON/ABS/REL", these are the last 3 contrasts of your model (post-hoc paired t-tests)
In practice in the Glm gui: you need x contrasts and one F test and you click on the first 2 boxes of the F-test.
and randomise, something like:
randomise -i all_FA_skeletonised -m mean_FA_skeleton_mask -o res_FA -d design.mat -t design.con -f design.fts -F 4 -c 2 --T2 -n 1000
Hope this helps,
Gwenaelle
