Hi everyone,
We have a study with 3 patient groups (who underwent different clinical interventions) and 3 time-points and we would like to test several effects using TBSS. I have been reading the mailing list and the GLM guide, but I still have some doubts about how to create the design and contrasts with Glm in order to use them later with randomise.
First, after TBSS processing, I have created a set of FA images subtracting one time point minus each other for each subject: T3-T2, T3-T1, T2-T1. Then to calculate the interaction between any pair of times and Group, I have created the following design and contrasts (assuming I only have 2 subjects per group and that we will later use the T3-T1 FA images):
Design
Group A Group B Group C
1 0 0
1 0 0
0 1 0
0 1 0
0 0 1
0 0 1
Contrast
1 -1 0 --->This would calculate: Group A (T3-T1) > Group B (T3-T1)
1 0 -1 --->This would calculate: Group A (T3-T1) > Group C (T3-T1)
0 1 -1 --->This would calculate: Group B (T3-T1) > Group B (T3-T1)
Is this correct so far?
Reading the GLM guide (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM ; in 1 level 4 factors) I think that if I select contrasts number 2 and 3 to do an F-test, I will get a contrast that will test for any significant effect in any of my interactions, i.e., an overall F-test. Is this right?
Then if I want to check for the effect of time in each group, i would specify these contrasts:
1 0 0----> Significant areas in which FA has increased in T3 from T1 in Group A
0 1 0----> Significant areas in which FA has increased in T3 from T1 in Group B
0 0 1----> Significant areas in which FA has increased in T3 from T1 in Group C
And if I want to see the overall effect of time (for T3-T1) in all groups then I would use these three contrasts in an F-Test. Can I do this?
Does this make any sense?
Many thanks in advance.
|