Hello experts.
I'm having trouble using Glm to create a design matrix to compare 3 groups of FA scans with TBSS.
I've followed the TBSS practical and have had success doing 2 group comparisons. In these cases my understanding is that the tfce_corrp_tstat1 file gives the (corrected) 1-p values for the case where Group 1 > Group 2, while tfce_corrp_tstat2 does the same for Group 2 > Group 1. I used the ttest design script provided with FSL to setup my design.mat and design.con files.
I'm now trying to use Glm to design a 3 group ANOVA. Right now my design.mat file is (given 4 subjects in each group):
1 0 0
1 0 0
1 0 0
1 0 0
0 1 0
0 1 0
0 1 0
0 1 0
0 0 1
0 0 1
0 0 1
0 0 1
and my design.con file is:
1 -1 0
1 0 -1
0 1 -1
I followed the advice in this thread to choose those matrices
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0802&L=FSL&D=0&1=FSL&9=A&I=-3&J=on&K=2&d=No+Match%3BMatch%3BMatches&z=4&P=423550
but like the previous user, Glm prevented me from using a single F-test to incorporate all three contrasts ["Warning: At least one EV is (close to) a linear combination of the others. You probably need to alter your design"].
If I run this in randomise without an F contrast file I get 3 tstat files which I presume are the results of 3 pairwise t-tests.
My real question is this: Am I approaching this in the right way? And if so, how should I go about making an F contrasts file?
Any advice or direction towards documentation is greatly appreciated. I've read the FEAT5 documentation regarding Glm and it has been helpful but hasn't answered all my questions.
Cheers,
Mike Jarrett
|