Dear FSLers,
I'm hoping to run TBSS on diffusion data collected over three visits (0, 6 and 15 months) on a sample of 85 subjects (45 patients, 40 controls). I'm a little stumped as how to set up the appropriate design matrix and contrasts for Randomise and would greatly appreciate any guidance.
From what I understand about Randomise and having looked through the FSL archive, there isn't a formal way to model change within subjects about 3+ timepoints, in order to compare this change between groups. My attempt at a design matrix (simplified to 3 controls, 3 patients) got as far as this:
con-time1 con-time2 con-time3 pat-time1 pat-time2 pat-time3 con1 con2 con3 pat1 pat2 pat3
1 0 0 0 0 0 1 0 0 0 0 0
1 0 0 0 0 0 0 1 0 0 0 0
1 0 0 0 0 0 0 0 1 0 0 0
0 1 0 0 0 0 1 0 0 0 0 0
0 1 0 0 0 0 0 1 0 0 0 0
0 1 0 0 0 0 0 0 1 0 0 0
0 0 1 0 0 0 1 0 0 0 0 0
0 0 1 0 0 0 0 1 0 0 0 0
0 0 1 0 0 0 0 0 1 0 0 0
0 0 0 1 0 0 0 0 0 1 0 0
0 0 0 1 0 0 0 0 0 0 1 0
0 0 0 1 0 0 0 0 0 0 0 1
0 0 0 0 1 0 0 0 0 1 0 0
0 0 0 0 1 0 0 0 0 0 1 0
0 0 0 0 1 0 0 0 0 0 0 1
0 0 0 0 0 1 0 0 0 1 0 0
0 0 0 0 0 1 0 0 0 0 1 0
0 0 0 0 0 1 0 0 0 0 0 1
Does this look correct? As for the contrasts, pairwise comparisons are feasible (i.e. visit 1 > visit 2, visit 1 > visit 3 etc), but want I really would like it to model change across all three timepoints and look for group differences in patterns of change. How would this be possible using Randomise?
Many thanks,
James
James Cole PhD
Huntington's Disease Research Group
UCL Institute of Neurology
|