Dear FSL experts,
I am a bit confused with the design matrix, contrast and demeaning in a TBSS FA correlative analysis and I would be so grateful if anybody could check my approach.
Issue: 1 group of patients, COV1 continuous, COV2 continuous, FACTOR1 gender (male, female).
My design matrix would be : according to http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM / "Even if you add a categorical covariate like Gender, simply add a covariate that is 1 for one gender, 0 for the other and mean center this (that is, it is treated the same way as for a continuous covariate)"
EV1=demeanCOV1 EV2=demeanCOV2 EV3=demeanFACTOR1
Since I am not interested in FA differences between Male and Female, I would not setup FACTOR1 as additional EVs (one for males, one for females).
Since I omit an additional EV for GROUP mean (all ones), I would use the -D flag in randomise.
My contrast would be [ 0 1 0 ] to look for positive correlations between FA and COV2, regressing out the effect of COV1 and GENDER (factor1).
Would this analysis be correct?
Thanks for your help,
best wishes,
Robert
|