Dear all,
I am interested in testing wheter the linear relationship between the dependent variable (skeleton FA image) and a covariate (for example, age) differs between two groups of subjects. For that, I have run TBSS analysis with -S option in tbss_3_postreg script (derive FA mean and skeleton from the actual subjects) and I use randomise for statistics. In order to define the contrast and matrix files, I follow the advice from 'Two group with continuous covariate interaction' at FSLWIKI GLM.
Then, I would like to test each group separately. For each group of subjects, I would like to test the covariate effect on skeleton FA image.
My question is the following: Should I use the same output from TBSS to feed into randomise? i.e. I split the desired skeleton FA images corresponding to a single group of subjects and I do the test with it. And I repeat with subjects of group 2 skeleton FA images. If I did it, I would be using the projection of FA images onto a skeleton which averages all subject of group 1 and group 2. Otherwise, I should do a new TBSS analysis with subjects of group 1 and another with subjects of group 2. What option should I choose?
Thanks you in advance.
M del Mar
|