Hello FSL experts,
I have a question about additional regressors for VBM and TBSS processing. If I have two group of participants, one group are the patients the other controls and I would like to correlate FA or GM with a behavioral measure would my designed matrix look like the following:
0 demeaned_behavioral_regressor
0 demeaned_behavioral_regressor
0 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
or
1 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
1 demeaned_behavioral_regressor
There seem to be different approaches and different solutions provided on the forum. In one post I read that if the behavioral correlate is believe to the caused by group identity (i.e. patients' disease duration scores), then design matrix should look like the last example. Is there a general rule of thumb how these correlations should be performed? Also, if the regressor is already demeaned when entered into the design matrix, should the randomise command be run with the -D flag? Some of the previous posts suggest that -D should be used, while other do not.
Thank you for your help in clarifying this.
Sincerely,
Anastasia
|