first of all, you should set up gender as a class factor (ie, you'd have 4 classes, con-male, con-female, etc). If you want to test for an intercept difference between Con and Scz regressing out the effects of age and age^2, you should first verify that there is not an interaction between diagnosis and age. Do this by testing for the interaction (using DODS); if there is no interaction, then run with DOSS. In neither case is it necessary to demean (though you may need to demean and normalize just to make the scaling in the design matrix better). If there is an interaction, then you'll need to demean and then be very careful how you interpret your data.

On 1/5/18 3:32 PM, Hossein Rejali wrote:

Hello Freesurfer experts,

I am doing a whole brain analysis using *mri_glmfit *to determine if there is a significant difference between two groups (controls-*con* VS Schizophrenic -*Scz*). I have set up the following fsgd file:

What I am wondering is if I should have a demeaning flag, there was not much documentation as to when or when not to do this but I was hoping I can get an answer here.

Also below is the contrast vector that I am using:

A follow-up question would be whether to use DODS or DOSS I am still confused which model to use? The independent variable is Myelin.

Kind Regards,

Hossein Rejali

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer