The second design is fine if there is no significant interaction between 1 or 2 and 3. You'll have to create the design matrix yourself or you can create an fsgd file with 7 classes

On 02/02/2017 11:38 AM, Tim Reess wrote:

Dear Freesurfer experts,

I would like to set up a design matrix for a design that is not QDEC-able. I have three categorical factors:

1.disease status: 2 levels

2.gender: 2 levels

3.scan site: 4 levels

The first design I set up was the following (the rows are just for illustration and don’t reflect the number of subjects that will enter the analysis):

designs4_fs_mailinglist.png

However, I have my doubts as to the “efficiency” of the design (in terms of number of regressors / DOF). Therefore, here is my second design:

Would this design be correct for assessing the group differences for disease status while regressing out gender and scanner site? My understanding is, that in this design I can’t assess the effects of interactions regarding scanner site. However, I think it is fair to assume that the scanner site doesn’t have a differential effect on either gender or disease status. Any feedback is highly appreciated. Thanks a lot. Best regards,

Tim

---

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