The problem is that your variable is always0 for the HC group. This creates a regressors of all 0s when using DODS. One way around this is to use DOSS (--fsgd your.fsgd doss) and change the contrast matrix. This may not be the model you want (it will assume that the change due to Antipsychotic is the same across all patient groups. You can get around this by creating your own design matrix rather than going thru an FSGD file

On 05/20/2017 05:30 PM, Katherine Damme wrote:

Hello FreeSurfer Group

I would like to make a two group comparison controlling for both gender and medication status. The only way that I have sorted out doing this is to make eight groups (ClinicalGroup x Sex x Meds) but I get an error that my design is poorly scaled:

---

ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08

---

Possible problem with experimental design:

Check for duplicate entries and/or lack of range of

continuous variables within a class.

If you seek help with this problem, make sure to send:

1. Your command line:

mri_glmfit --y lh.lgi.mgh --fsgd FSGD_GroupSexTP_AntiPsych.txt 

--glmdir lh.lgi.glmdir --surf average lh --C contrast.txt

2. The FSGD file (if using one)

3. And the design matrix above

---

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