posting the response to the list: The problem is that femalezero and
femaletwo only have 4 and 3 subjects, respectively. As you are trying to
estimate 4 slopes + an intercept, you will need at least 5 subjects per
group.
doug
On 08/16/2017 03:27 PM, Sullivan, Danielle Renee wrote:
>
> Hi Doug,
>
>
> I am just following up to this email. If you prefer I can repost on
> the FreeSurfer mail archive.
>
>
> Thanks,
>
> Danielle R. Sullivan (Miller), Ph.D.
>
> National Center for PTSD
>
> VA Boston Healthcare System Jamaica Plain
>
> Boston University School of Medicine
>
> OFFICE: (857) 364-4022
>
>
>
> *From:* Sullivan, Danielle Renee
> *Sent:* Friday, August 11, 2017 1:33:03 PM
> *To:* Douglas N Greve
> *Subject:* Re: [Freesurfer] Matrix ill-conditioned error for dods
> thickness mri_glmfit analysis
>
> I am still getting the same error. My apologies for the back and forth.
>
>
> Here are updated files (now there are 5 classes and contrast mtx has
> been changed to reflect that).
>
>
> Danielle R. Sullivan (Miller), Ph.D.
>
> National Center for PTSD
>
> VA Boston Healthcare System Jamaica Plain
>
> Boston University School of Medicine
>
> OFFICE: (857) 364-4022
>
>
>
> *From:* Douglas N Greve <gr...@nmr.mgh.harvard.edu>
> *Sent:* Friday, August 11, 2017 1:10:53 PM
> *To:* Sullivan, Danielle Renee
> *Subject:* Re: [Freesurfer] Matrix ill-conditioned error for dods
> thickness mri_glmfit analysis
> remove the definition too
>
>
> On 08/11/2017 12:46 PM, Sullivan, Danielle Renee wrote:
> >
> > Hi,
> >
> >
> > Thank you for the tips. I tried doing both and still run into the
> > error (although now I get a warning that the class femaleone is
> > defined but not used). Attached is my new FSGD (now all variables
> > normalized via a Z-score), XG.mtx, Contrast, and Terminal output.
> >
> >
> > Thanks again for the help!
> >
> >
> > Danielle R. Sullivan (Miller), Ph.D.
> >
> > National Center for PTSD
> >
> > VA Boston Healthcare System Jamaica Plain
> >
> > Boston University School of Medicine
> >
> > OFFICE: (857) 364-4022
> >
> >
> >
> > *From:* Douglas N Greve <gr...@nmr.mgh.harvard.edu>
> > *Sent:* Friday, August 11, 2017 11:12:21 AM
> > *To:* Sullivan, Danielle Renee
> > *Subject:* Re: [Freesurfer] Matrix ill-conditioned error for dods
> > thickness mri_glmfit analysis
> > there appear to be two problems. One is that you only have one instance
> > of femaleone. This might be ok if you had no covariates, but you can't
> > estimate a covariate slope from one data point. Second, the scale of
> > your covariates is much different, so you may need to normalize them. I
> > would try removing the lone femaleone and rerun
> >
> >
> > On 08/11/2017 10:35 AM, Sullivan, Danielle Renee wrote:
> > >
> > > Hi Doug,
> > >
> > >
> > > Thanks for your response and help.
> > >
> > >
> > > Attached is the matrix and the terminal output. Please let me know if
> > > you need anything else.
> > >
> > >
> > >
> > > Danielle R. Sullivan (Miller), Ph.D.
> > >
> > > National Center for PTSD
> > >
> > > VA Boston Healthcare System Jamaica Plain
> > >
> > > Boston University School of Medicine
> > >
> > > OFFICE: (857) 364-4022
> > >
> > >
> > > can you send the design matrix Xg.mtx and the full terminal output?
> > > On 08/08/2017 09:13 AM, Sullivan, Danielle Renee wrote:
> > > >
> > > > Hi FreeSurfer Experts,
> > > >
> > > >
> > > > I am trying to run a dods cortical thickness analysis examining the
> > > > interaction of group on a continuous variable while controlling for
> > > > other variables.
> > > >
> > > >
> > > > Specifically, I have 6 classes (gender by three groups [zero, one,
> > > > two]). I am interested in the group interaction with a continuous
> > > > variable while controlling for gender (among 4 other continuous
> > > > variables). I set up my fsgd file (see attached) and contrast file
> > (36
> > > > dimensions- F test):
> > > >
> >