Run mri_glmfit on both groups separately, using a contrast to compute the mean
of the group. This will create two directories, one for each group. In each
directory, there will be a variance measure (rvar.mgh) and a mean of the
contrast (contrast/gamma.mgh). You then need to compute the welch's t as
t = (gamma1-gamma2)/sqrt(rvar1/N1 + rvar2/N2)
where n1 and N2 are the sample sizes for each group. You can implement this in
matlab or from the command line. If matlab, you run
rvar1 = MRIread('group1/rvar.mgh');
gamma1 = MRIread('group1/contrast/gamma.mgh');
rvar2 = MRIread('group2/rvar.mgh');
gamma2 = MRIread('group2/contrast/gamma.mgh');
t = gamma;
t.vol = (gamma1.vol-gamma2.vol)./sqrt(rvar1.vol/N1 + rvar2.vol/N2);
MRIwrite(t,'welchst.mgh')
On 3/15/19 3:24 PM, Hannah CK wrote:
External Email - Use Caution
Hi Doug,
We actually do want the voxel-by-voxel results. Could you please give details
on how we would do that? Specifically, would we need to use the command line
to run the Welch's T/F or would we extract values for each voxel and put them
into another software to conduct the Welch's T/F?
Thank you again.
Greve, Douglas
N.,Ph.D.<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=from:%22Greve%2C+Douglas+N.%2CPh.D.%22>
Thu, 14 Mar 2019 08:45:56
-0700<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=date:20190314>
You can use mri_segstats to get whole-cortex means (use --slabel and specify
the lh.cortex.label and add --id 1). You can do the same thing with the
variances. The problem with doing it after mri_glmfit is that the variances
will not reflect the spatial averaging over cortex. If you really just want one
number for cortex, you should use mri_segstats to extract the mean values from
the stack (ie, the --y input), using --avgwf. This will give you a column of
numbers, one mean for each subject. Then you can process this however you want.
When I first responded, I thought that you want to do this on a voxel-wise
basis.
doug
On Thu, Mar 14, 2019 at 10:10 AM Hannah CK
<hckal...@gmail.com<mailto:hckal...@gmail.com>> wrote:
Hi Dr. Greve,
Sorry about that - I had followed the "Reply via email" link on the archive
page instead of posting to the forum directly.
Thank you for this response. Can extracting the means be done at the
whole-brain level? If we get the means and variances for each group, I'm
assuming we would need to use ROIs. Could you please clarify whether it's
possible to extract means across the whole brain?
Thank you,
Hannah
Re: [Freesurfer] Welch's t-test for HOV violation
2019-03-13 Thread Greve, Douglas N.,Ph.D.
Hi Hannah, please include the previous correspondence so that we have context.
Also, please remember to post to the list and not to us personally.
thanks!
doug
On 3/13/19 10:52 PM, Hannah CK wrote:
External Email - Use Caution
Hi Dr. Greve,
Thank you for this response. Can extracting the means be done at the
whole-brain level? If we get the means and variances for each group, I'm
assuming we would need to use ROIs. Could you please clarify whether it's
possible to extract means across the whole brain?
Thank you,
Hannah
___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu>
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Re: [Freesurfer] Welch's t-test for HOV violation
2019-03-11 Thread Greve, Douglas N.,Ph.D.
I used to have this about 20 years ago, but I stopped supporting it when it did
not fit cleanly into the GLM ... The shortest route is probably to do two
separate analyses with mri_glmfit, one for each group. This will output means
and variances for each group. Then use fscalc to compute maps of the Welch's t.
On 3/11/19 1:28 PM, Hannah CK wrote:
>
> External Email - Use Caution
>
> We are conducting t-test comparisons across groups in which there is a
> large discrepancy in group sizes. HOV is violated. I've searched for
> ways to run Welch's t-test in FreeSurfer (or a similar analysis that
> does not assume homogeneity of variance) but am not finding one.
>
> Could anyone please advise on how to do this?
>
> Thank you.
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu>
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
