Matlab's software probably resolves the precision issues better than ours does. With 250 degrees of freedom, you can get extremely significant values. You can look at the f.mgh file. If you load that and the gamma.mgh into matlab and compute t = sqrt(f).*sign(gamma) you should get the same t that you get in matlab. Did you try computing p-values in matlab?

doug

Joonkoo Park wrote:

Hi FreeSurfers,

I'm running a group-level analysis of the fMRI data (beta-maps of individual subjects normalized onto fsaverage). One example would be like:

mri_glmfit --y grouprfx/lh.con_0001.norm.mgh --surf fsaverage lh \ --glmdir grouprfx/lh.con_0001.osgm --osgm

I understand that this is simply performing a one-sample t-test on each vertex. However, the resulting image (lh.con_0001.osgm/osgm/sig.mgh) contains some values that are extremely high or low (1e11 or 1e-11). I guess this is a representation of Inf and -Inf, but this is strange because when I run one-sample t-test manually in Matlab, the resulting t-values should be within an interpretable range. I have many subjects (close to 250), but I don't think that should be the reason.

Any clues on why this is happening and how to resolve it?

Thanks in advance as always!

Joon

---

---

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