Hi Doug et al,
We're running into a problem doing a wrfx group analysis using FreeSurfer 4.5. Looks like the glmfit is having trouble with the weights. We tried running it for different subsets of our subjects, and get the same thing, so it's not one subject contaminating the analysis. Any guess where we should look for tracking this down? Presumably the cesvar files are screwed up somehow, which would imply something went wrong in the individual selxavg3's. But any ideas about what could lead to this? The individual overlays all seem to look ok.
[sidkundu@brodmann all-vs-null]$ mri_glmfit --y tal.ces.nii --osgm -- glmdir tal.wrfx.osgm --nii --mask ../tal.mask.nii --wls tal.cesvar.nii
$Id: mri_glmfit.c,v 1.138.2.15 2009/07/27 17:43:10 greve Exp $ cwd /home/sidkundu/freesurfer/sessions/Heading-TTP/1-24-12-TTP/ TTP_analysis/all-vs-null cmdline mri_glmfit --y tal.ces.nii --osgm --glmdir tal.wrfx.osgm --nii --mask ../tal.mask.nii --wls tal.cesvar.nii sysname Linux hostname brodmann.bu.edu machine x86_64 user sidkundu FixVertexAreaFlag = 1 UseMaskWithSmoothing 1 OneSampleGroupMean 1 y /home/sidkundu/freesurfer/sessions/Heading-TTP/1-24-12-TTP/ TTP_analysis/all-vs-null/tal.ces.nii logyflag 0 usedti 0 mask ../tal.mask.nii maskinv 0 glmdir tal.wrfx.osgm IllCondOK 0 DoFFx 0 wFile tal.cesvar.nii weightinv 1 weightsqrt 1 Creating output directory tal.wrfx.osgm Loading y from /home/sidkundu/freesurfer/sessions/Heading-TTP/1-24-12- TTP/TTP_analysis/all-vs-null/tal.ces.nii Saving design matrix to tal.wrfx.osgm/Xg.dat Matrix condition is 1 Pruning voxels by thr: 0.000000 Found 260736 voxels in mask Saving mask to tal.wrfx.osgm/mask.nii search space = 2.08589e+06 ERROR: MRInormWeights: value less than or eq to 0. c=34, r=2, s=33, v=0
I'm guessing that there is a voxel (or more) in cesvar that is 0 for a subject (or more). Usually such voxels are masked out by pruning the input ces. You can create your own mask with
mri_concat tal.cs.nii tal.cesvar.nii --prune --o junk.mgh mri_binarize --i junk.mgh --min .00000000000001 --o newmask.mgh
Then pass this mask to mri_glmfit with the --mask option.
doug
Finnegan Calabro wrote:
Hi Doug et al,
We're running into a problem doing a wrfx group analysis using FreeSurfer 4.5. Looks like the glmfit is having trouble with the weights. We tried running it for different subsets of our subjects, and get the same thing, so it's not one subject contaminating the analysis. Any guess where we should look for tracking this down? Presumably the cesvar files are screwed up somehow, which would imply something went wrong in the individual selxavg3's. But any ideas about what could lead to this? The individual overlays all seem to look ok.
[sidkundu@brodmann all-vs-null]$ mri_glmfit --y tal.ces.nii --osgm --glmdir tal.wrfx.osgm --nii --mask ../tal.mask.nii --wls tal.cesvar.nii
$Id: mri_glmfit.c,v 1.138.2.15 2009/07/27 17:43:10 greve Exp $ cwd /home/sidkundu/freesurfer/sessions/Heading-TTP/1-24-12-TTP/TTP_analysis/all-vs-null cmdline mri_glmfit --y tal.ces.nii --osgm --glmdir tal.wrfx.osgm --nii --mask ../tal.mask.nii --wls tal.cesvar.nii sysname Linux hostname brodmann.bu.edu http://brodmann.bu.edu/ machine x86_64 user sidkundu FixVertexAreaFlag = 1 UseMaskWithSmoothing 1 OneSampleGroupMean 1 y /home/sidkundu/freesurfer/sessions/Heading-TTP/1-24-12-TTP/TTP_analysis/all-vs-null/tal.ces.nii logyflag 0 usedti 0 mask ../tal.mask.nii maskinv 0 glmdir tal.wrfx.osgm IllCondOK 0 DoFFx 0 wFile tal.cesvar.nii weightinv 1 weightsqrt 1 Creating output directory tal.wrfx.osgm Loading y from /home/sidkundu/freesurfer/sessions/Heading-TTP/1-24-12-TTP/TTP_analysis/all-vs-null/tal.ces.nii Saving design matrix to tal.wrfx.osgm/Xg.dat Matrix condition is 1 Pruning voxels by thr: 0.000000 Found 260736 voxels in mask Saving mask to tal.wrfx.osgm/mask.nii search space = 2.08589e+06 ERROR: MRInormWeights: value less than or eq to 0. c=34, r=2, s=33, v=0
freesurfer@nmr.mgh.harvard.edu
-
Douglas N Greve -
Finnegan Calabro