My guess is that your covariates have almost no range to them and so look like
constant (similar to columns 1 and 2). You can try demeaning and rescaling the
covariates. You can do this by hand and create a new FSGD file or you can add
the following lines anywhere in the fsgd:
RescaleFlag 1
DemeanFlag 1
On 10/26/19 10:45 PM, Nillo, Ryan Michael R wrote:
Hello,
I ran mri_glmfit with the following fsgd file:
GroupDescriptorFile 1
Title CJD
Class MM1_MV1
Class MV2_VV2
Variables frontal
Input 7693 MV2_VV2 2.930667
Input 11985 MV2_VV2 2.410333
Input 13082 MV2_VV2 2.529333
Input 13425 MV2_VV2 2.593667
Input 13667 MM1_MV1 2.628667
Input 13900 MM1_MV1 2.675333
Input 14094 MM1_MV1 2.549333
Input 15566 MV2_VV2 2.666667
Input 15999 MM1_MV1 2.641
Input 16714 MV2_VV2 2.643
Input 16848 MV2_VV2 2.63
Input 16952 MV2_VV2 2.516
Input 7693a MV2_VV2 3.054
Input 11985a MV2_VV2 2.42
Input 13082a MV2_VV2 2.454
Input 13425a MV2_VV2 2.602
Input 13667a MM1_MV1 2.604333
Input 13900a MM1_MV1 2.601333
Input 14094a MM1_MV1 2.523333
Input 15566a MV2_VV2 2.713333
Input 15999a MM1_MV1 2.584333
Input 16714a MV2_VV2 2.673333
Input 16848a MV2_VV2 2.533333
Input 16952a MV2_VV2 2.486
My two contrast matrices are:
one.mtx = [ 0 0 1 0 ]
two.mtx = [ 0 0 0 1 ]
When running the mri_glmfit command "mri_glmfit --glmdir new.frontal.lgi --fsgd
new.frontal.fsgd.txt --fwhm 30 --surf fsaverage lh --y lh.pial_lgi.fwhm0.mgh
--C one.mtx --C two.mtx”, I get the following output:
INFO: gd2mtx_method is dods
Saving design matrix to new.frontal.lgi/Xg.dat
Computing normalized matrix
Normalized matrix condition is 12820.7
Design matrix ------------------
0.00000 1.00000 0.00000 2.93067;
0.00000 1.00000 0.00000 2.41033;
0.00000 1.00000 0.00000 2.52933;
0.00000 1.00000 0.00000 2.59367;
1.00000 0.00000 2.62867 0.00000;
1.00000 0.00000 2.67533 0.00000;
1.00000 0.00000 2.54933 0.00000;
0.00000 1.00000 0.00000 2.66667;
1.00000 0.00000 2.64100 0.00000;
0.00000 1.00000 0.00000 2.64300;
0.00000 1.00000 0.00000 2.63000;
0.00000 1.00000 0.00000 2.51600;
0.00000 1.00000 0.00000 3.05400;
0.00000 1.00000 0.00000 2.42000;
0.00000 1.00000 0.00000 2.45400;
0.00000 1.00000 0.00000 2.60200;
1.00000 0.00000 2.60433 0.00000;
1.00000 0.00000 2.60133 0.00000;
1.00000 0.00000 2.52333 0.00000;
0.00000 1.00000 0.00000 2.71333;
1.00000 0.00000 2.58433 0.00000;
0.00000 1.00000 0.00000 2.67333;
0.00000 1.00000 0.00000 2.53333;
0.00000 1.00000 0.00000 2.48600;
--------------------------------
ERROR: matrix is ill-conditioned or badly scaled, condno = 12820.7
--------------------------------
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 --glmdir new.frontal.lgi --fsgd new.frontal.fsgd.txt --fwhm 30
--surf fsaverage lh --y lh.pial_lgi.fwhm0.mgh --C one.mtx --C two.mtx
2. The FSGD file (if using one)
3. And the design matrix above
Attempting to diagnose further
SumSq: Min=2.828427 (col 1), Max=10.485683 (col 4)
... could not determine the cause of the problem
Can anyone give me some insight into what might be going wrong?
Thanks so much,
Ryan Michael Nillo
Staff Research Associate I
University of California San Francisco
Department of Radiology and Biomedical Imaging
External Email - Use Caution
_______________________________________________
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
