External Email - Use Caution
Hi freesurfer experts,
I am trying to run a GLM fit between cortical thickness and some other continuous marker. My cortical thickness were stacked before using the mris_preproc and surf2surf commands. I have an FSGD file with the same subjects as used during the stacking of the thickness maps, but I added a new continuous regressor (my independent variable of interest). However, some subjects have a missing value for this regressor, so I added NaN. This makes me run into the following error. When I replace the NaN values by random numbers in the FSGD, it all runs fine.
M = [ ... 1.00000 0.99577 0.70711 0.98278 0.97547 -0.97111 nan 0.99577 1.00000 0.70628 0.97775 0.97819 -0.95833 nan 0.70711 0.70628 1.00000 0.68252 0.70488 -0.70032 nan 0.98278 0.97775 0.68252 1.00000 0.95399 -0.96466 nan 0.97547 0.97819 0.70488 0.95399 1.00000 -0.92863 nan -0.97111 -0.95833 -0.70032 -0.96466 -0.92863 1.00000 nan nan nan nan nan nan nan nan ] Normalized matrix condition is 1e+08 Design matrix ------------------ <design matrix> ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08 -------------------------------- Possible problem with experimental design: Check for duplicate entries and/or lack of range of continuous variables within a class.
I made sure that my FSGD file does not have any weird characters, so i applied the cat fsgd.txt | sed 's/\r/\n/g' > new.fsgd.txt, but this also didn't help.
Thank you for your help, Julie
NaN will not work for this. You can't do incomplete designs with an FSGD. You can create your own design matrix and pass it with --X, otherwise you'll have to exclude those subjects or put some reasonable (eg, imputed) value
On 1/12/2021 6:35 PM, Julie Ottoy wrote:
External Email - Use Caution
Hi freesurfer experts,
I am trying to run a GLM fit between cortical thickness and some other continuous marker. My cortical thickness were stacked before using the mris_preproc and surf2surf commands. I have an FSGD file with the same subjects as used during the stacking of the thickness maps, but I added a new continuous regressor (my independent variable of interest). However, some subjects have a missing value for this regressor, so I added NaN. This makes me run into the following error. When I replace the NaN values by random numbers in the FSGD, it all runs fine.
M = [ ... 1.00000 0.99577 0.70711 0.98278 0.97547 -0.97111 nan 0.99577 1.00000 0.70628 0.97775 0.97819 -0.95833 nan 0.70711 0.70628 1.00000 0.68252 0.70488 -0.70032 nan 0.98278 0.97775 0.68252 1.00000 0.95399 -0.96466 nan 0.97547 0.97819 0.70488 0.95399 1.00000 -0.92863 nan -0.97111 -0.95833 -0.70032 -0.96466 -0.92863 1.00000 nan nan nan nan nan nan nan nan ] Normalized matrix condition is 1e+08 Design matrix ------------------ <design matrix> ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08 -------------------------------- Possible problem with experimental design: Check for duplicate entries and/or lack of range of continuous variables within a class.
I made sure that my FSGD file does not have any weird characters, so i applied the cat fsgd.txt | sed 's/\r/\n/g' > new.fsgd.txt, but this also didn't help.
Thank you for your help, Julie
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Greve, Douglas N.,Ph.D. -
Julie Ottoy