Hello all,
I keep running into the following error when attempting to run a qdec analysis:
ERROR: matrix is ill-conditioned or badly scaled, condno = 10196.8
Attempting to diagnose further
SumSq: Min=2.449490 (col 1), Max=26.457514 (col 7)
The scale is much different between columns 1 and 7, you may want to
normalize by subtracting the mean and dividing by the standard deviation.
My analysis is setup to have Gender and Delirium selected as discrete factors, and Age and RThickness selected as the nuisance factors. This corresponding LH analysis does work when using LThickness. Per an earlier suggestion in the archives, I have de-meaned the age which is what allowed me to use LThickness as a nuisance factor in the first place as it threw errors initially. I have included my .dat file to reference.
Below is what the error suggested to include:
* Command line:
mri_glmfit.bin --y /Volumes/data/Paul/CorticalT/subjects/qdec/Untitled/y.mgh --fsgd /Volumes/data/Paul/CorticalT/subjects/qdec/Untitled/qdec.fsgd dods --glmdir /Volumes/data/Paul/CorticalT/subjects/qdec/Untitled --surf fsaverage rh --label /Volumes/data/Paul/CorticalT/subjects/fsaverage/label/rh.aparc.label --C /Volumes/data/Paul/CorticalT/subjects/qdec/Untitled/contrasts/rh-Avg-Intercept-thickness.mat --C /Volumes/data/Paul/CorticalT/subjects/qdec/Untitled/contrasts/rh-Diff-Male-Female-Intercept-thickness.mat --C /Volumes/data/Paul/CorticalT/subjects/qdec/Untitled/contrasts/rh-Diff-1-0-Intercept-thickness.mat --C /Volumes/data/Paul/CorticalT/subjects/qdec/Untitled/contrasts/rh-X-Gender-Delirium-Intercept-thickness.mat
* FSGD file: I am not using one * Design matrix: attached
Any help is greatly appreciated!
Best,
Dan