I have examined the relation between LGI and a behavioural continuous scale (SCQ) on QDEC while controlling for sex, age and surface area in my model. When I want to do another analysis, doing exactly this but also controlling for diagnosis as well, QDEC gives the following error:

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.

If you seek help with this problem, make sure to send:

  1. Your command line:

    mri_glmfit.bin --y /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/y.mgh --fsgd /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/qdec.fsgd dods --glmdir /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0 --surf fsaverage lh --label /d/mjt/9/users/avideh/data/recon/fsaverage/label/lh.aparc.label --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-Avg-Intercept-pial_lgi.mat --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-Avg-pial_lgi-RBS-Cor.mat --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-Diff-ADHD-ASD-Intercept-pial_lgi.mat --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-Diff-ADHD-ASD-Cor-pial_lgi-RBS.mat --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-Diff-M-F-Intercept-pial_lgi.mat --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-Diff-M-F-Cor-pial_lgi-RBS.mat --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-X-Dx-Sex-Intercept-pial_lgi.mat --C /d/mjt/9/users/avideh/data/recon/qdec/lh_LGI_RBS_overallscore_NDDs_controlling_SexAgeSA_fhwm0/contrasts/lh-X-Dx-Sex-Cor-pial_lgi-RBS.mat

  2. The FSGD file (if using one)

  3. And the design matrix above

Attempting to diagnose further

SumSq: Min=0.488835 (col 11), Max=3.834060 (col 6)

 ... could not determine the cause of the problem

My understanding is that QDEC allows for controlling for 2 discrete variables, thus I don't know what I am doing wrong if I also control for diagnosis (in addition to sex)?

I would really appreciate your help, 

Thanks so much, 

Avi