Can you send all the things that mri_glmfit suggests sending?
On 2/6/19 5:16 AM, Michal Tanzer wrote:
External Email - Use Caution
Hello FreeSurfer Developers,
I’m attempting to compute mri_glmfit to correlate cortical thickness with my dependent var while controlling for age, gender, iq and MRIcenter (2 centers). I don’t have any interactions or assumptions on group differences so I am using DOSS design (with fsgd file)
I have two questions regarding the analysis, which I couldn’t find in the archives ( I copied the error and attaching the fsgd file)
When I am trying to control for the different variables using 4 classes (Center1Female, Center2 Female, Center1Male, Center2,Male) (Correlation matrix 0, 0 , 0 , 0 , 0, 0, 1) I am getting the error:
ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+10
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 lh.ext.thickness.10.mgh --fsgd EXT2.fsgd doss --C EXT2.mtx --surf fsaverage lh --cortex --glmdir lh.ext2.glmdir 2. The FSGD file (if using one) 3. And the design matrix above Attempting to diagnose further SumSq: Min=0.000000 (col 1), Max=0.000000 (col 0) Column 1, all values are 0 --
However when I am not using the 2 sites as classes and using just the gender as a class (0,0,0,0,1) everything runs without errors. Is there any other way to control for the MRIsite?
Thank you for your help with this matter Best wishes Michal
-- Michal Tanzer, PhD
Postdoctoral fellow; Clinical Psychologist. Research Department of Clinical, Educational and Health Psychology, University College London, United Kingdom
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer