Try demeaning the continuous variables. By this I mean to compute the mean of the age across all subjects regardless of group, then subtract this mean from the age of each individual. doug
On 6/10/14 10:00 PM, Thomas DeRamus wrote:
Sorry to be a bother, but I seem to be having an issue with my analysis.
I have a control and a patient group, and I'm trying to covary gender the majority is skewed towards one gender.
I'm looking across 4 metrics (thickness, surface area, volume, and lgi) of the cortex.
My continuous variables include Age for all metrics, Total incracranial volume for the volume analysis, total intracranial surface area for the area measure (cubed root squared transform of eTICV), and Mean thickness for the left and right hemisphere (each separate for their respective hemispheres) for thickness.
My goal is to look at changes in these as a function of age, relative to total ICV/SA metrics (because my patient population has arrested brain development in theory), and diagnosis. So I'm examining the following interactions: DiagnosisXAge, AgeXICV, DiagnosisXICV, DiagnosisXAgeXICV
Which originally gave me 6 columns for the "total vol/area/thick" analyses: Control, Patient, ControlAge, PaitentAge, ConICV, PatICV if I'm not mistaken and 4 for the non-tICV lgi analysis.
On the recommendation of the tutorial, I coded my change my 2 class variables to 4 to account for gender (so now its MaleControl, MalePatient, FemaleControl, Female Patient)
Which should give me 12 columns right? ConMale, ConFemale, PatMale, PatFemale, ConMaleAge, ConFemAge, PatMaleAge, PatFemaleAge, ConMaleICV, ConFemICV, PatMaleICV, PatFemaleICV
This analysis will run for volume, area, and lgi (no icv so 8 columns), but I keep getting the following Message from the terminal for thickness:
ERROR: matrix is ill-conditioned or badly scaled, condno = 16115.7
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:
- Your command line:
mri_glmfit --ylh.diagnosisxagexleftThickslopetest.thickness.10.mgh --fsgd GenderGroupingSBM_glmLeftthick.fsgd dods --C diagnosisnonlGIICmeas.mtx --C agesxdiagnosisnonlGI.mtx --C diagnosisnonxagexICmeas.mtx --C ICmeasurexdiagnosisnonlGI.mtx --surf fsaverage lh --cortex --glmdir GenderGrouplh.Controlled_diagnosis_agethickness.glmdir 2. The FSGD file (if using one) 3. And the design matrix above
I see this has been a problem for others, but I'm not sure if its a scaling issue, a problem with my FSGD file, or that mean thicknesses correlate too much with age.
I've attache the FSGD file in question along with the contrast matrixes and the X file (cue theme) from the analysis.
Any input would be greatly appreciated, thanks.
-- *Thomas DeRamus* UAB Department of Psychology, Behavioral Neuroscience Graduate Research Trainee Civitan International Research Center 1719 6th Ave S, Suite 235J, Birmingham, AL 35233 _Phone_: 205-934-0971 _Email:_ tpderamus@gmail.com mailto:tpderamus@gmail.com, faustus@uab.edu mailto:faustus@uab.edu
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer