Hello,
I have a question regarding the interpretation qdec's output for the custom ROI's thickness estimates. I ran an analysis comparing cortical thickness in children with and without ADHD (categorical variable), controlling for age and demeaned ICV, using 5.0. After controlling for multiple comparisons with the monte-carlo simulation, there were 2 ROIs in the right hemisphere that were significantly different across groups: a regions near parstriangularis and near medial orbital cortex. I then created custom ROIs for these regions, and extracted thickness estimates for each participant for each of these ROIs.
I then decided to do a "sanity check" of the data -- run a linear regression on these extracted thickness estimates to see if the mean thickness in each region differed as a function of diagnosis (again controlling for age and demeaned ICV). By logic, I should see differences in thickness estimates for these 2 ROIs across the 2 groups, because that's how these ROIs were defined, right? But in reality, there is no difference at all between groups, with p-values around .9!
I tried saving the ROIs by both manually tracing and with the mri_surfcluster command -- I get very similar estimates of thickness (correlation coefficients of .9 across 32 subjects). So that's not the cause of error. Both manual and automatic extraction fails to find differences across subjects that I described above.
What am I missing? is the original GLM in QDEC that finds differences between groups doings something other than looking for differences in each voxel/region across groups?
Thanks in advance for the help! Maria
*******************************
Maria Kharitonova, Ph.D. Postdoctoral Research Fellow Laboratories of Cognitive Neuroscience Boston Children's Hospital Division of Developmental Medicine Harvard Medical School