Hello, I've noticed a behavior in QDEC (v1.1 under FS v4.1.0) that appears to be a bug regarding the "slopes" output to the terminal, and the corresponding regression lines displayed in the scatterplot, when using more than one continuous covariate.
I ran an analysis on cortical thickness on a group of patients using just age and duration of illness as continuous covariates (and listed in that order on the 'Design' tab). Looking at the map for "Does the correlation between thickness and age, accounting for duration_of_illness, differ from zero?", the blue blobs are reported to have negative slopes, and the red blobs have positive slopes, as expected. (This is based on the slope reported to the terminal, as well as the physical slope of the dashed regression line in the scatterplot when I click on a vertex). However, looking at the map for "Does the correlation between thickness and duration_of_illnes, accounting for age, differ from zero?", I found that when I clicked on a vertex in a blue blob the slope reported to the terminal was usually POSITIVE. That is, vertices in blue blobs in this case typically have a slope REPORTED as positive, ACCORDING to the output to the terminal as well as the slope of the regression line in the scatterplot. And vice-versa for vertices in red blobs. Upon further investigation it appears that the bug is that the output to the terminal, and the regression line in the scatterplot, are using the slope value from the first of the two continuous variables regardless of which correlation is actually selected as the active map. An analogous bug exists if you add a class variable to the mix. That is, the "slopes" output to the terminal for each class (using DODS), and used in the regression line of the scatterplot, are from the first of the two continuous variables, regardless of the active map. cheers, -Mike H. P.S. In the process of this, I found stumbled across the following: If you use the "Subject Exclusions" feature to exclude some subjects based on a discrete (class) variable, and attempt to run a DOSS analysis using a different discrete factor in combination with two continuous covariates, Qdec crashes with the following message: *** glibc detected *** free(): invalid next size (fast): 0x11303558 *** Abort But, if you attempt a DOSS analysis using just one of the two continuous covariates instead, no problem... -- Michael Harms, Ph.D. -------------------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 Renard Hospital, Room 6615 Tel: 314-747-6173 660 South Euclid Ave. Fax: 314-747-2182 St. Louis, MO 63110 Email: mha...@wustl.edu -------------------------------------------------------------------- _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
