Hi Xiaoyu,

again (see also other email) you should not run left hemispheres on the right. This could introduce a processing bias, especially when you force all left to be healthy and right to be diseased (or vice-versa).

Instead run your images normally and then for the univariate , simply compute the diff between the left and right ROI's (always healthy - diseased). Then do a normal GLM analysis on this signed difference.

Best, Martin

On Tue, 2018-08-28 at 21:27 +0000, Wang, Xiaoyu wrote:

External Email - Use Caution         Hello Freesurfer Developers,   I’m attempting to use Univariate LME analysis to compare the symptomatic and asymptomatic hemispheres of early stage Parkinson’s disease patients.   I re-organized the data so that the symptomatic hemisphere is the left hemisphere and the Asymptomatic is the right (by flipping the orig.mgz before recon-all).   I followed the LME model tutorial for univariate on the freesurfer wiki and created my qdec.tabel.dat to have the same layout as in the example with group 1 as left hemisphere and group 2 as right hemisphere.   I have a design matrix to do a simple linear model containing a group by time interaction with the contrast of [0 0 0 1 0 0]   After running the F-test however, I noticed that the p-value was really inconsistent across atlases. I used multiple atlases and noticed that the p-value can be as low as 1.00E-30 for many structures. This is too good to be true especially since after sending the thickness value to a statistician she came back with values greatly different from mine. However, the Bhat values were similar.   Can someone give me a hint as to what is going on?   Xiaoyu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer