 |
 |
Dear Andrea, Thanks for providing details. Getting convergence in 22 iteratoins can be a good thing or otherwise could only be ascertained by looking at the data. I do have some suspicions but it would be easier if I get a hand on one of your inverted DCMs. Best if you send it to me through dropbox or as an email attachment through personal message so as not to inundate others' inboxes? Best wishes, Adeel On Thu, Apr 4, 2019 at 5:31 PM Andrea Wang <[log in to unmask]> wrote: > Dear DCMers, > > Currently I'm doing some analyses using the method described in > Large-scale DCMs for resting-state fMRI ( > https://www.mitpressjournals.org/doi/pdf/10.1162/netn_a_00015). At this > stage, I'm simply trying to replicate the method (same ROI selection) using > my own resting state fMRI dataset (8 min resting state). > > However, I got some unexpected results - the DCM models of all subjects > converged at 22nd step, and all the DCM.Ep.A matrix are almost symmetric > (only tiny value differences between the upper and lower triangular matrix). > > One thing to mention is that I didn't use GLM to extract the time courses, > because I was not able to extract time courses of all ROIs by using the > regressor of DCT between 0.0083-0.15 Hz+WM+CSF+head motion and an F > contrast to test for DCT effects. > > I simply used DPABI to extract time courses and implemented them according > to DCM required structure. The functional connectivity of these time course > is very similar to the ones shown in the Large-scale DCM paper and also the > original ROI-proposing paper The Restless Brain ( > https://www.liebertpub.com/doi/10.1089/brain.2011.0019). I thus assumed > the time courses are correct. > > Did anyone come across similar problem? If so, could you please give a > hand? Thank you very much! > > Best, > Andrea >