Dear all,
I am encountering a very common problem:namely dealing with low explained variance in my DCMs. I have tried re-estimating my DCMs with empirical priors (using the spm_dcm_peb_fit) functions, unfortunately this doesn't yield better results. In one of the recent posts Amirhossein Jaffarian has suggested another workaround:
" One way to improve your results (for those that the model not fitted well) is to take estimation for a good fitted model (more explained variance) as a prior for those otherwise. The issue you asked caused by (effectively) local minima of optimization scheme which can be addressed by good initialization of the inversion algorithm in your group study. "
Now my question is how to implement this in DCM? What parameters and which lines of code need to be changed for that?
Thank you very much in advance!
|