Better handling of NPD MH proposal covariance
Created by: nerdcha
Hi there, thanks a bunch for all your work on Dynare! Do you accept pull requests?
This is a tweak I always make to my local copy. Obviously an NPD Hessian is less than ideal, but with this change you can run the MCMC for a few thousand draws, which almost always makes it obvious where the problem is. The mode_check
feature is a powerful tool, but doesn't catch everything.