I am running dynare on linux (redhat) and MCMC works fine and I get plausible results. If I do the same thing in parallel mode I get an acceptation ratio of exactly zero for all blocks and later on close to singular matrices (compute_mh_covariance_matrix at 74 and amrginal_density at 54 and 67 ). The program contiunes until the end afterwards.
Hi, could you post your dynare project?
I found the mistake. I was using a steady state file using global variables. In the parallel environment this seems to have caused the confusion.
great, let me know of parallel MCMC works then