I’ve already used parallel computing in another model, and it works fine. But with DSGE-VAR, at least in the Dynare version, I’m using (4.5.7 and also 4.6.1), it looks like it’s having a problem. The RW-MH accept rate is at 0% for any number of chained blocks I ask to compute. However, when I don’t use parallel computation, Dynare estimates the DSGE-VAR perfectly. In the same model version, without DSGE-VAR estimation, parallel computing works. I searched and couldn’t find if parallel computing doesn’t even work with DSGE-VAR or an error in this specific version of Dynare.
My question is if this is a Dynare-specific bug, especially in this version, or if parallel computing doesn’t work for this kind of estimation specification?