Dynare forums

Hi,


I am estimating DSGE model using Bayesian estimation using DYNARE.

I wonder what is the optimal number of replication for MH algorithm (default = 20000) and number of parallel chains for Metropolis Hasting algorithm (default = 2).

Cheers,


Sigitas

Hi Sigitas,

This is a crucial question. The number of MH replications must be such that the Markov chain converge to its stationary distribution (the posterior density). There is no theoretical results about this number (we only have asymptotic results on the convergence of a mcmc). You should read chapter 12 in the book by Robert and Casella ("Monte Carlo Statistical Methods") or section 11.4 in the book by Gelman, Carlin, Stern and Rubin ("Bayesian Data Analysis").

Best,
Stéphane.