I’m at the start of an estimation project and would like to use the Laplace approximation instead of MCMC everywhere (i.e., when computing marginal likelihood and computing the posterior distribution of the parameters). Is it possible to have Dynare shut down the MCMC and just do Laplace?
just use the option mh_replic=0, so Dynare will stop after the computation of posterior mode and don’t go into MCMC.
Note however that the numerical optimization used to find the posterior mode may fail and stop in the wrong place. Beware of parameters with virtualy 0 standard error (if the optimization stops against a constraint) or when the improvements drop too rapidly.