Quick question to load_mh_file

[Palumenis]

Hi all,

is it the same if I

a) estimate a model with 100.000 draws (mh_replic=100000)
b) estimate a model with 50.000 draws (mh_replic=50000) twice with (load_mh_file)?

I suppose yes, but just want to make sure. Thanks in advance!

[jpfeifer]

This will largely (but not completely) be the same. Before Dynare 4.5 the load_mh_file option will not continue with the same state of the random number generator, leading to usually small differences (if your chain converged to the ergodic distribution). When using the load_mh_file option, you should use mode_compute=0 together with the mode_file option to make sure you continue with the same proposal density derived from the mode as in the previous estimation (In newer Dynare versions mode-finding should be pseudo-deterministic so that the same proposal should be used, but better be safe than sorry).