Estimation and load_mh_file

Hi. I estimated a model and the mode check plots are ok, but it seems that some parameters have not reached the convergence yet (but are very close). In particular, there is a parameter which posterior shows 2 peaks, and the calculated mode points to the highest peak (which seems the “true” mode).
So I used :

estimation(order=1,mode_file=mymodel_mh_mode, datafile=file.xlsx, mode_compute=0, mh_replic=500000, mh_nblocks=3,mode_check,mh_tune_jscale,smoother,consider_all_endogenous,load_mh_file);

I give a new name to the model, but it is the same. I copy the files in “metropolis” folder and rename them for the new name.
When I ran the model again, I was expecting that the my previous run (mymodel_mh_mode) would be loaded and the estimation will add more draws using the others that I already have. But the calculated mode changes and, in particular, the calculated mode for that parameter points now to the lowest peak , although the posteriori shows a similar shape than before.

I am not sure if I understood what’s is happening. Would anyone give me some help?

The mymodel_mh_mode you use as the mode-file is not the mode from mode-finding, but from the Metropolis-Hastings.

Thanks, professor Pfeifer.

One more question: when I specify the mode_file mymodel_mh_mode (or mymodel_mode) in estimation command, does dynare use the last draw of MH (or mode-finding in the case of mymodel_mode) as initial point for the optimizer? In this case the “estimated_params_init” block are ignored if present?