Simulations for mode computation

[Susana]

Hi all!

In order to avoid any dependence from the initial values, I'm trying to compute the mode for my model sampling 100 times from the prior distribution, and then using those values as the estimated_params_init. The problem I encounter is that after 15-20 simulations I start to get an error saying that my "Blanchard Kahn conditions are not satisfied: indeterminacy". The error seems somehow random and not related to the parameter values that I'm selecting given that the first draws work fine. Has somebody encountered a problem like this? Thanks a lot for your help.

[jpfeifer]

This is a common problem that your model is not stable or not determinate for some regions of the prior parameter space. See also the paper of An/Schorfheide. This is not a problem with Dynare.

[Susana]

Thanks!

[flyingpig]

Hi all!

In order to avoid any dependence from the initial values, I'm trying to compute the mode for my model sampling 100 times from the prior distribution, and then using those values as the estimated_params_init. The problem I encounter is that after 15-20 simulations I start to get an error saying that my "Blanchard Kahn conditions are not satisfied: indeterminacy". The error seems somehow random and not related to the parameter values that I'm selecting given that the first draws work fine. Has somebody encountered a problem like this? Thanks a lot for your help.

Hi, I am recently also learning how to do what you described. Would you please share with your .mod file, I would like to learn how to do this. Please, please.