Bayesian estimation problem

[federico]

Hi everyone,

I run a Bayesian estimation of my model and I got a strange results and maybe sameone could help me.

In the attachment you can find the mode check of same parameters and you could clearly see that parameter chi_sb give me same issue.

As far as I know the red dot points tell me that the blanchard khan conditions are not satisfied.
Anyway If I run a simulation in one of that points everything is fine and I obtained coherent results.

After the estimation , I run the identification command to get same extra information and i obtained this error


==== Identification analysis ====

Testing prior mean

-----------
Parameter error:
The model does not solve for prior_mean with error code info = 19

info==19 %! The steadystate routine thrown an exception (inconsistent deep parameters).
-----------

Try sampling up to 50 parameter sets from the prior.

-----------
Identification stopped:
The model did not solve for any of 50 attempts of random samples from the prior
-----------

and finally using the model_diagnostic command I got

model diagnostic can't obtain the steady state


Since It's the first time that I obtained this strange results I have no idea how to detect my problem.
NB: If I get rid of chi_sb in the estimation block leaving the parameter calibrated and run the estimation I have no problem and the mode search results succesful

Any idea?

[jpfeifer]

It's hard to tell. The red dots do not signify that the Blanchard Kahn conditions are not satisfied, but general problems with the model. It could be BK, but it could also be that the steady state could not be found. Identification at the prior mean does not help you as you are interested in the posterior.

I find it puzzling that model_diagnostics complains about the steady state. How could Dynare compute the posterior if it cannot compute the steady state. You may need to post your mod- and datafile.