simulation after bayesian estimation

[shufecj]

Hello all,

I am solving a model using Bayesian estimation with dynare. After running the code, I get the mode of parameters and standard deviations of exogenous shocks. In principle, I should be able to simulate the model when I set these parameters and standard deviations equal to the mode. Is it right?
However, when I try to simulate the model, dynare shows “Blanchard Kahn conditions” doesn’t hold. Why is that?
It really puzzles me. Thanks a lot!

Best!
Jie Chen

P.S. Attachment is the codes and data I use. The file “code.mod” is to run the Bayesian estimation. The file “labroinvest.xls” is the data one should use when run the “code.mode”. It will take about half an hour to get the mode. The file “sim.mod” is to run the simulation.
The mode I get is stored in “code_mode.mat”. But I don't know how to load a ".mat" file.

[jpfeifer]

Please upload everything into a zip-file and upload it. I need the mode-file.

[shufecj]

Please upload everything into a zip-file and upload it. I need the mode-file.

I have upload the zip file and it contains the mode-file now. Thank you so much!

[jpfeifer]

Are you sure you are working with a correct mode-file? When I use

Code: Select all

mode_file=code_mode

in your estimation file code.mod I get the same error, making it very unlikely that the mode-file came from this mod-file.

[shufecj]

Are you sure you are working with a correct mode-file? When I use

Code: Select all

mode_file=code_mode

in your estimation file code.mod I get the same error, making it very unlikely that the mode-file came from this mod-file.

Sorry for the late reply. My computer broke down and I have to fix it.
It is strange that you can't run the code when using "mode_file=code_mode" as I can run it. (see attachment as the screenshot of the result as well as the code named "code1") Can you help me figure out why it is the case. My version of dynare is 4.3.3.
Thank you very much!

Jie Chen

[jpfeifer]

Your model is somewhat strange in that it requires a unit root to be counted as an explosive root. This is what estimation does by setting

Code: Select all

options_.qz_criterium=1-1e-6;

The simulation code in contrast uses

Code: Select all

options_.qz_criterium=1+1e-6;

by default.
You have to use

Code: Select all

stoch_simul(qz_criterium=0.9999999)

[shufecj]

Your model is somewhat strange in that it requires a unit root to be counted as an explosive root. This is what estimation does by setting

Code: Select all

options_.qz_criterium=1-1e-6;

The simulation code in contrast uses

Code: Select all

options_.qz_criterium=1+1e-6;

by default.
You have to use

Code: Select all

stoch_simul(qz_criterium=0.9999999)

It works. It really helps me. Thank you so much!