Problems encountered when doing the Bayesian Estimation

[lssp]

Dear Professor Pfeifer,

I have encountered three problems when doing the Bayesian Estimation, would you please give me some hints? Thank you so much!

(1) The result of my Bayesian Estimation displays that the persistence of some shocks(e.g. technology shock, government expenditure shock) are always more than 0.99. Is this because there are some mistakes when I doing the Bayesian Estimation ? like data dealing or prior value setting?

(2) It takes me more than four hours doing the Bayesian Estimation once, is this normal ? What can I do to shorten the time expended on this process ?

(3) It is known that the number of shocks should be equal to the number of observe variables, but what if the number of shocks I want to add in the model are more than the number of observe variables ? I have read some papers that contain this circumstance, but how can I do the Bayesian Estimation in this condition ?

I am looking forward to your reply, thank you very much !

[jpfeifer]

1) Most probably your observation equations are still wrong. The only way a model can account for an unhandled constant is by estimating a unit root
2) Yes, that is normal if you have a lot of draws. Providing an analytical steady state can speed things up
3) As long as the parameters are all identified, you can have more shocks than observables.

[lssp]

1) Most probably your observation equations are still wrong. The only way a model can account for an unhandled constant is by estimating a unit root
2) Yes, that is normal if you have a lot of draws. Providing an analytical steady state can speed things up
3) As long as the parameters are all identified, you can have more shocks than observables.

Dear Professor Pfeifer,

Thank you so much for your reply, I really appreciate that.

Best Wishes !