Parameter Restriction in Bayesian Estimation

[earsmall]

Hi all. I have been trying to reproduce Lubik and Schorfheide's 2007 results.
In this paper, authors compare 2 models, unrestricted vs restricted by using posterior odds ratio.

[earsmall]

Additionally, the restriction is that psi3 = 0 which is included in the monetary policy rule.

R = rho_R * R(-1) + (1-rho_R)*( psi1*pi + psi2*y + psi3*e ) + eps_R

I think if we impose restriction, psi3=0, priors of other parameters, rho_R, psi1 and psi2 should be changed.
How do you think?

[jpfeifer]

Yes, you should compare the marginal data densities. The restrictions are imposed by not declaring the respective parameters as estimated parameters. That is, you can declare them in the

Code: Select all

parameters

statement, but not in the

Code: Select all

estimated_params

block.

Why would you want to change the priors when imposing a restriction?

[earsmall]

Thank you for your comment.

I also think that prior must be changed when imposing restriction
but according to the authors, they did use the same priors.