Parameter bounds

Dynare help (legacy posts)
1

The dynare documentation it says that parameter bounds are specified for the maximum likelihood estimation.

I suppose when we can also specify the parameter bounds in case of Bayesian estimation using the same format. However, the online manual always refers to them as bounds for maximum likelihood estimation.

could somebody confirm this.

thanks
abhishek.

2

Hi Abhishek,

You don’t need to specify parameter bounds in case of bayesian estimation. These bounds are implicitly defined by the support of the prior densities. So if you need tighter bounds for a parameter you just have to change the prior density (for instance by decreasing the prior variance).

Indeed, it may be a very bad idea to add parameter bounds in case of bayesian estimation because this will make the estimation of the marginal posterior density of the data (used for model comparison) wrong.

Best,
Stéphane.