Diffuse (improper) priors in Dynare?

Estimation
1

Dear Team Dynare,

Is there a straightforward way to impose improper diffuse priors on parameters for estimation in Dynare?

I’m particularly interested in imposing improper priors on standard errors for shocks. For a shock parameter (or any other parameter that is strictly positive but otherwise unbounded), a truly diffuse prior should be an improper prior with p_prior(sigma_shk) proportional to 1/sigma_shk (where sigma_shk is the standard deviation of the shock).

As it is, the closest I’ve come to a truly diffuse prior for shock parameters is a flat prior, but this is a poor substitute.

Any assistance here is appreciated.

Thank you!

2

At the current stage, Dynare only supports priors of the documented distributional classes. What you describe does not fall into these classes. You would have to modify Dynare routines to implement this yourself.