comparing posterior parameters to priors

[tmoursi]

The Dynare wiki: DYNARE USER GUIDE by Tommaso Mancini Griffoli, 2007: contents errata, comments, additional material… states “When comparing the posterior distributions, displayed at the end of the program, one observes that, for several parameters, they are very close to the prior.” Is there a rule (even of thumb) to determine how close is close? Or is it “entirely” subjective?

Thanks in advance for your help.

Tarek Moursi

[jpfeifer]

No, there are no guidelines here. It is important to make sure that i) the MCMC correctly sampled from the posterior, i.e. that convergence has been obtained and ii) that the parameters are actually identified. If these conditions are met, you simply know that the data is not that informative for updating your prior.