Choice of priors in estimation
Posted: Tue Oct 06, 2009 10:07 amDear all.
Before starting the search for the posterior mode in estimation Dynare shows the initial value of the log posterior (or the objective function).
Two questions:
1. Is this initial objective function value computed at the mean of the assumed parameter prior distributions?
2. If yes, is it correct to claim that a set of prior distributions that produces a higher initial objective function value will also produce a higher posterior marginal data density (e.g. higher Modfied Harmonic Mean value) and that therefore this is a good way for one to choose the priors?
Thanks a lot a regard!
George
Before starting the search for the posterior mode in estimation Dynare shows the initial value of the log posterior (or the objective function).
Two questions:
1. Is this initial objective function value computed at the mean of the assumed parameter prior distributions?
2. If yes, is it correct to claim that a set of prior distributions that produces a higher initial objective function value will also produce a higher posterior marginal data density (e.g. higher Modfied Harmonic Mean value) and that therefore this is a good way for one to choose the priors?
Thanks a lot a regard!
George