Posterior distribution too flat

Dear all,

I have a model with 3 shocks and use three observables for the Bayesian estimation. After estimation, I found the standard deviation of shock “v”, sigma_v, has a very flat posterior distribution, although the mode_check of this parameter looks OK. Does this mean sigma_v is not identified by the given observables? The posterior distribution is attached. PriorsAndPosteriors.eps (116.6 KB) Thanks!

Judging from the plots, the posterior distribution typically yields much higher values for sigma_v than the prior. Sigma_v is identified; the reason that the posterior looks so flat is that the graph is scaled to fit the prior mode (which is much larger because of the narrow posterior distribution).

Hello julianjohs! in fact I have a similar problem and I want to know whether it is common that posterior mean for sigma could be much greater than prior? And posterior distribution is flat in comparison with prior. Thank you!

  1. You did not state what sigma is.
  2. If the posterior mean were commonly bigger than the prior one, people would have adjusted their prior by now.
  3. There are no general statements about the identification strength of particular parameters as they depend on the model

Thank you for your reply!