Dynare forums

by **jmadeira** » Sat Dec 29, 2007 2:51 pm

Hi,

I'm having trouble interpreting output. I'm not sure how are the graphs suposed to look like (particularly the MCMC diagnostics).
Does anyone know a more detailed reference on this subject besides the user guide?

I also notice that my acceptation rates are very low (almost zero), independently of the mh_jscale used. I suspect this is a problem and this is the reason why the posterior mode is so close to the prior (and why it has very small standard deviation).

Does anyone have any suggestions or comments that could help me?
Best,
Joao

by **reubenpjacob** » Tue Jan 01, 2008 1:22 pm

the fact that the posterior mode is close to the prior mode, has nothing to do with MCMC. the mode is found by the optimiser and not the M-H algorithm. MCMC uses the mode as a starting point.
try reducing the jscale to a very low value and see if the acceptance rate improves.

by **jmadeira** » Tue Jan 01, 2008 2:41 pm

Hi,

Thanks Reuben, you were a great help!
Best,
Joao

by **jlhuillier** » Thu Jan 03, 2008 12:33 am

Where do those outputs (MCMC diagnosis) appear? I am using version 4 and all I get if the graphs of the priors, and then priors/posteriors (+ eventually smoothed shocks/variables).

Thanks ! Jean-Paul

by **jmadeira** » Thu Jan 03, 2008 3:26 am

Hi,
I'm using version 3.065 and they show up right after the end of the metopolis hastings simulations.

Dynare forums

interpreting output

interpreting output

Thanks!

Who is online