Dynare forums

Dear Michel,
How does the option smoother work with MLE? If you don't use the option, the output of oo_ is

mle_mode
mle_std
SmoothedVariables
FilteredVariables
SmoothedShocks

If you use the option, all you get is

mle_mode
mle_std

Is it possible to get some HPDinf and HPDsup with MLE?

Thanks,
Junior

jmaih wrote:Is it possible to get some HPDinf and HPDsup with MLE?

Thanks,
Junior

Hi Junior,

You should get the smoothed shocks as an output of the maximum likelihood estimation, but you cannot have HPDinf and HPDsup in this case because these intervals are computed from the metropolis hastings. We may add asymptotic confidence bands (or bootstraped confidence bands) in the case of a maximum likelihood estimation...

If you really need such intervals you may follow the bayesian approach (with a metropolis) using uniform priors over all the parameters, this will be equivalent to a constrained ML approach, and the intervals will be exact at finite distance.

Best,
St

Thanks a lot Stephane!

Junior