Estimation and Stoch_simul

Hi Prof. and to all,

I have two questions regarding the HBayesian estimation (which I could not figured out by reading the dynare.pdf)
I know after running the estimation(mode_compute=4, …mh_replic=3000, … ) command I can run the stoch_simul(…) command.

  1. What is the difference between plugging the option [moments_varendo] inside the estimation command
    (which computes the theoretical moments of the posterior distribution) , and
    NOt plugging this option but running the stoch_simul(…) command instead of it (which as I know computes theoretical moments
    using posterior means (as I know it) .

my second question is :

  1. following question (1) is there a way I can run the stoch_simul not with the posterior mean but rather with the posterior_median.

these two I could not figure out from the dynare.pdf (though the second question might be very simple)

Thanks a lot for any suggestions I can get !

Hi can someone help me with the questions in the previous post pls ?

Prof. Pfeifer your comment would be appreciated!

  1. Estimation will provide the posterior mean of the objects while stoch_simul will provide the objects at the posterior mean. Due to Jensen’s Inequality, these are two conceptually very different things.
  2. Add

xparam1 = get_posterior_parameters('median'); M_ = set_all_parameters(xparam1,estim_params_,M_);
between estimation and stoch_simul.

  1. I got the second question. Thank you very much Professor. Very helpful.

  2. I am not quite clear about the first issue Professor.

1.a) When we talk about the theoretical moments following an estimation
which one is commonly used - the option [moments_varendo] or [stoch_simul] Professor ?

If the [stoch_simul] is used commonly, then when is [moments_varendo] used and vice versa ???

Thank you very much for your time in this learning process Professor.

Please have a look at [Variance Decomposition)

Thanks Prof. Pfeifer!

It’s clear now what you meant !

As I obtained the moments with estimation(…,moments_varendo,…) I checked the results. But what I get for the correlations for example is:

a cell arrray of the form ‘‘oo_.PosteriorTheoreticalMoments.dsge.correlation.Mean.y.y’’

A quick observation is that:
[1]. there are 5 values within that , and
[2]. none of them is = 1 (as I would expect the correlation(y,y) = 1 no matter whether is average at th mean or mean of the average )

Regarding [1] Could it be due to the default option of [ar=5] ??
I do not understand [2] though .

From the manual on

[quote]Auto- and cross-correlation of endogenous variables. Fields are vectors with correlations
from 1 up to order[/quote]

You would get the 1 you are looking for with the


I see.
That’s a separate option of ‘‘estimation’’ - command.

Many thanks Professor.