Stoch_simul & ME shocks after estimation

Stochastic simulations
1

Dear Prof Johannes,

When the estimation includes some measurement errors, does the stoch_simul after estimation accounts for measurement error shocks?

If so, how we can repeat stoch_simul with ME shocks since shock block does not give us any choice to ME shock.

Another issue is the unconditional variance composition with option moments_varendo may change from run-to-run. Is it valid? What should we report if we obtain somewhat difference among runs?

Thank you and best regards,

2
  1. What do you mean with account for measurement error? What do you expect?
    In Dynare 4.5 measurement error is not taken into account for variance decompositions (which is essentially the only step in stoch_simul where they would matter). You need the unstable version/4.6. See https://github.com/DynareTeam/dynare/issues/1486
  2. What do you mean with run-to-run? As the MCMC is stochastic, some random variation is expected. But those variations should be small. If they are big, then either you do not have enough subdraws from estimation or your MCMC has not converged in the first place.
3

Thank you for your answer Prof Johannes!

I do expect I can do stoch_simul along with ME shocks as currently we only declare structural shocks in the “shock block”.

Should I explicitly declare eps_me attach to each observation equation and add them to shock block to run stoch_simul?

Thank for your help!

4

If you declare measurement error like a normal structural shock, then everything works, regardless of the Dynare version.