Simul v.s. stoch_simul

If I can plot some irfs from stoch_simul but all the output from simul is NaN, it means my model is not right, is it?

I am so frustrated.

Not necessarily. Maybe you don’t have enough periods in simul. Remember that you must do the simulation until the variables go back very near steady state.

What are you trying to get from comparing stoch_simul with simul?



Thanks for your reply, Michel. I just want to set a loop outside dynare code to do some searching to match the theoretical IRFs and Empirical IRFs. Simul is better for this purpose but I am really struck with NaNs . Thus I switch to stoch_simul and it can give out some plot, which looks look.

Further, If simul needs extra periods to go back to steady state, how many periods does it need? Say, if I just want IRF for periods and simul gives out 24 numbers, How can I decide which part can be used?

I guess that by empirical IRFs, you mean IRFs obtained from VAR analysis, right?

simul is no better than stoch_simul for that purpose… and simul assumes deterministic perfect foresight.

For simul, you should start with 100 or 200 periods and check whether convergence back to stready state happens any earlier. If that is the case, you can use less periods.

With simul and rplot, if you don’t want to plot all the periods, you can use
will plot ‘int’ periods