I have I kind of technical question. I know dynare works just with continouos shocks. Anyway by using just first order approximation, one should be able to work with discrete shocks by holding the 0 mean assumed by Dynare. Is this correct?
Thank you very much
Yes, Dynare does not make any assumption over the shape of the shocks distribution, at least at first and second order.
At first order, your shock distribution just has to be zero-meaned.
At second order, your shock distribution has to be zero-meaned, and you explicitely set the 2nd moment.
The third order case is a bit different: Dynare needs to know the 3rd moment of the shock distribution, and instead of letting the user giving that moment, it assumes that it has the moments of a normal distribution.
For higher orders (only in Dynare++) there is also an assumption of normality.
Ok, thank you very much. Very clear explanation!
I know that by using second order approximation Dynare consider the second moments of the exogenous shock as well, but I don’t get any significant difference by changing the variance of my shocks? Do you have any idea of why is that?
And I used “stoch_simul(order=2,nocorr, nomoments,replic=1,simul_seed=123,periods=2000, IRF=50)” to plot the IRF but without any result. Could you help me?
Thank you very much.
At second order, only the constant term in the decision rule (the delta square in the reference manual, the so-called correction term) will change when the variance of shocks changes. This is a well-known property of the 2nd order approximation.
Concerning the IRFs, if they are not displayed, it probably means that they are explosive (this happens frequently at 2nd order). You can try the “pruning” option, which implements an alternative way of computing IRFs which guarantees that they are not explosive.
Thank you very much. Sorry if I asked the same questions on another post, no need to answer them.
Thank you again