Drawing from two different distributions

[mkader]

Hi there,

I have a data sample from 1981 to 2015 for Canada. Long story short I am trying to analyse the importance of terms of trade shocks. As an additional exercise I am trying to understanding the importance of terms of trade shocks across time. In this regard, I have divided my sample into two halves (1981 to 1995 & 1996 to 2015). I have also re-estimated AR(1) process for the terms of trade shocks separately for the two sub-samples from the data and obtained the persistence level and the standard deviation. So, everything thing is kept the same except the volatility of the terms of trade shocks.
Now, in terms of simulations for the above exercise, I want to draw from two different distributions in each sub sample and analyse the unconditional moments and the variance decomposition of these two sub samples. I have no clue how to do that in Dynare.

I would really appreciate any help on this.

Best.

[jpfeifer]

If the autocorrelation of the shock process is the same across periods and you are using a first order approximation, this is easily possible. Certainty equivalence assures that you can use the same decision rules even if the volatility changes. What you can then do is generate a shock matrix of appropriate length using the shocks drawn for the different periods. You then feed them into the simult_ function as in e.g. https://github.com/JohannesPfeifer/DSGE_mod/blob/master/RBC_news_shock_model/RBC_news_shock_model.mod
You can then compute the moments of the simulated series.