Getting the same IRF graphs with different variances of uncertainty shock help!


I am running a DSGE model with a third order approximation for the uncertainty shock. One problem that I have is that : when I set different values for the variance of uncertainty shock in the simulation part, results of the IRF graphs remain the same. And I could not find which part is wrong. Could you please help me with this?

My code is attached. try4.mod (3.1 KB)

Thank you very much for your attention !!

Your variance specification needs to be consistent. You set the standard deviation with

var evola; stderr exp(rho_vola);

but then simulate with a two unit shock at

shock_mat(1,strmatch('evola',M_.exo_names,'exact'))=2;   %exp(rho_vola)

It’s usually better to have the shocks block set to 1 standard deviation and then simulate multiples of this standard deviation. See

Many thanks for replying me!

Is it correct that If I want set to different values for the shock, the change should be made both at the var part and the simulation part. BUT when I try to directly set both of them as 2 or other values (get rid of stander deviation), the results of IRF graphs are still the same as that of 1 stander deviation. I do not understand why this happens.

Thanks a lot.

Then please provide that version.

try4.mod (3.1 KB) try_3.mod (3.1 KB)
The first file that I set the value of uncertainty shock (evola) equal to 2, and the second one is with uncertainty shock of 1 stander deviation. after running these files, the IRF graphs are the same. BUT when I change to order 1 or 2, I could get different IRF results with different shock values. Is there some steps that I am missing or wrong ? Do I write the right code for the third order approximation?

You need to normalize the variance of the shock process to

var evola=1;

But in the simulation, you need to set the shock size, i.e. the actual realization to 2. Also not that some of your IRFs are not the result of the actual shock, but rather the convergence behavior towards the stochastic steady state that results from pruning. This is a tricky issue. See