Impulse response function

[Thisisusername]

Hi all,

1. Can dynare simulate the data rather than just compute the IRF? Say I have a series of shock e_t, and ask dynare to simulate the corresponding variables C_t, Y_t, etc.

2. Another question is does IRF affected by the distribution of the shock? Suppose the process z = rho*z(-1)+e_t, where e_t have different distributions but with the same std derivation. According to the definiton, the IRFs should be the same, but then I found it a bit strange.

Thanks a lot!

[h0ps]

1) Have you tried the rplot function? If your variable names are x, pi and r respectively you can write (at the end)

Code: Select all

rplot x;
rplot pi;
rplot r;

[jpfeifer]

1) If you use the periods-option the rplot command is correct
2) For perturbation solutions of order n, only the the first n+1 moments matter. This can be seen in e.g. Schmitt-Grohe/Uribe (2004). Due to approximating the system using a Taylor approximation, only the first n+1 moments appear. All higher order moments are neglected. Thus, at order=1 only mean and variance of the shocks matter. This corresponds to a normal distribution that is fully characterized by those moments.