The RATS Software Forum

Hi,

When calibrating a simple DSGE model we introduce shocks by setting for example frml z = rho*z{1} (in logs). But this refers to 1.0 shocks. How do we introduce 1 standard deviation shocks?

Thanks

The easiest way to handle that is to rescale the "F" matrix. If the vector of standard deviations of the shocks is D, then using F*%DIAG(D) for the "F" matrix will give the desired results. The alternative is to endogenize the shock by

Thanks a lot.

What value as an initial guess for sigmau would ne plausible then?

Thank you

Wouldn't that depend upon the situation? Since DSGE assumes the shocks in the non-identity equations (which will now be the "U" variable) have unit variance, sigmau would be the desired standard deviation for the shock to the productivity variable.