I have a DSGE model with multiple shocks, two of which (called v_u and v_d) I allow to be correlated. When I do Bayesian IRFs, the responses if the endogenous variable to v_u incorporate effects of v_d as well because of the correlation. Is there some way to suppress that, so I can see the responses of the endogenous variables to v_u, holding v_d =0? I tried the irf_shocks = option, but that seems only to change what is reported, not the calculations. Thanks.
That is not easily feasible. You need to take a stance to which variable to assign the correlation. Dynare does that by using a Cholesky decomposition. In your case, v_d seems to be ordered first and v_u second. If you were to alter the declaration order, that would switch.
Thanks, that seemed to work. I had declared v_u before v_d, but when I switched the order and declared v_d first, the responses to v_u seem to be conditioned on v_d = 0, which is what I wanted.