I have estimated my model and obtained the Bayesian IRFs for positive shocks hitting the economy, e.g.
a_z = ZRHO_Z*a_z(-1) + eps_z.
However, I was now thinking that it might be more interesting to show the IRFs for some negative shocks, i.e.
a_z = ZRHO_Z*a_z(-1) - eps_z.
What would be the fastest way to this, as I don’t want to re-estimate the model with 2,000,000 draws? Can I just multiply oo_.PosteriorIRF.dsge.Mean.xxxx, oo_.PosteriorIRF.dsge.HPDinf.xxxx and oo_.PosteriorIRF.dsge.HPDsup.xxxx by -1 to obtain the opposite response?
Simply put a minus symbol before the innovation in the equation of the autoregressive process. Note that obviously this would not change your estimation results (except for the smoothed innovation you redefine by changing the sign).
Many thanks for getting back to me. Would this mean I’d have to re-estimate the model again with the minus sign? I wanted to show the estimated mean IRFs for a negative shock. Sorry if I have misinterpreted your answer.
Sorry, I misread your post. The model is linear so the IRFs must be symetric (with respect to the abscissa when you change the sign). So I suppose it is safe to simply put a minus sign before the curves characterizing the posterior IRFs.