GSA of IRFs with a higher order approximation


I am interested in a global sensitivity analysis toolbox. Particularly, I was wondering if it is possible to investigate IRFs sensitivity to parameters for the model with a higher order of approximation?

Since in the higher-orders IRFs are typically defined to be in relative terms of stochastic steady-state, I guess that GSA does not produce stochastic simulation to approximate IRFs? If that is the case, is it possible to somehow include stochastic simulation to generate IRFs (maybe avoid irf_calibration block?) and conduct further analysis using GSA?

Thanks a lot for the help!

Update: I estimated what parameters ensure that the impulse IRF of various second moments is positive or negative. the result is that there are not any parameter values, suggesting that the impulse stays at zero. Thus, I guess irf_calibration uses only the first order of approximation. In that case, my question is whether I can introduce either higher-order IRFs or results from a stochastic simulation that can be used in GSA toolbox?

Can you provide me with the file?
@rattoma may know more.

Thanks for the response but sorry for the silence on my side. I somewhat thought the post died.:slight_smile:

As for the example, I think my case works for any model with 3rd perturbation. So one could just look at a seminal case of Basu and Bundick 2017, as produced by your effort. As for the question, one could ask which parameters are important to generate a positive response of VIX, s.t.

cond_var_R_E, eps_a, +

dynare_sensitivity currently only works at first order. In the future, you should get a warning: dynare_sensitivity.m: provide warning if order is reset (!2009) · Merge requests · Dynare / dynare · GitLab

Thank you

I also opened a ticket for adding that feature: Expand *_calibration to higher order (#1847) · Issues · Dynare / dynare · GitLab