I am interested in obtaining generalized impulse responses functions and forecast error variance decomposition [as in Pesaran & Shin (1998)].
I found a post on the "old" statalist stating that the general "irf create" command follows this setup, but I am not convinced.
As far as I could find, the documentation does not specify how the IRF's are create when the general irf command is used (without orthogonalizing the shocks).
Old post: http://www.stata.com/statalist/archi.../msg00451.html
Original Paper with methodology: http://83.143.248.39/faculty/nulku/E...n%20(1998).pdf
My particular questions are:
1) Does the "irf create" command indeed utilize the Pesaran & Shin approach where the distribution (variance-covariance matrix) of the shock vector is estimated using historical data?
2) How would I go about calculating the forecast error variance decompostion? (If the answer to question #1 is yes)
Thanks!
I found a post on the "old" statalist stating that the general "irf create" command follows this setup, but I am not convinced.
As far as I could find, the documentation does not specify how the IRF's are create when the general irf command is used (without orthogonalizing the shocks).
Old post: http://www.stata.com/statalist/archi.../msg00451.html
Original Paper with methodology: http://83.143.248.39/faculty/nulku/E...n%20(1998).pdf
My particular questions are:
1) Does the "irf create" command indeed utilize the Pesaran & Shin approach where the distribution (variance-covariance matrix) of the shock vector is estimated using historical data?
2) How would I go about calculating the forecast error variance decompostion? (If the answer to question #1 is yes)
Thanks!