[Gretl-users] Re: R: Query about FEVDs for BVAR
Am 31.07.2025 um 13:23 schrieb GIULIA CHIARA:
Thank you for your quick response.
If you could, at some point, when it's convenient for you, provide me
with an example, it would be extremely helpful since I'm not very
familiar with coding.
Thank you in advance for your help. I wish you a good day.
Luca and I will definitely add the FEVD calculation to the BVAR package
quite soon. In the meantime and for the record, here's the current
internal structure of the results (since this isn't fully spelt out in
the help document):
- As the help mentions, the output bundle (with user-chosen name)
contains a sub-bundle called "s_irfs".
- This s_irfs sub-bundle holds a bunch of matrix members, each of which
contains one of the IRFs.
- The names of these matrix-es are of the pattern:
summ_<originname>_tow_<targetname>,
where <originname> refers to the variable where the shock comes from,*
and <targetname> of course is the variable that reacts.
- The number of rows of each matrix is given by the irf horizon that was
chosen, starting from 0 (=contemporaneous impact).
- There are five columns with posterior statistics: mean, std.dev., 5%
quintile, median, 95% quintile.
So for example, if your results bundle is called "res", you can access
'res.s_irfs.summ_interest_tow_inflation[,4]' to get the posterior median
of the structural IRF from the equation of the variable "interest" to
the variable "inflation", as a column vector up to the specified horizon.
I'm not saying that we guarantee this internal structure to stay like
that forever, but that's the current state of affairs.
cheers
sven
---
*: (of course, for structural IRFs the shock source is not the same
thing as a variable, so the variable name is just used to identify the
equation which is considered)