package(vars) and generalized impulse response functions

[Vimal B]

Hi,

I was wondering whether the package(vars) can estimate generalized impulse
response functions for VARs / VECMs.

The general option on package(vars) available is:-

suppose I want to give a shock to 'a' and observe the response of 'b', then

irf(x, impulse=a, response=b, n.ahead=10,
boot=TRUE,ci=0.95,runs=100,seed=NULL...)

The vignettes and manuals says that the default is 'ortho=TRUE' (and the
other option is cumulative=NULL).

Or is there any other package one would suggest?

Thanks,

Regards,
Vimal

        [[alternative HTML version deleted]]

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.

[Matthieu Stigler]

Hi Vimal

Do you refer to generalized impulse response functions such as in Koop
et al (1996) and Pesaran and Shin (1998) where focus is on
investigating the distribution of any shock at n, taking into account
further  shocks at n+1, n+2...?

If yes, this is not to my knowledge available in R. Package vars does
not provide it at least, options ortho just asks whether a Choleski
decomposition (with ordering as in the input matrix) should be done,
and cumulative whether effects of shocks should be added.

Note btw that even if very interesting, it seems that GIRF did not
have big sucess. I personally saw only a few paper using it, and I'm
not aware of any textbook presenting it, while they all discuss in
details usual IRF.

Best

Matthieu



2009/11/4 Vimal B <[hidden email]>:

> Hi,
>
> I was wondering whether the package(vars) can estimate generalized impulse
> response functions for VARs / VECMs.
>
> The general option on package(vars) available is:-
>
> suppose I want to give a shock to 'a' and observe the response of 'b', then
>
> irf(x, impulse=a, response=b, n.ahead=10,
> boot=TRUE,ci=0.95,runs=100,seed=NULL...)
>
> The vignettes and manuals says that the default is 'ortho=TRUE' (and the
> other option is cumulative=NULL).
>
> Or is there any other package one would suggest?
>
> Thanks,
>
> Regards,
> Vimal
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only.
> -- If you want to post, subscribe first.
>

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.

[Vimal B]

Thanks Matthieu. Yes, there were some ordering issues in my VEC model and I
had wanted to estimate a girf so as to avoid using choleski decomposition.

While I see a reason why girfs may not be so popular (as more often in
economics *a priori* reasoning should allow for a reasoned out ordering of
the matrix), when something is completely data driven, *a priori* reasoning
might completely fail in giving a good fit. Hence the attempt at girfs.

Thanks again for the clarification.

Regards,
Vimal

विमल | বিমল  | વિમલ  | ವಿಮಲ್  | വിമല്‍  | ਵਿਮਲ | விமல் | తెలుగు  | ومال

Mobile: [phone]
Email: [hidden email]
Blog: http://vimsaa.blogspot.com



2009/11/4 Matthieu Stigler <[hidden email]>

> Hi Vimal
>
> Do you refer to generalized impulse response functions such as in Koop
> et al (1996) and Pesaran and Shin (1998) where focus is on
> investigating the distribution of any shock at n, taking into account
> further  shocks at n+1, n+2...?
>
> If yes, this is not to my knowledge available in R. Package vars does
> not provide it at least, options ortho just asks whether a Choleski
> decomposition (with ordering as in the input matrix) should be done,
> and cumulative whether effects of shocks should be added.
>
> Note btw that even if very interesting, it seems that GIRF did not
> have big sucess. I personally saw only a few paper using it, and I'm
> not aware of any textbook presenting it, while they all discuss in
> details usual IRF.
>
> Best
>
> Matthieu
>
>
>
> 2009/11/4 Vimal B <[hidden email]>:
> > Hi,
> >
> > I was wondering whether the package(vars) can estimate generalized
> impulse
> > response functions for VARs / VECMs.
> >
> > The general option on package(vars) available is:-
> >
> > suppose I want to give a shock to 'a' and observe the response of 'b',
> then
> >
> > irf(x, impulse=a, response=b, n.ahead=10,
> > boot=TRUE,ci=0.95,runs=100,seed=NULL...)
> >
> > The vignettes and manuals says that the default is 'ortho=TRUE' (and the
> > other option is cumulative=NULL).
> >
> > Or is there any other package one would suggest?
> >
> > Thanks,
> >
> > Regards,
> > Vimal
> >
> >        [[alternative HTML version deleted]]
> >
> > _______________________________________________
> > [hidden email] mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> > -- Subscriber-posting only.
> > -- If you want to post, subscribe first.
> >
>

        [[alternative HTML version deleted]]


_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.