impulse response

A forum for questions and answers regarding the CATS cointegration procedure
atarca
Posts: 25
Joined: Thu Dec 08, 2011 8:48 pm

impulse response

Unread post by atarca »

Hi,

Is it possible to use the EXPORT option of @CATS in order to obtain the impulse-response graphs by the below procedures?

@mcvardodraws(model=CatsModel)
@mcgraphirf(model=CatsModel,page=one,varlabels=vlabels,shocklabels=vlabels, center=median,percent=||.025,.975||)

Or is there any other way to obtain them with bands?

Tarkan
TomDoan
Posts: 7779
Joined: Wed Nov 01, 2006 4:36 pm

Re: impulse response

Unread post by TomDoan »

CATS just does point estimates. One problem with doing Monte Carlo integration with cointegrated models is that the "beta" has a non-standard conditional distribution. If you're willing to take beta as given the remainder of the parameters can be handled using standard techniques---see the MONTEVECM.RPF example. Making beta also unknown requires Gibbs sampling.
atarca
Posts: 25
Joined: Thu Dec 08, 2011 8:48 pm

Re: impulse response

Unread post by atarca »

TomDoan wrote:CATS just does point estimates. One problem with doing Monte Carlo integration with cointegrated models is that the "beta" has a non-standard conditional distribution. If you're willing to take beta as given the remainder of the parameters can be handled using standard techniques---see the MONTEVECM.RPF example. Making beta also unknown requires Gibbs sampling.
Thank you very much for your explanations, Tom.
Cheers
Tarkan
Post Reply