Artur T. schrieb: > Allin Cottrell schrieb: >> On Tue, 23 Mar 2010, Artur T. wrote: >> >> >>> I am just doing an empirical class project on the PPP-UIP relationship. >>> After setting up the long-run relationships and imposing >>> (over-)identifying restrictions on beta, I want to conduct an >>> impulse-response analysis on the cointegrating relationships in order to >>> check how long it needs the long-run relationships to get back to its >>> equilibrium. >>> >>> Is this possible in gretl? Or has anybody a script for this or a similar >>> problem? >>> >> Use the --full option to the restrict command and you'll get a >> complete VECM model from which you can generate impulse responses. >> You won't be able to get a bootstrapped confidence interval with >> beta restricted, but you can get a point estimate. >> >> open denmark >> vecm 2 1 LRM LRY IBO IDE >> v2 <- restrict >> b[1] + 1 = 0 >> end restrict --full >> >> Then open v2 in a GUI window and get impulse responses via the >> Graph menu. >> >> Allin Cottrell >> _______________________________________________ >> Gretl-users mailing list >> Gretl-users(a)lists.wfu.edu >> http://lists.wfu.edu/mailman/listinfo/gretl-users >> > Hey Allin, > > Sorry, I think I did not explain it correctly. > > I know that I can restrict the model and run the impulse response > analysis. What I was actually asking for is the so called persistence > profile approach proposed by Pesaran/Shin ("Cointegration and speed of > convergence to equilibrium" 1996 in Journal of Econometrics). > > > There they shock the cointegrating relations directly and not the > individual variables (as I understood it). >
Of course it's doable in gretl :-) However, that doesn't mean much anymore since gretl's scripting and matrix language by now has become so flexible that it's equivalent to saying "why don't you program it yourself?" Seriously, though, IIRC (no guarantees) the point of the persistence profiles is that you hit the system with a sort of "average" innovation, and then you sit back and watch as the equilibrium deviations converge to zero again. So I wouldn't call it shocking the CI relations directly, but rather "shocking all the involved variables simultaneously". Thus the question is how to choose this "average" system-wide innovation, and that's proposed in the article. (I don't know right now.) I'm not sure if it's worth implementing this for a class project, but hey I'm not your instructor. There must be some Ox or Rats code floating around in the web for this. good luck, sven