[R] Partial structural Change in STRUCCHANGE PACKAGE
Hi, I am using the Strucchange package in R to test for structural change in regression coeffcient. Given a model y = b0 + b1*X + b2*Z, the Fstats test whether there is a change in both b1 and b2 over a time period. Is there any way where I can restrict the test to hold b2 constant and test for break in only b1? That is, instead of a pure structural change, could I test for partial structural change in only one of the coefficient estimate? Warm Regards, Yen __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Partial structural Change in STRUCCHANGE PACKAGE
Hello, one way could be to compute the residuals of the regression y=b0 + b2*Z, let's call them U, and then test the structural change on the model U=c0+c1*X. Maybe there's a better way. Romain. Le 18.02.2005 20:03, Yen H., Tong a écrit : Hi, I am using the Strucchange package in R to test for structural change in regression coeffcient. Given a model y = b0 + b1*X + b2*Z, the Fstats test whether there is a change in both b1 and b2 over a time period. Is there any way where I can restrict the test to hold b2 constant and test for break in only b1? That is, instead of a pure structural change, could I test for partial structural change in only one of the coefficient estimate? Warm Regards, Yen -- Romain FRANCOIS : [EMAIL PROTECTED] page web : http://addictedtor.free.fr/ (en construction) 06 18 39 14 69 / 01 46 80 65 60 ___ Etudiant en 3eme année Institut de Statistique de l'Université de Paris (ISUP) Filière Industrie et Services http://www.isup.cicrp.jussieu.fr/ __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Partial structural Change in STRUCCHANGE PACKAGE
On Fri, 18 Feb 2005 11:03:13 -0800 (PST) Yen H., Tong wrote: Hi, I am using the Strucchange package in R to test for structural change in regression coeffcient. Given a model y = b0 + b1*X + b2*Z, the Fstats test whether there is a change in both b1 and b2 over a time period. Is there any way where I can restrict the test to hold b2 constant and test for break in only b1? That is, instead of a pure structural change, could I test for partial structural change in only one of the coefficient estimate? Not in Fstats(). Fstats() computes a sequence of Wald statistics such that in the case of partial changes I would have to compute partially segmented models which I haven't implemented because I haven't seen a real application where a partial change approach is intuitive. (How do you know that the other coefficients are really stable?) However, you can compute a sequence of LM statistics using gefp() and selecting the coefficients you want to assess via the parm argument. The trick is that you only estimate the model once under the null and thus avoid the problem of partially segmented models. But I haven't provided an efpFunctional for the supLM test, yet. But if you want to use only a certain trimming parameter, that is very easy to set up. Contact me off-list, if you want to do so. Z Warm Regards, Yen __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Partial structural Change in STRUCCHANGE PACKAGE
On Fri, 18 Feb 2005 20:14:23 +0100 Romain Francois wrote: Hello, one way could be to compute the residuals of the regression y=b0 + b2*Z, let's call them U, and then test the structural change on the model U=c0+c1*X. The approach of Andrews (1993, Econometrica) would be preferrable where you explicitely estimate partially segmented models with a breakpoint shifted across the sample period. This woul usually lead to a varying coefficient of Z as well. But as I explained in my previous mail, this is not currently implemented in strucchange. Maybe there's a better way. Personally, if I really believed in a partial change model, I would compute a statistic (supLM, Cramer-von Mises, maximum, etc.) from the corresponding cumulative score process(es) computed by gefp(). Z Romain. Le 18.02.2005 20:03, Yen H., Tong a écrit : Hi, I am using the Strucchange package in R to test for structural change in regression coeffcient. Given a model y = b0 + b1*X + b2*Z, the Fstats test whether there is a change in both b1 and b2 over a time period. Is there any way where I can restrict the test to hold b2 constant and test for break in only b1? That is, instead of a pure structural change, could I test for partial structural change in only one of the coefficient estimate? Warm Regards, Yen -- Romain FRANCOIS : [EMAIL PROTECTED] page web : http://addictedtor.free.fr/ (en construction) 06 18 39 14 69 / 01 46 80 65 60 ___ Etudiant en 3eme année Institut de Statistique de l'Université de Paris (ISUP) Filière Industrie et Services http://www.isup.cicrp.jussieu.fr/ __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Partial structural Change in STRUCCHANGE PACKAGE
Hi, Thanks for all the thoughtful replies. Let me think on this further. Warm Regards, Yen On Fri, 18 Feb 2005, Achim Zeileis wrote: On Fri, 18 Feb 2005 20:14:23 +0100 Romain Francois wrote: Hello, one way could be to compute the residuals of the regression y=b0 + b2*Z, let's call them U, and then test the structural change on the model U=c0+c1*X. The approach of Andrews (1993, Econometrica) would be preferrable where you explicitely estimate partially segmented models with a breakpoint shifted across the sample period. This woul usually lead to a varying coefficient of Z as well. But as I explained in my previous mail, this is not currently implemented in strucchange. Maybe there's a better way. Personally, if I really believed in a partial change model, I would compute a statistic (supLM, Cramer-von Mises, maximum, etc.) from the corresponding cumulative score process(es) computed by gefp(). Z Romain. Le 18.02.2005 20:03, Yen H., Tong a écrit : Hi, I am using the Strucchange package in R to test for structural change in regression coeffcient. Given a model y = b0 + b1*X + b2*Z, the Fstats test whether there is a change in both b1 and b2 over a time period. Is there any way where I can restrict the test to hold b2 constant and test for break in only b1? That is, instead of a pure structural change, could I test for partial structural change in only one of the coefficient estimate? Warm Regards, Yen -- Romain FRANCOIS : [EMAIL PROTECTED] page web : http://addictedtor.free.fr/ (en construction) 06 18 39 14 69 / 01 46 80 65 60 ___ Etudiant en 3eme année Institut de Statistique de l'Université de Paris (ISUP) Filière Industrie et Services http://www.isup.cicrp.jussieu.fr/ __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html