[R] gefp() boundaries?

[Achim Zeileis]

On Thu, 6 Oct 2011, bonda wrote:

> Thank you very much for the answer. If I take Poisson model and follow
> "Generalized M-fluctuation tests for parameter instability", A. Zeileis and
> K. Hornik, Statistica Neerlandica (2007) Vol. 61, N. 4, p. 500-501 (section
> 4.3):
>
> data("Boston")
> n <- 506;

I guess that this is data("Boston", package = "MASS")?

> my.X <- as.matrix(cbind(1, Boston["crim"], Boston["age"]));
> my.model <- glm(tax ~ crim + age, family = poisson, data = Boston);

Applying a Poisson model for count data to a tax rate seems to be very 
awkward.

> my.psi <- estfun(my.model);
> my.mu <- fitted(my.model);
> J <- sum(my.mu*my.X%*%t(my.X))/n;

This should be the covariance matrix, not a scalar.

> my.process <- apply(as.matrix(my.psi), 2, cumsum)/sqrt(J*n);
>
> gprocess <- gefp(tax ~ crim + age, family = poisson, data=Boston);
>
> then my.process and gprocess$process have to be the same?

No. Even if you would have computed the J that you would have been 
interested in (and computed its root correctly), then gefp() still uses a 
different covariance matrix estimator by default.

See ?gefp and the 2006 CSDA paper for more details. Looking at the actual 
code may also help.


> Best regards,
> Julia
>
> --
> View this message in context: http://r.789695.n4.nabble.com/gefp-boundaries-tp3872529p3878886.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>