[R] Chow Test

[Achim Zeileis]

On Wed, 6 Jun 2007, Bernd Stampfl wrote:

>
> Hello R-users!
> I tried to find a package to run a CHOW TEST. As a reference package I found
> the STRUCCHANGE package. Do you know if it works well

If you have concerns regarding the reliability, you can check the
underlying source code and read the accompanying publications (which also
comment on reliability and reproducibility).

> otherwise can you recommend a different one?

If the breakpoint is known, you can easily compute the Chow test by hand:
set up a factor that codes the two regimes and then fit the un-segmented
and segmented regressions:
  fac <- my_time > my_break
  fm0 <- lm(y ~ x1 + x2 ...)
  fm1 <- lm(y ~ fac / (x1 + x2 ...))
  anova(fm0, fm1)
If you want other covariances in the test, you can also use waldtest()
from the "lmtest" package.

If the breakpoint is not known in advance, a supF test (aka supChow) over
all conceivable breakpoints is more appropriate. This is also provided in
"strucchange" along with a rich set of other testing (and dating)
techniques.

Best,
Z

> Thanks, Bernd
> --
> View this message in context: http://www.nabble.com/Chow-Test-tf3878416.html#a10990270
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at 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
> and provide commented, minimal, self-contained, reproducible code.
>
>