Antwort an: [email protected]
> Tobias Rehbock
>
> > I have a problem using the granger causality test in Stata 6.0.
> > I've testet two time series named DD and DUS using
> > the command-line:
> >
> > granger DD DUS, lags(2)
> >
> > and, alternativly
> >
> > granger DD DUS, lags(3)
> >
> > For the first case, I received F=5,98 and p=0,0032 and for the
> > second case F=5,45 and p=0,0014. But there must
> > be a mistake, because it is not possible, that the second F-value
> > ist smaller than the first while the second p-value
> > also ist smaller than the first.
>
> Not quite. Remember that the F distribution depends on two
> separate numbers of degrees of freedom. If the numerator
> and denominator df differ, this is possible.
Oh, you are right! Thank you very much.
>
> > Then I tried the same test with another software (EViews4) and
> > the results were amazing: there were not only other
> > F- and p-Values, but the found causalities were the other way round.
> >
> > My questions are: Where is my mistake? And: Is it possible, that
> > granger.ado doesn't work well?
> >
>
> -granger- was published in STB-51. One substantial bug in it,
> if I read the code correctly, is
> that while the syntax permits -if- and -in-, the regressions
> within are always performed for all possible observations.
>
> I guess this doesn't explain your problem, but it needs
> fixing.
>
> I suggest that you send your data and the precise EViews
> output to the program author.
I also think, this would be the correct way and I've done this yesterday. In the meantime I got another answer from
the statalist with a new granger-test (granger2.ado by Joly Patrick) and it seems to be the right one. I think, you red
it, too.
Thank you very much for your help! Tobias
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
