Greene does note (page 509, 4th ed) that if the disturbances in the two groups
are normally distributed, then the Goldfeld-Quandt statistic is exactly
distributed as an F. If not, then the F distribution is only approximate and an
alternative with known large sample properties - such as White's test - may be
preferred.
Scott
----- Original Message -----
From: "Nick Cox" <[email protected]>
To: <[email protected]>
Sent: Tuesday, February 03, 2004 5:58 PM
Subject: st: RE: Goldfeldt-Quant versus -hettest-
> A Google on "Goldfeld-Quandt Stata" [sic] throws
> up various recipes to try at home.
>
> Not your question, but I'll put in a plug for
> a user-written plot program called -rdplot-.
>
> It's no doubt eccentric, but given a choice
> I'd always choose a purpose-driven graph over
> a test statistic with P-value. Of course, you
> can have both.
>
> Nick
> [email protected]
>
> Richard Williams
>
> > Various textbooks discuss the use of the Goldfeldt-Quant test for
> > heteroskedasticity. For those of you who are familiar with
> > it, it is a
> > little clunky, and requires some arbitrary decisions on how
> > to split the
> > data. Stata, on the other hand, has the nice easy to use -hettest-
> > command, which does the Breusch-Pagan / Cook-Weisberg test for
> > heteroskedasticity, and Stata also offers some other tests.
> > Is there any
> > particular reason I would still prefer GQ given that
> > -hettest- and other
> > options are available? i.e. are there situations in which GQ is more
> > appropriate or will pick up problems that other tests will
> > not?
>
> *
> * 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/
*
* 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/
