Date: Thu, 6 Mar 2003 09:18:18 +0100
Reply-To: Asesoría Bioestadística
<bioestadistica@eresmas.net>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Asesoría Bioestadística
<bioestadistica@eresmas.net>
Subject: Re: Test of Homoscedasticity
Content-Type: text/plain; charset=iso-8859-1
Hi
> The first test to detect heteroskedasticity is usually a plot of the
> residuals vs. the predicted. I use a plot of studentized residuals
> (*SRESID) on the y axis, and the predicted values (*PRED) on the x axis. Do
> this with the /SCATTERPLOT command after the /METHOD command in syntax, or
> in the PLOTS section of the regression dialogue box (I think *PRED is only
> available in syntax). For example:
>
> /METHOD=ENTER ind_var
> /scatterplot (*SRESID *PRED).
>
> The plot should have what is referred to as a random or "cloud" appearance.
> A fan shape like > or < indicates heteroscedasticity. There are several
> statistical test for heteroscedasticity: Goldfeld-Quandt and White's test
> are two that I use. McClendon does a good job of discussing White's using
> SPSS syntax, and Wilkinson, Blank and Gruber do an excellent job of
> discussion Goldfeld-Quandt (though this is for SYSTAT).
>
> So if you have the visual appearance of heteroscedasticity and you want to
> do a statistical test, then visit Raynald for some excellent resources:
> http://pages.infinit.net/rlevesqu/spss.htm#Heteroscedasticity (the article
> by Gwilym Pryce is currently not avilable)
I checked today an it IS available. I wrote 2 macros based on this paper:
http://pages.infinit.net/rlevesqu/Syntax/RegressionRepeatedMeasure/Breusch-PaganAndKoenkerTest.txt
Performs those 2 tests of heteroscedasticity and a plot similar to the one you
mentioned before (ZRESID and ZPRED instead of SRESID and PRED)
http://pages.infinit.net/rlevesqu/Syntax/RegressionRepeatedMeasure/WhiteTestStatisticsAndSignificance.txt
Performes another test (White)
Regards
Marta Garcia-Granero
> ----- Original Message -----
> From: "Dieudonne Anderson Loundou" <loundou@marseille.inserm.fr>
> Newsgroups: bit.listserv.spssx-l
> To: <SPSSX-L@LISTSERV.UGA.EDU>
> Sent: Wednesday, March 05, 2003 9:13 PM
> Subject: Test of Homoscedasticity
>
> > Dear All,
> >
> > I want to test homoscedasticity of residuals for my regression linear
> model in
> > spss. Can anybody know how can I do this ?
> >
> > Thanks,
> >
> > ---------------------------------------------
> > Ce messge a été envoyé par SMM Ste Marguerite Mail.
> > http://www.marseille.inserm.fr
> >
