Date: Tue, 1 Oct 2002 17:09:31 +0200
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Isabell Ottfried <firstname.lastname@example.org>
Subject: the use of breusch-pagan test and the white standard errors
I already got very helpful input from Marta how to use the syntax of the Breusch-Pagan test. Thanks Marta!
Now, I have a more general question about the use of a) the Breusch-Pagan test and b) the White standard errors, respectively:
a) Gwilym Pryce writes in her paper about heteroscedasticity that Z’s can be all or some original regressors which ONE THINK CAUSE the heterscedasticity. However, how do I know which regressors causes hetero? Is it a common approach in a moderated regression model to integrate all regressors as well as all interaction effects in the test? More specifically, I have a moderated regression model with 10 predictors and 9 interaction effects, do I test only the 10 predictors or additionally the 9 interaction effects? I found nothing about this issue in the literature. May be somebody has made similar experience and can be a helping hand!? References are also very welcome that tackles this issue.
b) Gwilym Pryce offers for a “normal” regression model a syntax for the White standard errors if one detect problems with hetero. However, can I use this approach in analogy for the moderated regression analysis?
Your help is sincerely appreciated. Thousand thanks in advance.
Neue Freunde finden!
Spielend in Kontakt kommen und neue Freunde finden in der freenet.de Community!
Jetzt durchstarten unter: http://www.freenet.de/tipp/community/