# Re: st: how to tell if there is heteroscedasticity, thanks!

 From "Joanne Marshall" <[email protected]> To [email protected] Subject Re: st: how to tell if there is heteroscedasticity, thanks! Date Sat, 13 Jan 2007 17:15:31 +0000

thank you for the help, maarten.
do we look at the p values for the im test and bp test to determine hetero, and if we do, what p values are we looking for in order to determine that?

and also how can we undertake the type 1 error 5% test, i think i have seen this before, but never knew how to work it out!

thank you very much

regards
jo

From: Maarten buis <[email protected]>
To: [email protected]
Subject: Re: st: how to tell if there is heteroscedasticity, thanks!
Date: Sat, 13 Jan 2007 16:24:56 +0000 (GMT)

Jo:
According to the im test you do not reject the null hypothesis of
homoscedasticity, and according to the hettest you do. In general I am
not a big fan of testing model assumptions prior to doing the tests you
are really interested in. What you do in that case is perform a
sequence of tests, each with a probability of a type I error of 5%, so
the total probability of a type I error is not 5%. Furthermore some of
these tests can be not powerful enough to detect problems, thus giving
you a false sense of security. For diagnozing my model, I would go for
graphs, e.g. -rvfplot-.
Hope this helps,
Maarten

--- Joanne Marshall <[email protected]> wrote:
> i am trying to run a heteroskedasticity test for my regression here,
> using command
>
> estat imtest (IM TEST)
> estat hettest ( BREUSCH PAGAN)
>
> and have got the following result:
>
> Cameron & Trivedi's decomposition of IM-test
>
> ---------------------------------------------------
> Source | chi2 df p
> ---------------------+-----------------------------
> Heteroskedasticity | 167.27 155 0.2368
> Skewness | 45.45 18 0.0004
> Kurtosis | 9.72 1 0.0018
> ---------------------+-----------------------------
> Total | 222.44 174 0.0077
> ---------------------------------------------------
>
> .. estat hettest
>
> Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
> Ho: Constant variance
> Variables: fitted values of wage
>
> chi2(1) = 191.17
> Prob > chi2 = 0.0000
>
>
> with the above information, how can i tell if my model is hetero or
> not then?

-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------

___________________________________________________________
*
* 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/
_________________________________________________________________