Can you supply a literature reference where the theory
is worked out for these tests with pweights and/or
they are applied in a way that makes the theoretical
extension clear?
On the other hand, it could be that you are the first
person to work out the theory, which explains why the
code doesn't yet exist.
More positively, the best way to look at these
matters is often graphical. The commands in the -modeldiag-
package are not so fussy. Correspondingly, there is
onus on the user to take responsibility for what she
is doing. -search modeldiag- for location.
Nick
[email protected]
Eva Gottschalk
> Well, I'd like to perform the Cook Weisberg test for
> heteroskedasticity,
> which can be done by using the
> - hettest
> command after an unweighted regression.
> I'd also like to discover influential cases with cooks d, but the
> command
> - predict X, cooksd
> also doesn't work after a pweighted regression.
Friedrich Huebler <[email protected]>:
> > Which commands are you referring to? Please be more specific. The
> > combined subject index in my Stata manual has the following entries
> > under "regression diagnostics":
> >
> > - ladder
> > - logistic
> > - predict
> > - regress postestimation
> > - regress postestimation time series
Eva Gottschalk
> > > I am calculating an OLS regression with surveydata that is
> > weighted
> > > for the overrepresentation of east germany (that I control for
> > with
> > > a dummy in my reg). I'd like to test my model by using regression
> > > diagnistics, but most of the commands won't work with weighted
> > data.
> > > Why is that? So how can I test the Gauss-Markov assumptions?
*
* 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/
