Re: st: heteroskedastisity and panel data problem (again)

[Marck Bulter <177316mb@student.eur.nl>]

Brian P. Poi wrote:
> On Mon, 10 Dec 2007 18:58:24 +0100, Marck Bulter 
> <177316mb@student.eur.nl> wrote:
>
>
> > I am working on a panel data project. I would like to test for
> > heteroskedastisity in a unbalanced panel data set. I searched the list,
> > and found the following info:
> >
> > 1. http://www.stata.com/support/faqs/stat/panel.html
> > 2. And I tried the following
> > http://www.stata.com/statalist/archive/2005-03/msg01029.html, to include
> > the nonhetero.
> >
> > Following the suggestions in the links above, I tried the following:
> >
> > xtgls pstrmon price mat age coup pstrmonprev pstrprev intr ivol compl ,
> > igls panels(heteroskedastic)
> > estimates store hetero
> > xtgls pstrmon price mat age coup pstrmonprev pstrprev intr ivol compl
> > estimates store nonhetero
> > local df = e(N_g) - 1
> > lrtest hetero . , df(`df')
> >
> > In the last stage it gives the following error:
> >
> > lrtest hetero . , df(`df')
> > nonhetero does not contain scalar e(ll)
> > r(498);
> -xtgls- is not saving the log-likelihood for the second model, even 
> though -xtgls- without any options produces maximum likelihood 
> estimates.  We will fix this in a future adofile update.
>
> In the meantime, Marck can simply include the "igls" option with his 
> second model:
>
> . xtgls pstrmon price mat age coup pstrmonprev pstrprev intr ivol 
> compl, igls
>
> In this case 'igls' will not affect the estimates, but it will cause 
> -xtgls- to save the log-likelihood, so that -lrtest- will work.
>
>
>   -- Brian Poi
>   -- bpoi@stata.com
>
>
> *
> *   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/
Dear Brian Poi,

The additional option works,

Likelihood-ratio test                                  LR chi2(107)=   
9385.67
(Assumption: nonhetero nested in hetero)               Prob > chi2 =    
0.0000



I assume that the null is homoskedasticity?

regards,
Marck




*
*   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/