Re: st: hetereoskedasticity/serial correlation with RE xtprobit xttobit?

["Arne Risa Hole" <arnehole@gmail.com>]

Just one more point: as has been noted earlier on Statalist (see e.g.
http://www.stata.com/statalist/archive/2007-01/msg01034.html) -robust-
SEs are not necessarily that helpful when the model is non-linear. The
estimated coefficents in logit and other non-linear models will in
genaral be inconsistent in the presence of heteroscedasticity so
fixing the SEs won't help - you only get 'correct' SEs for the 'wrong'
coefficients. See the paper by Freedman (2006) for a discussion.

Arne

David A. Freedman (2006) `On The So-Called "Huber Sandwich Estimator"
and "Robust Standard Errors"', The American Statistician, 60(4), pp. 299-302.


On 09/07/07, Austin Nichols <austinnichols@gmail.com> wrote:
> Jessica <jholmes@middlebury.edu>:
> Thanks for your gracious reply.
>
> The -cluster- option, not -robust-, is what makes your inference
> robust to autocorrelation (both offer robustness to het).  You might
> try a standard fixed effects model, and compare SEs as follows:
>
> g real=exp(lnreal1)
> loc x "male yrsgrad yrsgradsq mar alumnicloserelative med_income
> artsent bankfin commedia comptech consulting education environmental
> govtpp healthcaremed intlang law nonprofitsocialservices
> profservicesbus salesmarketing  ownbusiness mileslt250 reunion soc
> affinity acad devug arts campus sports otheract socsci natsci art
> nongrad loans_z grants_z  finaid_miss djia_pctch campaignyr hock_pc
> rank hock_pc_male hock_pc_yrsgrad hock_pc_sports rank_male
> rank_yrsgrad"
>
> xtreg real `x' if stillug==0, fe i(pidm)
> xtreg real `x' if stillug==0, fe i(pidm) robust
> xtreg real `x' if stillug==0, fe i(pidm) cluster(pidm)
> clogit donate `x' if stillug==0, group(pidm)
> clogit donate `x' if stillug==0, group(pidm) robust
> clogit donate `x' if stillug==0, group(pidm) cluster(pidm)
>
> but see various programs from Baum and/or Schaffer (e.g. -xtivreg2-
> and -ivreg2- and -xttest3- from SSC) for various tests for linear
> regression and correct inference in the presence of AC without het.
> The help file for -ivreg2- is a good starting point. A forthcoming
> article in Stata Journal by Baum and Schaffer and Stillman provides
> even more detail--watch for it...
>
> On 7/9/07, Holmes, Jessica <jholmes@middlebury.edu> wrote:
> > First, I want to thank everyone for their helpful responses to my query.
> > You all have been a fantastic resource!
> >
> > I am still stuck on a test for autocorrelation, but this has been a
> > great help. If the -robust- standard errors correct for both
> > heteroscedasticity and autocorrelation and they are available in Stata
> > 10, perhaps my problem is solved?
> >
> > Thanks all!
> > Jessica
> *
> *   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/
*
*   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/