Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Statistical tests under heteroskedasticity


From   "Airey, David C" <david.airey@vanderbilt.edu>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   Re: st: Statistical tests under heteroskedasticity
Date   Tue, 15 May 2012 23:01:31 -0500

.

The abstract seems a bit strong for the content! Still, in a linear regression model if not a logistic regression model, vce(robust) is a useful option for a well specified model where the variances are not assumed to be constant. 

-Dave

> Indeed, that is exactly the argument presented in:
> Freedman, D.A. (2006). On the So-Called “Huber Sandwich Estimator” and “Robust Standard Errors”. The American Statistician, 60(4),  299-302. 
> http://www.stat.berkeley.edu/~census/mlesan.pdf
> 
> 
> Cam
> > From: david.airey@vanderbilt.edu
> > To: statalist@hsphsun2.harvard.edu
> > Date: Tue, 15 May 2012 10:09:49 -0500
> > Subject: Re: st: Statistical tests under heteroskedasticity 
> > 
> > .
> > 
> > I think it is important (for people like me) to remember to tack on to the end of this correct statement:
> > 
> > (1) provided the model for the expectations (means, for linear regression) is approximately correct, and
> > (2) provided the sample size is large enough
> > 
> > The vce(robust) option is not going to help a poor model, or a model missing important covariates.
> > 
> > If desired, Stata provides more conservative robust options (vce(hc3)) for linear regression with smaller samples.
> > 
> > > <>
> > > On May 15, 2012, at 2:33 AM, Lukas wrote:
> > > 
> > > > I use Stata 11.2 and I want to run several statistical tests to analyse my regression results. In my regression, I use the -robust- option but now I became aware, that heteroskedasticity invalidates my statistical tests such as hypothesis tests as well. 
> > > > 
> > > > Is there a way to run the -ttest- and -testparm- commands under heteroskedasticity without invalidating their results?
> > > 
> > > Calculating robust standard errors -- or more properly, a robust VCE -- 'robustifies' all computations done with that VCE, including t-tests and F-tests. So if the only departure from i.i.d. errors is heteroskedasticity, the 'robust' option makes all subsequent test, testparm, lincom, etc. valid.
> > > 
> > > Kit


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index