Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: RE : Heteroskedasticity and fixed effects (was: st: RE: Re: Weak instruments)


From   "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: RE : Heteroskedasticity and fixed effects (was: st: RE: Re: Weak instruments)
Date   Thu, 17 Jul 2008 16:28:29 +0100

Martin,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> Maarten buis
> Sent: 17 July 2008 15:55
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: RE : Heteroskedasticity and fixed effects (was: 
> st: RE: Re: Weak instruments)
> 
> --- Gaulé Patrick <patrick.gaule@epfl.ch> wrote:
> > In both cases where is the harm in using robust standard errors and 
> > what's the point to test for heteroskedasticity?
> 
> The harm comes from making people feel more secure about 
> their results than they should be. The point made by Freedman 
> is that it is not going to do them any good, but only the 
> name -robust- suggest that they are somehow protected against 
> all kinds of evils.

You don't mean this literally, right?  For example, if you think a linear model is reasonable and you want to use OLS, but you don't want to rely on more assumptions than you really need, then using OLS + heteroskedastic-robust standard errors (instead of OLS + classical SEs) can't hurt and - if heteroskedasticity is actually present - could help.  This counts as "doing them some good", I think.

Or to repeat Patrick's points 1 and 2, and to make explicit the implicit point 3:

1)  If the model is seriously in error, robustifiying will not help getting better
estimates of the coefficients. Getting standard errors right is irrelevant.

2) If the model is nearly correct, robustifying makes virtually no difference

3) If the model is mostly correct, but the assumption of homoskedasticity is implausible, undesirable, or unsupported, then robustifying helps.

It's not a full prescription for how to go about modelling - of course! - but it's still reasonable guidance, I think.

--Mark

> As Rich remarked earlier, the use of looking for 
> heteroskedasticity (I am a big fan of looking at residuals 
> rather than testing) is that it can be an indication of other 
> problems in your model.
> 
> -- Maarten
> 
> 
> -----------------------------------------
> Maarten L. Buis
> Department of Social Research Methodology Vrije Universiteit 
> Amsterdam Boelelaan 1081
> 1081 HV Amsterdam
> The Netherlands
> 
> visiting address:
> Buitenveldertselaan 3 (Metropolitan), room Z434
> 
> +31 20 5986715
> 
> http://home.fsw.vu.nl/m.buis/
> -----------------------------------------
> 
> 
>       __________________________________________________________
> Not happy with your email address?.
> Get the one you really want - millions of new email addresses 
> available now at Yahoo! http://uk.docs.yahoo.com/ymail/new.html
> *
> *   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/
> 


-- 
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index