Jay,
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> Verkuilen, Jay
> Sent: 17 July 2008 21:33
> To: [email protected]
> Subject: RE: RE : Heteroskedasticity and fixed effects (was:
> st: RE: Re: Weak instruments)
>
> Schaffer, Mark E wrote:
>
> >The general motivation behind the advice seems reasonable -
> why assume
> something that you don't have to assume, if making the
> >assumption is potentially costly and relaxing the assumption
> is nearly costless? Of course, this doesn't excuse us from
> the >responsibility of giving our model "a good long look",
> but we should do that anyway.
>
> This assumes that the solution -is- costless. It may well not
> be. As the Carroll et al paper I sent a link to in an earlier
> email notes
> (http://www.stat.tamu.edu/ftp/pub/rjcarroll/sandwich.pdf),
> most of the discussion of robust VCE is in asymptotic terms.
> With small to medium samples and different models your
> mileage may most definitely vary.
Quite true - that wass one of the reasons for the "if". I should have
said "IF".
Another reason is that giving up on modelling the heteroskedasticity may
be giving up too early. If FGLS really is feasible, then just using
robust standard errors is not costless - you're paying a price in terms
of the efficiency gains possible with FGLS. (Which is why I don't
entirely agree with the Stock & Watson quote that Patrick posted.)
> I generally think robust VCE is a good thing to have, mind
> you, but something that should not be blindly applied to
> cover up a crappy, misspecified model.
Indeed!
--Mark
>
>
> *
> * 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/
