Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: a question about areg with robust versus areg with robust and cluster options

From (Vince Wiggins, StataCorp)
Subject   Re: st: a question about areg with robust versus areg with robust and cluster options
Date   Sun, 06 Nov 2005 23:32:25 -0600

Maria Gesheva <> asks,

> When looking through the stata help I came over the following FAQ:
> ''Why is it dangerous to use the robust cluster() option on areg?
> (areg estimates the same fixed-effects model as xtreg, fe)'' but the
> content had been removed.

The FAQ should be clearer.  It is not just that the content has been
withdrawn, but that the FAQ itself has been withdrawn (and has been for some
time).  Despite what the title of the FAQ implies, there is nothing wrong with
using the robust VCE with -areg-, or for that matter, with Stata's other
fixed-effects estimator -- -xtreg, fe-.  In fact, as of the release of 
Stata 9, -xtreg, fe- supports both the -robust- and -cluster()- options.

The reason the FAQ cannot be completely removed from the website is that old
versions of Stata still contain references to the FAQ in the index files
referenced by -search-.  The FAQ can also currently be found when searching
the FAQs on the Stata website, but that was not intended and will be fixed.

Maria Gesheva <> goes on to ask,

> I would like to ask the same question plus a related one: when is it
> appropriate to use areg only with the robust option and when areg
> with both robust and cluster options. [...]

As with all other estimators, that depends on what Maria believes about her
disturbances.  If she is merely concerned about heteroskedasticity, then the
linearization VCE provided by option -robust- will suffice.  If Maria does not
believe her disturbances are independent, but rather exhibit within-group
correlation, then she will need the clustered version of the linearization
VCE, i.e., the -cluster()- option.  As always, remember that the asymptotics
of the cluster-robust VCE require a large number of groups.

-- Vince

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index