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

Re: st: RE: Robust Standard Errors


From   "Mark Schaffer" <M.E.Schaffer@hw.ac.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: Robust Standard Errors
Date   Fri, 21 May 2004 09:42:53 +0100

Larry,

Subject:        	st: RE: Robust Standard Errors
Date sent:      	Thu, 20 May 2004 16:11:16 -0700
From:           	"Chavis, Larry Wilson" <chavis_larry@gsb.stanford.edu>
To:             	<statalist@hsphsun2.harvard.edu>
Send reply to:  	statalist@hsphsun2.harvard.edu

> Thanks Richard for pointing out that FAQ. I have not been using the
> cluster option.  So it seems according to the FAQ that it is possible
> that the decrease is due to "something odd going on between the
> residuals and the x's" (or I could be extremely lucky and it is due to
> random variation!)  So I should probably look at the correlation
> between my X's and the residuals.  Maybe it is because most of my
> variables and fixed effect dummies have lots of 0 values.

Do you have fixed effects dummies in which you have a single 1 and 
the rest 0s ("singleton dummies")?  About a year ago Vince Wiggins 
wrote a very useful note about the effect of these on robust var-cov 
estimates:

http://www.stata.com/statalist/archive/2003-06/msg00646.html

in which he concluded that the reported standard errors are still 
useable even though the var-cov matrix isn't of full rank.  Might be 
totally unrelated to your problem, but you never know.

--Mark

>  Thus when
> the "sandwich" is being summed the effects of some residuals are
> diminished. This sounds plausible to me but I'll have to work through
> the math.
> 
> Thanks again,
> Larry
> 
> -----Original Message----- 
> From: owner-statalist@hsphsun2.harvard.edu on behalf of Richard
> Williams Sent: Thu 5/20/2004 3:45 PM To:
> statalist@hsphsun2.harvard.edu Cc: Subject: Re: Robust Standard Errors
> 
> 
> 
>  At 03:22 PM 5/20/2004 -0700, Chavis, Larry Wilson wrote:
>  >My robust standard error estimates where much less than the original
>  >estimates.  For instance the coefficient T was estimated as -.015
>  and the >original standard error was .0541, while the robust standard
>  error is >.0034.  This is more than a 10 fold decrease, while what I
>  could find in >the literature suggested that robust standard errors
>  are usually larger >than the original standard errors.
> 
>  Are you using a cluster option?  This FAQ might or might not be
>  relevant:
> 
>  http://www.stata.com/support/faqs/stat/cluster.html
> 
> 
>  *
>  *   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/
> 
> 

Prof. Mark E. Schaffer
Director
Centre for Economic Reform and Transformation
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS  UK
44-131-451-3494 direct
44-131-451-3008 fax
44-131-451-3485 CERT administrator
http://www.som.hw.ac.uk/cert

*
*   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/



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