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

Re: st: Clustering vs. Indicator variables in a Logit


From   "David" <dj_knight01@yahoo.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Clustering vs. Indicator variables in a Logit
Date   Mon, 08 Nov 2004 20:00:32 -0000

Thank you, Mark and David, for your quick replies.

-David Bernstein
 

--- In statalist@yahoogroups.com, "Mark Schaffer" 
<M.E.Schaffer@h...> wrote:
> David,
> 
> Subject:        	st: Clustering vs. Indicator variables in a 
Logit
> To:             	statalist@h...
> From:           	DBernstein@l...
> Date sent:      	Mon, 8 Nov 2004 10:59:45 -0500
> Send reply to:  	statalist@h...
> 
> > Hello All,
> > 
> > I understand that clustering specifies that observations be 
independent
> > across groups while allowing for changes in variance within a 
group.  My
> > question is how is this different than controlling for a group 
with  an
> > indicator (0/1) variables.
> > 
> > For example, If my data contains patient data for 12 hospitals 
and my LHS
> > variable is (0/1) for recovery and my RHS variables include 
characteristic
> > variables, treatment type, etc.  What is the difference between 
clustering
> > on hospital or creating an indicator variable for each hospital?
> 
> The indicator variable approach is a particular functional form 
that 
> allows for intra-group correlation.  With clustering, you allow 
for 
> arbitrary intra-group correlation.  For example, if you were 
working 
> with panel data, an indicator variable approach would be like 
> assuming intra-group correlation that never dies out over time, 
> whereas the cluster approach allows for any time of serial 
> correlation within the group.
> 
> Also, as Dave pointed out, the clustering approach gives you 
robust 
> standard errors and doesn't affect the coefficients (unless you go 
> down a 2-step GMM route or something like that).
> 
> Last point - you can combine the two approaches, i.e., use 
indicator 
> variables in the estimation, and also request standard errors that 
> are robust to any intra-group correlation that the indicator 
variable 
> approach didn't properly account for.
> 
> Hope this helps.
> 
> --Mark
> 
> > 
> > Thanks in advance for any commentary,
> > 
> > David J. Bernstein, Ph.D.
> > dbernstein@l...
> > 
> > 
> > 
> > *
> > *   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/



*
*   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