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

st: suest and clogit


From   "Weesie, Jeroen" <J.Weesie@fss.uu.nl>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   st: suest and clogit
Date   Wed, 11 Sep 2002 07:48:43 +0200

Glenn Hoetker <ghoetker@uiuc.edu> asked about the support for -clogit- 
in -suest-, while -clogit- itself does not have a robust/cluster() option. 

Glenn (of course) rightly observes that clogit does not support the robust
and cluster() options supported by most Stata estimation commands. As far as
I can tell, there is no intrinsic (statistical or otherwise) reason for
this, but adding this feature to clogit is not a trivial amount of work,
especially for the "multiple successes per group" case. For the "one success
per group" case, as in the application to McFadden's choice model, however,
robust/cluster() is very simple. 

The scores are just the residuals depvar-pr(success), summed 
over all responses in groups. For the cluster() option we should sum over 
all responses in all groups in some cluster(). These computations are 
strictly analogous to those already provided in other computations for the
cluster() options, and are available via the programmer's command _robust. 

I added the support for clogit in the one-success-per-group case to suest
because (1)I needed this for myself -- many STB/SJ submissions arise from
requirements in my own work, -- research or consultation -- and (2) I
realized that one of the central applications of -suest- is to compute a
"better" version of the Hausman test for the IIA test for the Luce-McFadden
choice model (clogit), and hence likely will be needed by other users
interested in suest. 

I hope this suffices. 
Jeroen 

*
*   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