Roberto:
Unfortunately the inverse mills ratio is derived from the
normal distribution, while logit and ordered logit use
the logistic distribution, so this won't work. (Stata will
do what you say, and it won't complain, but it doesn't
make sense.) This approach seems more plausible to me with
-probit- and -oprobit-, since these use the normal
distribution.
HTH,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting adress:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
-----Original Message-----
From: [email protected] [mailto:[email protected]]On Behalf Of Roberto Fontana
Sent: donderdag 5 oktober 2006 15:35
To: [email protected]
Subject: st: Inverse mills ratio and ordered logit
Dear Statalisters
I am running a two step estimation to check for selection bias. In the first
stage I have a logit. In the second stage I use an ordered logit with some
explanatory variables plus the inverted mills ratio calculated from the first
stage.
I am wondering whether this is a correct way to proceed (i.e. whether I get
the corrected standard errors in the second stage). Would there be additional
problems if, instead of the ordered logit, I had a multinomial logit in the
second stage?
In general, can anyone suggest any published article which tried a similar
procedure without OLS in the second stage.
Thank you very much for yor help.
Roberto
--
Roberto Fontana (PhD)
CESPRI - Bocconi University
Via Sarfatti 25
20136 MILAN
ITALY
Tel: (+)39 02 58363037
Fax: (+)39 02 58363399
*
* 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/
