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

st: RE: event history and research design question


From   Benoit.Laplante@inrs-UCS.uquebec.ca
To   statalist@hsphsun2.harvard.edu
Subject   st: RE: event history and research design question
Date   Fri, 22 Nov 2002 13:19:35 -0500

This is not really a stata question, but here is a short tentative answer.

I can not see how your problem could be dealt with using delayed entry.

If the variables describing third party intervention are few, discrete, and
have few categories, a simple solution might be to combine them with
presence or absence of third party intervention to create a single variable,
estimate the model with this new series of dummies and test your hypotheses
using linear combinations of coefficients. If your variables are continuous,
you may have to do a bit more modelling.

Benoit Laplante
Centre interuniversitaire d'etudes demographiques
INRS/Universite du Quebec
Montreal Qc Canada

-----Message d'origine-----
De : andrew enterline [mailto:ajenter@hotmail.com]
Envoyé : vendredi 22 novembre 2002 11:36
À : statalist@hsphsun2.harvard.edu
Objet : st: event history and research design question


Hello,

I have a sample of 213 civil wars.  My dependent variable of interest is 
civil war duration.  A subset of this sample 213 civil wars--18 civil 
wars--experience a third party intervention.   I measure several independent

variables pertaining to these 18 cases with intervention.  Unfortunately, 
these independent variables are not measured for the 195 cases in which 
intervention did not occur.  However, I'd prefer to include the information 
for the dependent variable (i.e., the duration) for these 195 cases in order

to serve as a benchmark against which to assess the intervention subset.  
One of my colleagues suggested using the -stset- command for "late entry" to

retain this information on the dependent variable for these cases, but avoid

including zeros for observations where the intervention variables are 
unmeasured.  Another suggestion was to drop an event history approach 
altogether in favor of testing differences of means for different subsamples

in the data, but this approach has limitations as well.

Would anyone have any suggestions about how I might address this problem, or

citations to similar designs in the literature?  I am open to any 
recommendations whatsoever.

thank you ahead of time,
Andrew Enterline

a.j.enterline
University of North Texas
Dept. of Political Science
Box 305340
Denton, TX 76203
(v) 940.565.2313
(f)  940.565.4818
email: ajenter@unt.edu
homepage: http://www.psci.unt.edu/enterline/enterline.htm





_________________________________________________________________
The new MSN 8: smart spam protection and 2 months FREE*  
http://join.msn.com/?page=features/junkmail


*
*   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