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

st: sample selection problem in a discrete-time event history analysis(competing risks)


From   Teresio Poggio <teresio.poggio@poste.it>
To   statalist@hsphsun2.harvard.edu
Subject   st: sample selection problem in a discrete-time event history analysis(competing risks)
Date   Mon, 26 Sep 2005 20:53:16 +0200

Dear Statalisters,

I'm studying housing tenure at first marriage among Italian couples, using retrospective information collected from married women in a cross-sectional national survey.

I'm doing a discrete-time event history analysis for women and men, modeling competing risks for tenure of their first accomodation (home ownership, rent, rent free, cohabiting with parents) versus not marrying.
In practice, I'm using mlogit procedures (Stata 8; 9 is shipping) on a person-year data matrix, following the approach proposed by Allison (1982).

I have a sample selection problem as information on accomodation at first (or only) marriage was not collected in the survey for divorced people, widows and not married cohabiting partners. Nor for people that married twice or more.
The problem is related only to my dependent variable, housing tenure at first marriage. While information on year of the first marriage and relevant time-varying independent variables are available for all the records.

In short, my (women, for instance) sample is organized in this way:

a) 1,173 never married women
Included in the analysis (at "risk" of marriage)

b) 11,486 married women, both partners at their first marriage
Included in the analysis (outcome known)

c) 1,467 (total) divorced, widowed women and not married women,in couple
Excluded from the analysis because tenure at first marriage is unknown.

d) 124 married women, but not at their first marriage.
Excluded from the analysis because tenure at first marriage is unknown.

e) 143 married women, but their husbands are not at their first marriage
Excluded form the analysis for theoretical reasons and analytical simplicity

Simply discarding (c), (d) and (e) will result in seriously biased estimates. So I'm looking for a way to control for this selection.
Can anyone provide suggestions?

Best,

Teresio

Allison, P. D.
1982 Discrete-time methods for the analysis of event histories, in Sociological Methodology, a cura di S. Leinhard, San Francisco, Jossey-Bass, pp. 61-98.



-----------------------------------------------------------
Teresio Poggio
Dept. of Sociology and Social Research
University of Trento
Via Verdi, 26 - 38100 Trento - Italy
E-mail: teresio.poggio@soc.unitn.it
teresio.poggio@poste.it
-----------------------------------------------------------










*
* 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