Date:         Thu, 2 Nov 2006 15:29:04 +0100
Reply-To:     e.janssen@free.fr
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Eric Janssen <e.janssen@free.fr>
Subject:      Re: Recurrent Event Counting Process
Comments: To: "Butler, Deborah {FLNA}" <deborah.butler@fritolay.com>
In-Reply-To:  <5A5BDC9663F87547AA55455B55CE59D35D6ED8@PEPWMV00015.corp.pep.pvt>
Content-Type: text/plain; charset=ISO-8859-1

> My idea is based on the fact that a discrete time survival analysis can be
restructured to be solved using ordinary logistic regression. At least one
reference for this is Paul Allison.

Check:
Allison P. D. (1984) "Event history analysis. Regression for longitudinal event
data", Sage University Paper #46, Series Quantitative Applications in the
social sciences.

Allison P. D. (1982) "Discrete-time methods for the analysis of event histories"
in S. Leinhardt (ed.) Sociological Methodology, San Francisco: Jossey-Bass, pp.
61-98.
A quite comprehensive introduction of the method.

Goodies:
- you can use the time IV as linear or categorical;
- you can apply multinomial logreg;
- no need to check equiprobablity assumptions as in Cox model
Baddies:
- an event may occur more than once within a single time interval - but there
are tricks to manage it properly.

Doug Massey and colleagues used discrete time event analysis in their studies of
the Mexican Migration project data base - a technique applied in Mexico for
other longitudinal surveys (EDER and ERMEU).