st: Re: competing risk modells in stata 8/9

[Lars Kroll <lars@climweb.de>]

Dear Layna and Elisabeth,

you both asked about competing risk modells in stata 8 or 9.

I hope this helps:

Intro
=====
I recently tried something similiar. First of all I would suggest to use 
a kind of st regression modell like cox, weibull (this depends on the 
possible shapes of all functions, if you don't know it use cox, if it's 
about mortality try gompertz).

Problem
========
You want to compare different effects of say x1,x2,x3 on different 
transistions say y0->y1, y1->y0, y0->y2, y1->y2, etc.

Solution
========
You need to stset each transition seperate (and therefore a variable 
which contains the state_{i,t} = [y0,y1,..]), estimate your modell, 
store your estimates, and so on..
Using

. est tab ALLYOURSAVEDESTIMATES, eq(1)

you are able to compare the effects

Syntax
======
(Two origin states one destination state, it it very important to 
specify time0 in the case of competing risk modells..)

// Y=0->Y=1
. stset time if inlist(state,1,0), id(id) ///
   failure(state==1)
. sort id _t
. by id: gen time0 = _t[1]
. stset time if inlist(state,1,0), id(id) ///
   failure(state==1) time0(time0)
. stcox y*
. est store state_0to1
. drop time0

// Y=2->Y=1
. stset time if inlist(state,1,2), id(id) ///
   failure(state==1)
. sort id _t
. by id: gen time0 = _t[1]
. stset time if inlist(state,1,2), id(id) ///
   failure(state==1) time0(time0)
. stcox y*
. est store state_2to1
. drop time0

// Results
est table state_* , equations(1)

References:
===========
Box-Steffensmeier and Jones (2004). Event History Modelling. Cambridge. 
pp.155

Blossfeld, Rohwer (2001). Techniques of Event History 2nd PR. Lawrence 
Erlbaum Associates, ?.

Mario Cleves, William Gould, Roberto Gutierrez (2004). 	An Introduction 
to Survival Analysis Using Stata, Revised Edition. College Station: TX.


--
____________________________
Lars E. Kroll
- Tutor für Statistik -

ANSCHRIFT:
Freie Universität Berlin
Garystr. 55
14195 Berlin
Raum 202

KONTAKT:
Website  www.lkroll.de
*
*   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/