[S] Cox Model

Jaime =?iso-8859-1?Q?G=F3mez?= (jgvillas@posta.unizar.es)
Tue, 14 Jul 1998 10:14:35

Dear S-Plus users,

I am trying to estimate the Cox Proportional Hazards Model. I have data
about spells of employment and unemployment and I would like to estimate
the probability of becoming employed in function of several regresors.

The characteristics of the data are the following. Not all the people enter
the risk set (for the first time)at the same time and they do not stay in
it for the same time either (that is the period of risk is not always the
same for all the subjects). I have multiple events for each subject. The
risk periods are discontinous due to the fact that the ocurrence of an
event (the person is employed) puts the subject out of the risk set (I
assume they are in the risk set only when they are unemployed). All the
covariates may be (and some of them are) time varying.

As far as my knowledge reaches I have to use the formulation of the model
as a counting process, as was proposed by Andersen and Hill. My questions
are: is this correct? is it possible to estimate this model with S-Plus?
how do I have to arrange the data for this? how can I test the model? how
can I check the proportionality assumption? is there any other
distinguishing feature of this model in relation with the basic one?.

Any help, suggested material or suggestions in general are welcome.


Jaime Gómez.
Department of Business Economics.
Faculty of Economics and Management.
University of Zaragoza. SPAIN.
E-mail: jgvillas@posta.unizar.es
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news