[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Accounting for sample selection in discrete-time survival models

From   "Luis Ortiz" <>
To   <>
Subject   st: Accounting for sample selection in discrete-time survival models
Date   Thu, 17 Apr 2008 15:07:08 +0200

Dear Statalisters,

I am carrying out survival analysis on the transition from a state A to
another state B. But I suspect many of the variables possibly affecting this
transition are also quite determinant in having people in my population of
study (risk set). In other words, I suspect population in state A is not
randomly selected, and many of my covariates may not be just affecting the
transition from state A to state B, but also the mere inclusion of many
individuals in state A.

My even of study occurs in continuous time but observed survival times are
grouped into intervals in my data. Given this restriction, I am using the
command 'CLOGLOG', as suggested by Stephen Jenkins in his lessons. Does
anyone know if 'cloglog' allows for any option that corrects for this
problem of sample selection I have previously mentioned? Alternatively, is
there any option that I could resort to?

Moreover, I suspect that my main event of study naturally competes with
another one. Both events are quite likely dependent on each other. Given the
fact that I am forced to use discrete-time models, I am using a MULTINOMIAL
LOGIT MODEL, as recently suggested by Stephen Jenkins to Pavlos C. Symeou's.
Again, my question is similar to the previous one: do you know the
possibility of carrying out a similar analysis but accounting for the fact
that my sample is possibly not randomly selected?

Many thanks for your attention.

Luis Ortiz

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index