Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Issues re. survival analysis

From   Eleonora Paesen <>
Subject   st: Issues re. survival analysis
Date   Fri, 17 Dec 2010 13:41:21 +0000

Dear Statalisters,

I would have a few questions regarding survival analysis:

1. I have used st-commands in STATA in order to prepare my data file
to run a Cox proportional hazard model and among my covariates I use
the fraction of cumulative time spent in unemployment.
That means that my variable is constructed as 'cumulative time spent
in unemployment/_t'.
Does anybody of you know if STATA returns different Cox estimates in
the case the model is run on:
i) a 'contracted' data set (with a single line for spells that have a
duration longer than 1 month - unless the spell has been st-splitted)
ii) an 'expanded' data set (with a line of for each month of a spell
which is obtained with the following command:

stsplit time, at(1(1)'value retrieved from t of last observed exit')

The reason of my question is that obviously the value of '_t' changes
in the two data set and would like to know if estimation would be
equivalent on models run in the two cases.

2. Let's say I am interested in the effect of the fraction of
cumulative time spent in unemployment on the transition to second
births. This means that the observation window for this particular
piece of analysis starts from the time when my respondents had a first
child. How can I account in my Cox proportional hazard model for the
effect of the entire employemnt history. I think it is reasonable to
suppose that the risk of having a second child is also influenced by
the employment experience prior the birth of the first child. Is there
a way to account dinamically for this effect?

3. After using a Cox proportional hazard model I would like to move to
a parametric framework. Does anybody could give me any reference of a
good text or article on the criteria one could adopt in descerning
among different parametric model against the estimations obtained by
my Cox model?

Many thanks in advance for your help and suggestions!
*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index