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

Re: st: partially retrospective survival analysis

From   Steven Samuels <>
Subject   Re: st: partially retrospective survival analysis
Date   Wed, 9 Apr 2008 16:31:19 -0400

Alan, the situation you describe is an example of left-censoring. Recall that right censoring at a time T provides information that the event time Y >T. The same applies to left-censoring-- Y<T. This is generally handled by parametric models, as the likelihood contribution of a left-censored event is F(T), where F(t) = P(Y<t). In Stata, the programs -stpm- (use the leftvar() option and -intreg- (and maybe some others) will do this. I cannot find a Stata command to implement Bruce Turnbull's 1976 nonparametric estimate, but -stpm- gives flexible models.


(Turnbull, B.W. (1976) The empirical distribution function with arbitrarily grouped, censored and truncated data. J. R. Statist. Soc. B, 38, 290-295

On Apr 9, 2008, at 3:23 PM, Feiveson, Alan H. (JSC-SK311) wrote:

Hi - Has anyone seen models for or tried doing a survival analysis when
for some observations the data had been observed after the event already
occurred? For example, trying to predict time to heart attack given some
ECG measurements, but for some patients, the measurements were obtained
after the patient already had been in the hospital after the attack. So
in one sense the time to event is actually negative and censored at
zero. Probably a better approach would be some sort of
discrete/continuous mixture model where with a certain probability, the
event has already occurred and given that it hasn't, a standard survival
model takes effect.

Al Feiveson

* For searches and help try:
*   For searches and help try:

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