Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Varying survival distributions & interval censoring (2)


From   "Stephen P Jenkins" <stephenj@essex.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Varying survival distributions & interval censoring (2)
Date   Wed, 23 Jun 2004 12:04:18 +0100

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> siyama@who.int
> Sent: 23 June 2004 11:07
> To: statalist@hsphsun2.harvard.edu
> Subject: st: Varying survival distributions & interval censoring (2)
> 
> 
> My thanks to Stephen Jenkins for his pointer below, I found 
> the notes quite detailed and helpful. 
> 
> The "Discrete" entry in the [ST] Manual reads:
> 
> Discrete-time survival analysis concerns analysis of 
> time-to-event data whenever survival times are either
> (a) intrinsically discrete (e.g. numbers of machine cycles), or
> (b) grouped into discrete intervals of time ("interval 
> censoring"). If intervals are of equal length, then the same 
> methods can be applied to both
> (a) and (b): survival times are positive integers.
> 
> In the data set I am handling, every subject has his/her OWN 
> interval of time, that is their actual failure happens 
> between two home-visit dates that the fieldworker makes.  The 
> home-visit dates allow us to calculate the exact ages of the 
> subjects who fall in one of 3 categories:
> 
> i.	Left-censored: the fieldworker initiates the home visits and
> observes the failure has already taken place.
> 
> ii.	Interval-censored: failure has take place between two 
> home visits,
> and the fieldworker has record of both 	dates.
> 
> iii.	Right-censored: the fieldworker completes the last home 
> visit and
> failure has not been observed.
> 
> 
> I am not sure exactly on how to structure the data to reflect 
> the 3-states above to use with "streg" or a similar model 
> that would allow me to vary the distribution of the failure 
> time (between exponential, weibull, lognormal, loglogistic 
> and generalised gamma).


If I understand you correctly, you won't be able to use -streg-. You
will have to program yourself, using -ml-, where the sample
log-likelihood expression will include a number of different components
corresponding to each of (i)-(iii) above. [The text by Klein and
Moeschberger, /Survival Analysis/ has a taxonomy of different likelihood
contribution types for the different types of spell that you mention.]
At the stage at which you characterise the contributions, you will have
to also decide which failure time distribution to assume.

Stephen
-------------------------------------------------------------
Professor Stephen P. Jenkins <stephenj@essex.ac.uk>
Institute for Social and Economic Research
University of Essex, Colchester CO4 3SQ, U.K.
Tel: +44 1206 873374.  Fax: +44 1206 873151.
http://www.iser.essex.ac.uk   



*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index