Re: st: Including left-censored data in survival model

[Ronan Conroy <rconroy@rcsi.ie>]

On 11 Feb 2011, at 21:28, Heidi wrote:

> Hello all,
> 
> I'd like to estimate a Cox hazards model with my dataset, but it looks
> like a good portion of my observations are left-censored and failed
> in/prior to the first period. (I do have the year that they failed,
> but no access to data prior to my time 0.) Having checked _st, it
> looks like those subjects have already been dropped from the analysis.
> However, these left-censored subjects were mostly also in a particular
> category, so dropping them has made the other two categories seem much
> more likely to fail.

There is a confusion of time metric here. Your year zero is not in the same time metric as your failure time, since some subjects failed before time zero. 

You need to identify a time zero which marks the same entry point for all participants (diagnosis, for example, in medicine). Your covariates are measured at some time after this, and this needs to be factored into the analysis. You can use -enter- and -origin- to define the time at which covariate measurements start and at which risk starts for all subjects. 



Ronán Conroy
rconroy@rcsi.ie
Associate Professor
Division of Population Health Sciences
Royal College of Surgeons in Ireland
Beaux Lane House
Dublin 2


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/