Hi all,
I would like to ask a statistical question regarding stratification in survival analysis.
I'm running a discrete-time event-history model of time-to-drop-out on a sample of adult students. I've decided to stratify by enrolment status (part-time or fulltime in any given person-term), as most of the effects of the other covariates seem to be different for these two groups. I would like to know, however, whether it is necessary to restrict my data-set so that no subject contributes person-terms to both models. This basically comes down to a general question of whether it is appropriate or advisable to stratify by a time-varying covariate. While there are actually very few instances of subjects switching enrolment status in my sample, I have tentatively opted to censor anyone upon the first time they switch enrolment status (in the same way that a single event model deals with competing events by censoring upon the occurrence of an alternative event). Is it necessary that one ensure any stratification in a survival analysis is done by only time invariant predict!
ors? If so, is there a better way of dealing with the problem than the one that I have adopted? Thank you very much, Kevin.
*
* 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/
