I have a person who has survival data with time-varying covariates. He's
interested in the predictive power of a test. There's no problem getting the
hazard ratio associated with the most recent test results. However, he was
wondering if there were some way of expressing this as something equivalent
to positive and negative predictive value.
This doesn't meet the standard definition for predictive value calculation
(since the length of follow-up after the test result is variable, and
patients have multiple tests) but does anyone out there know of any work in
expressing data like this in a form analogous to predictive value?
We thought of defining something like 'death within 6 months' as an endpoint
(this disease is quite rapidly fatal). But is there some more elegant
formulation that would make fuller use of the varying lengths of time at
risk?
I can't bear to think about this right now, or about anything else for that
matter. It's late Friday afternoon, and I need a swim and a sauna.
Ronan M Conroy ([email protected])
Lecturer in Biostatistics
Royal College of Surgeons
Dublin 2, Ireland
+353 1 402 2431 (fax 2764)
--------------------
Bombing our way to a perfect world...
*
* 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/
