R help - Jul 2002 - survfit on coxph object with weights

Hi,
I am working on a study where we need to predict individual survival
curves from a cox-model fit with sampling weights. In both R and Splus
the survfit.coxph code starts with 

if(!is.null((object$call)$weights))
  stop("Survfit cannot (yet) compute the result for a weighted model")

My question is does anyone have code to get the expected survival curve,
or even just the base cumulative hazard (H^{\hat}_{0}(t)) ? 

I am surprised that it is not implemented yet since a formula is given
on page 35 of Dr. Therneau's report, "a package for survival analysis
in
S" for the base cumulative hazard with weights. I am looking at the
source for survfit.coxph to see if I can fix it myself, but if someone
has they can share I can get my report out and go back and send a patch
later.

Thanks for any help.

Nicholas

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

If you don't need the confidence intervals, which is the part of the formula
that I got hung up on, then there is an easy way to get the survival curve.
(A continual succession of other things to do has kept me from getting back to
this problem.)

   fit <- coxph(Surv(.....  etc etc
   details <- coxph.detail(fit)
   
The coxph.detail routine was originally written for debugging; it returns
nearly every intermediate internal computation that coxph does.

  The cumulative hazard can be plotted as
  
  	plot(details$time, cumsum(details$hazard))
  	   
This is for a hypothetical subject whose covariates are details$means[1,].
For a general subject with covariates x, multiply the cumulative hazard
by exp(sum(fit$coef * (x-details$means)))

	Terry Therneau
	
	  	   
  	   

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._