[R] survreg and predict probability of failure

[Therneau, Terry M., Ph.D.]

-- begin included message ---
Hi,
I am fitting a weibull model as follows
my models is

s <- Surv(DFBR$Time,DFBR$Censor)

wei <- survreg(s~Group+UsefulLife,data = DFBR,dist="weibull")



How can i predict the probabilty of failure in next 10 days, for a new data with group =10 
and usefuleLife =100
Thx
shilpi  		 	   		

--- end inclusion ---

You need to create the cumulative failure curve:

  scurve <- predict(wei, newdata=data.frame(Group=10, UsefulLife=100), type="quantile",
	    p=seq(.1, .9, length=9))
  plot(scurve, seq(.1, .9, length=9), type='b')

This will give you the time points at which 10%, 20%, ..90% of the units are expected to 
fail.  If you want more precision repeat the process to "zero in" on the quantiles that 
are close to your time point.

Terry Therneau