R help - Jun 2008 - expected risk from coxph (survival)

Hello,
 
When I try to to obtain the expected risk for a new dataset using coxph in the
survival package I get an error. Using the example from ?coxph:
 
> test1 <- list(time=  c(4, 3,1,1,2,2,3),+                
status=c(1,NA,1,0,1,1,0),+                 x=     c(0, 2,1,1,1,0,0),+           
sex=   c(0, 0,0,0,1,1,1))> cox<-coxph( Surv(time, status) ~ x +
strata(sex), test1)  #stratified model> > new<-list(time=  c(5,
1,1,2,2,4,3),+                 status=c(1,NA,1,0,0,1,1),+                 x=    
c(0, 2,1,1,1,0,0),+                 sex=   c(0, 0,0,0,1,1,1))> >
predict(cox,new,type="expected")Error in predict.coxph(cox, new, type
= "expected") :   Method not yet finished
 
I assume that this is something that has simply not yet been incorporated into
the survival package. Does anyone know of a way to calculate the expected risk
for a new data set? Is this even possible? I would appreciate any help that you
could give me.
 
Cheers,
Reid
 
 
 
_________________________________________________________________


	[[alternative HTML version deleted]]

Reid Tingley <r_tingley <at> hotmail.com>
writes:
> When I try to to obtain the expected risk for a new dataset using coxph in
the
survival package I get an error.
> Using the example from ?coxph:
# Example rewritten by DM; please do not use HTML mail
library(survival)
test1 <- list(time=  c(4, 3,1,1,2,2,3),
            status=c(1,NA,1,0,1,1,0),
            x=     c(0, 2,1,1,1,0,0),
            sex=   c(0,0,0,0,1,1,1))
cox<-coxph( Surv(time, status) ~ x + strata(sex), test1)  #stratified model
new<-list(time=  c(5, 1,1,2,2,4,3),
          status=c(1,NA,1,0,0,1,1),
          x=     c(0, 2,1,1,1,0,0),
          sex=   c(0,0,0,0,1,1,1))
predict(cox,new,type="expected")


 #   }
#    else if (type == "expected") {
#        if (missing(newdata)) 
#            pred <- y[, ncol(y)] - object$residuals
#        else stop("Method not yet finished")
Looks like this is "by design"; see the code above.

You might try to use cph and predict.Design from Frank Harrell's Design
package
instead.

Dieter