2
$\begingroup$

I'm creating a calibration plot for my breast cancer prognostic Cox model, which doesn't include any fancy transformations, using the calibrate() function in the rms package for R. Should I use the hare method or KM? Similar literature to what I'm doing always uses the KM method, but I'm curious if hare is more advantageous.

There are 3 options I have: use the KM method, use the hare method, or plot both hare and KM on one figure.

Here's sample code:

library(rms)
library(survival)
library(mice)

remove(veteran)
data(veteran)
veteran$trt=factor(veteran$trt,levels=c(1,2))
veteran$prior=factor(veteran$prior,levels=c(0,10))
survmod=with(veteran,Surv(time,status))
fit=cph(survmod~celltype+karno,data=veteran,surv=TRUE,u=100,x=T,y=T)
cal.KM <- calibrate(fit, u=100, cmethod='KM', m=20, B=20)
cal.hare=calibrate(fit,u=100,cmethod='hare',B=20)
plot(cal.hare)
plot(cal.KM,add=TRUE)

Thank you for your help!

$\endgroup$
3
$\begingroup$

Sorry I didn't see this question when you first posed it.

Any method that requires binning of continuous variables is to be avoided. The often-used method of stratifying predicted values and then computing ordinary Kaplan-Meier estimates was always problematic - you could see this by how noisy the results appeared. I never use calibration methods based on binning/categorization any more. hare has more precision because it effectively uses interpolation across levels of predicted values, unlike stratification.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.