calibration plots

#1
Hi all,

1.does it make any sense that a random forest survival analysis plot doesn't go from 0-100% prediction?
(i'm using the PEC package in R)
2. is there any possibility to calculate AUC of the calibration plots in order to compare them? is there a better way?


1569948439133.png
 

hlsmith

Not a robit
#2
In my mind i liken the calibration curve plot to the hosmer lemeshow test. Thus i wonder if you could examine the observed versus the predicted values?
 
#3
In my mind i liken the calibration curve plot to the hosmer lemeshow test. Thus i wonder if you could examine the observed versus the predicted values?
thanks, however, in my understanding, Hosmer-Lemehshow doesn't deal with survival data (unless some transformation done)