I have built a Cox PH model. I obtained the predicted survival probabilities. My questions are:

  1. How to get the predicted survival time, or time to event, using survival probabilities?
  2. I need validation using lift charts, as in how many events is the model capturing in the first three deciles.

There are several ways to do this in R. One is to use the survest function in the rms package in conjunction with that package's cph function.

Lift charts are not used to validate models but instead to find the subset of higher risk individuals will get you the biggest bang for the buck. Lift charts do not use deciles or other binning methods but deal with continuous predictions. You can even make a lift chart on the relative log hazard scale without bothering to compute cumulative risks.

It may not be a good idea to compute how many events you can capture per se, as these are functions of the censoring distribution, but rather to estimate the proportion of cumulative risk that be captured with a subset of high-risk subjects.


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.