Predict_survival_function returns implementation error.

[dlee07]

Hi,

So I trained a few parametric regression models. Through some scoring metrics, the coxph model with a piecewise equation had the best performance.

My data is looking at subscribers/churners who started from August 2022 to October 9th 2022. The maximum duration in Months is 14 for censored subjects.

I want to predict the additional months of survival for the censored subjects. I tried to use the coxph.predict_survival_function and pass in the duration column of the censored subjects. But I get this error:

Am i doing this right or is this an implementation issue regarding the coxph fitter?

[dlee07]

Hi @CamDavidsonPilon ,
just wanted an update on this question

[CamDavidsonPilon]

It hasn't been implemented for the piecewise model, likely because it's difficult.

[dlee07]

Hi @CamDavidsonPilon

Thank you for the response. An unrelated question: Should the average predicted survival curve of an AFT Model be close to the kaplan meier curve? This has a concordance index of 0.78

In addition, what is the benchmark ICI and E50 that I should be aiming for here?