Replies: 1 comment
-
Hi there, have you tried using |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
I am currently working on predicting customer survival for a gas supplier. They require that I compute the conditional probabilities for survival, e.g given I made it to month 3 whats the prob of survival to month 4,5 and 6 ect. I managed to compute these using Kaplan-Meier however they said they would like the inclusion of covariates in the model. I then thought cox-ph model would be best. However I was using the method: for i in range of my timeline, kmf.survival_function / kmf.predict(i) to compute these probabilities. However, using cox I have noticed I get a predicted survival function for each entry. Has anyone a solution to this problem or any ideas on how to solve it
Beta Was this translation helpful? Give feedback.
All reactions