0

I want to predict the customer lifetime value for each client while many clients are still active, hence right censored data.

I'm using the survival library for my censored data. I then run the coxph model to fit the training data and obtain the survival probabilities with:

predict(cox, newdata = df_survival, type = 'survival')

Since I'm interested in the customer lifetime value I want to transform these survival probabilities to lifetime values. Is there a way to obtain lifetime values from the results of fitting the Cox PH model?

Thanks!

Michieldo
  • 169
  • 1
  • 2
  • 15
  • Can you be more specific about what you mean by lifetime values? CoxPH models usually give you output like "5 year survival" for what proportion will still be "alive" at the given timepoint. Another usual output is median survival. – pgcudahy Jul 27 '20 at 11:30

1 Answers1

1

Cox proportional hazards models can't give you an estimate of survival time, but parametric survival models can. Choosing the correct model takes some knowledge of the shape of your dataset. See this thread and Applied Survival Analysis (chapter 8)

pgcudahy
  • 1,542
  • 13
  • 36