I need to predict number of days past due date to collect cash for a product based company which sells their product to more than 5000 clients. We receive ~100K invoices on weekly basis and the prediction needed at the invoice level. The predictors would be invoice amount, client information (like demographics, finances, etc.) and their historical payment behavior.

Which statistical technique would be best fit in this scenario and why?

  • $\begingroup$ Have you considered a proportional hazards model? $\endgroup$
    – mdewey
    Nov 16, 2016 at 12:15
  • $\begingroup$ No, I have not. What's that? $\endgroup$
    – P.K.
    Nov 16, 2016 at 12:20
  • $\begingroup$ See @peterflom's answer $\endgroup$
    – mdewey
    Nov 16, 2016 at 13:10
  • $\begingroup$ To get started with survival analysis I suggest you take a look at UCLA. They have examples in different programming languages and are pretty easy to follow. $\endgroup$
    – Marcel10
    Nov 16, 2016 at 13:53

1 Answer 1


Because it's time to event, you want survival analysis. Because you have both invoice level and person level data, you want a multilevel analysis. Fortunately, such models do exist.

E.g see Yau, Multilevel Models for Survival Models with Random Effects in Biometrics, vol 57, p 96-102.

There are ways to model these in SAS, R (coxme package) and Stata

  • 1
    $\begingroup$ Addition to the R suggestion: another option is the frailtypack package. Even the survival package has a function to include a frailty term in your survival model. $\endgroup$
    – Marcel10
    Nov 16, 2016 at 14:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.