Why the fit_interval_censoring returns two different survival function?

[eduardo-janotti]

Hello.

I've been trying to use fit_interval_censoring function and going through lifelines' documentation to understand the results.

Every time that i use this, i receive two different survival functions, one called upper and the other lower, but i can't find anywhere in the documentation, or even in other places, the reason why the algorithm returns two different survival functions instead of just one, i can only find that it uses the Turnbull estimator. However, i couldn't find examples where the estimator produced two functions. Would you mind explaining why it returns two functions?

[CamDavidsonPilon]

This is a good question! Something that should go in the docs too.

For Turnbull-interval estimator, there's not a single optimal solution, but a range of optimal solutions: any curve that fits in between the two bounds you are provided back is optimal.

A perhaps simpler example: given the sequence of data 0, 8, 10, 20 - what is the median? Well, any number between 8 and 10 would be a median (so really I should ask "what is a median?").

[eduardo-janotti]

thanks for the prompt answer, that really helped clarify things.