Tim

+1, it would be very nice to add information about surfaces, like what is the risk of children playing on a playground

I am also looking for the train_val.prototxt -- can this be made available?

I'm also looking for a train_val.prototxt -- thanks!

@CamDavidsonPilon, I would very much appreciate your review (and approval, conditional on acceptance). My organization is really struggling without a sound method to generate confidence intervals on a difference in...

@CamDavidsonPilon, thank you for your response. I'm sorry, I overlooked `lifelines.utils.restricted_mean_survival_time`. I now validated that the 2 implementations produce the same point estimate on a test set. My newer implementation...

@CamDavidsonPilon, I'm hoping to hear back from you about whether you or I should include the other outputs from RMST, and therefore how to proceed with the difference in RMST...

I fixed an issue with RMST where, if the followup interval is greater than the last observed event, the confidence intervals were NaN. The latest commit resolves that issue and...

Great point, this won't work for all fitters. The current implementation depends on a `fitter.event_table`, so it only works for fitters that have an `event_table` property. Using data from `event_table`,...

upon further thought, should we push one step further upstream and give all fitters an `event_table`? It seems like a handy attribute and that making fitters more similar would be...

Unless I'm missing something, my last commit moved the `event_table` property into the BaseFitter class so that all fits get an `event_table` property. This passes tests -- is it correct...