Cameron Davidson-Pilon

Hi @Himscipy - this seems like a great introduction to the subject of renewal processes! There's not support in lifelines _yet_, but it's been on my "wouldn't it be great...

Hi @Dekermanjian - that's a heavily requested feature, so I think I'll expose this somehow in the next release.

Hi @tuliosarmento, try adding `import autograd.numpy as np` before the class definition

Hi @rchwn, Definitely there is a bug here. I wonder if it has something to do with the left-censored dataset (i.e. RMST is making some right-censoring assumption). Anyways, the two...

I did find a bug (fix will be in latest version, see #1278), and I suggest using `restricted_mean_survival_time(kmf, t=max_detected, return_variance=True)` for best accuracy.

@leavor this are possible to build quickly with the library: https://lifelines.readthedocs.io/en/latest/jupyter_notebooks/Custom%20Regression%20Models.html There's an example of the Exponential dist there. I chose not to include all these as they would confuse...

hm, that behaviour sounds strange, but must be data specific. I've been looking into this function recently, and have seen difference between transforms. I'll investigate further however.

That would be appreciated! It would be nice to understand the behaviour more.

@taoxu2016 is correct, and another change needs to be made: `transform` should read `time_transform`: ``` lifelines.statistics.proportional_hazard_test(lung_coxPH, lung_dataset.drop(['inst', 'ph.ecog', 'ph.karno', 'pat.karno', 'meal.cal'], axis = 1).dropna(), time_transform = 'km').print_summary() ```

So I dug deeply into this problem. In version 3.0 of `survival`, released 2019-11-06, a new, more accurate version of the `cox.zph` was introduced. This avoided an assumption of [variance...