Non-Parametric
Estimators that make no assumption about the shape of the distribution:
the survival curve is read straight from the data. Each estimator is
exported as a ready-made instance – KaplanMeier, NelsonAalen,
FlemingHarrington and Turnbull – whose fit(x, c, n, t)
returns a NonParametric model.
That model carries the survival, hazard and quantile functions,
confidence bounds and bands, the restricted mean and plotting.
The theory is in Non-Parametric Estimation and worked examples are in Non-Parametric SurPyval Modelling. The log-rank test and the two-group restricted-mean difference, which compare non-parametric estimates between groups, are documented in Comparison Tests and Validation Metrics.
Non-Parametric Class
The fitted model every estimator returns.
Non-Parametric Estimators
Kaplan-Meier, Nelson-Aalen and Fleming-Harrington handle observed,
right-censored and left-truncated data, and raise a ValueError on
left- or interval-censored or right-truncated data; Turnbull handles
every combination of censoring and truncation.
Other Non-Parametric Functions
Zero-failure (success-run) testing, and the plotting positions used by
probability plots and by the MPP parametric fitting method.