Closed erikbern closed 6 years ago
Split up all models into three classes
They are all mostly special cases of each other so a bunch of glue code lets us translate models.
This mostly simplifies plotting etc but the glue code in itself adds some complexity
Fitting a Weibull to synthetic Weibull data:
Fitting a Kaplan-Meier to the same data:
(not surprisingly, larger uncertainty estimates)
Coverage increased (+2.09%) to 87.906% when pulling 6da2043a420f77b2d4c6c362383c91ab108b419a on refactoring into 0da313914d24e93110034979f956e2369dbae11c on master.
Split up all models into three classes
They are all mostly special cases of each other so a bunch of glue code lets us translate models.
This mostly simplifies plotting etc but the glue code in itself adds some complexity
Fitting a Weibull to synthetic Weibull data:
Fitting a Kaplan-Meier to the same data:
(not surprisingly, larger uncertainty estimates)