The KaplanMeierFitter.fit_interval_censoring function has a few lines commented-out regarding setting self.confidence_interval_survival_function_, and correspondingly its output is incomplete. I am noting a requirement that would make this function's interface compatible with your other fitters and fitting methods.
Current behavior
After fitting, the survival_function_ attribute of the model contains two columns, with column names f"{label}_lower" and f"{label}_upper", ignoring the ci_labels parameter.
Expected behavior
The behavior exhibited by other fitters is for the survival_function_ property to have one column with name label, and for the confidence_interval_survival_function_ property to have columns containing what is now in survival_function_, but with names that respect ci_labels.
Your documentation notes the following about KaplanMeierFitter.fit_interval_censoring:
This is new and experimental, and many features are missing.
Hello!
The
KaplanMeierFitter.fit_interval_censoring
function has a few lines commented-out regarding settingself.confidence_interval_survival_function_
, and correspondingly its output is incomplete. I am noting a requirement that would make this function's interface compatible with your other fitters and fitting methods.Current behavior
survival_function_
attribute of the model contains two columns, with column namesf"{label}_lower"
andf"{label}_upper"
, ignoring theci_labels
parameter.Expected behavior
survival_function_
property to have one column with namelabel
, and for theconfidence_interval_survival_function_
property to have columns containing what is now insurvival_function_
, but with names that respectci_labels
.Your documentation notes the following about
KaplanMeierFitter.fit_interval_censoring
:So, you may already know this.