Development build. This page describes main, not a released package. The latest published Lodestar.Survival is 0.2.0 — read its documentation.

Home › Survival › Survival estimators

Each subject's cumulative hazard at every event time: lifelines' predict_cumulative_hazard.

public double[] PredictCumulativeHazard(ReadOnlySpan<double> design)

Parameters — design holds the subjects' covariates row-major, FeatureCount per row; empty for the one subject of a fit with no covariate.

Returns — row-major, one row per subject and one column per EventTimes entry: each subject's covariates, with a one for the intercept, times the cumulative coefficients.

Exceptions — ArgumentException when design is not whole rows of finite covariates. ArgumentException naming design also when it, with any intercept column, or its rows times the event times, is more cells than one array holds (#1614, #1616).

Example — an untreated and a treated patient, at the last of the ten event times.

using Lodestar.Survival;

double[] treated = [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1];
double[] months = [3, 10, 5, 4, 8, 18, 12, 6, 9, 20, 7, 15];
bool[] died = [true, true, true, false, true, true, true, true, true, false, true, true];

AalenSummary fit = AalenAdditive.Fit(treated, months, died, featureCount: 1);

double[] hazard = fit.PredictCumulativeHazard([0.0, 1.0]);
double untreated = Math.Round(hazard[9], 6);   // => 2.45
double onTreatment = Math.Round(hazard[19], 6);   // => 0.366667

Remarks — the prediction is linear in the covariates, which is what "additive" means: the treated patient's hazard is the untreated one's plus the treatment's cumulative coefficient, 2.45 − 2.083333. Nothing keeps it non-negative or non-decreasing; a design far from the fitted one can give a cumulative hazard that falls.

Applies to — net10.0, netstandard2.0.

See also — AalenSummary, AalenSummary.PredictSurvivalFunction.