Home › Survival › Survival estimators

Each subject's survival at given times: lifelines' predict_survival_function.

public double[] PredictSurvivalFunction(ReadOnlySpan<double> design, ReadOnlySpan<double> times)

Parameters — design holds the subjects' covariates row-major, FeatureCount per row; empty for the one subject of a fit with no covariate. times are the positive, finite times to read at.

Returns — row-major, one row per subject and one column per time: exp(−H) of PredictCumulativeHazard.

Exceptions — ArgumentException when design is not whole rows of finite covariates, or a time is not positive and finite. ArgumentException naming design also when it is more cells than one array holds (#1614).

Example — two subjects at ten months.

using Lodestar.Survival;

double[] dose = [1.0, 2.0, 0.5, 3.0, 0.0, 1.5, 0.0, 2.5, 1.0, 0.5];
double[] months = [5, 8, 12, 3, 15, 9, 20, 6, 11, 14];
bool[] died = [true, true, false, true, true, true, false, true, true, false];

AftSummary fit = AcceleratedFailureTime.Fit(AftModel.Weibull, dose, months, died, featureCount: 1);

double[] survival = fit.PredictSurvivalFunction([0.0, 2.0], [10.0]);
double untreated = Math.Round(survival[0], 6);   // => 0.973697
double twoUnits = Math.Round(survival[1], 6);    // => 0.004345

Remarks — an empty times reads at no time and returns an empty array: a parametric curve has no event times of its own to default to, where CoxSummary.PredictSurvivalFunction falls back on its baseline's.

Applies to — net10.0, netstandard2.0.

See also — AftSummary, AftSummary.PredictCumulativeHazard.