`hazardous.metrics`.aj_calibration#

hazardous.metrics.aj_calibration(y_calibration, times, pred_calibration, alpha=2, reduction='mean', min_prop_at_risk=0.05)#

Overall AJ calibration score aggregated across all events.

Computes the per-event AJ calibration scores via aj_calibration_per_event(), then reduces them to a single number:

\[\text{AJ-Cal} = \frac{1}{K} \sum_{k=1}^{K} \text{AJ-Cal}_k \quad \text{(reduction='mean')}\]

or the sum (resp. max) when reduction='sum' (resp. reduction='max').

Each per-event score integrates the pointwise error \(AJ_k(t) = |\bar{F}_k(t) - \hat{F}^{AJ}_k(t)|\) between the mean predicted cumulative incidence \(\bar{F}_k(t) = \frac{1}{n} \sum_{i=1}^n \hat{F}_k(t \mid \mathbf{x}_i)\) across the calibration set and the marginal Aalen-Johansen reference \(\hat{F}^{AJ}_k(t)\) fitted on the same set (Kaplan-Meier via km_calibration() for event 0).

Parameters:

y_calibrationarray-like of shape (n_samples, 2): Survival outcomes of the calibration set, with columns "event" and "duration".
timesarray-like of shape (n_times,): Time points at which the CIFs were predicted.
pred_calibrationarray-like of shape (n_samples, n_events+1, n_times): Predicted incidence probabilities at times for the calibration set.
alphaint, default=2: Exponent applied to the pointwise difference before integration.
reduction{“mean”, “sum”, “max”}, default=”mean”: How to aggregate per-event scores into a single value.
min_prop_at_riskfloat, default=0.05: Lower bound on the proportion of the set still at risk required to include a timepoint in the integral. Stops the integration once fewer than this fraction of subjects remain at risk. Set to 0 to integrate over the full time grid.

Returns:

scorefloat: Aggregated AJ calibration score.

hazardous.metrics.aj_calibration#

`hazardous.metrics`.aj_calibration#