Skip to content
EvalSuite
Documentation menu

Confidence intervals

Percentile bootstrap

Implementedconfidence.bootstrap_percentile

Definition

Interval from the empirical quantiles of a statistic recomputed on resampled data.

Formula

[θ̂*(α/2), θ̂*(1 − α/2)]

Inputs and outputs

  • data
  • statistic
  • n_resamples
  • random_state

Returns: BootstrapResult

Assumptions

No assumptions beyond valid, aligned inputs of the documented types.

Limitations

  • Can under-cover for skewed or biased statistics; BCa corrects for bias and skewness.

Python API

PythonSince v0.1.0
import evalsuite as es

es.bootstrap_ci("f1", y_true, y_pred, method="percentile", random_state=0)

References

  1. Efron, B. (1979). Bootstrap methods: another look at the jackknife. The Annals of Statistics, 7(1), 1–26.

Implementation status