Skip to content
EvalSuite
Documentation menu

Statistical tests

Kruskal–Wallis test

Implementedstatistics.kruskal_wallis

Definition

Rank-based test comparing three or more independent groups.

Formula

H = (12 / N(N + 1)) Σ Rⱼ²/nⱼ − 3(N + 1)

Inputs and outputs

  • k independent samples

Returns: StatisticalTestResult

Assumptions

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

Limitations

No metric-specific limitations are documented yet. Interpret the value alongside the task, data, and other metrics.

Python API

PythonSince v0.2.0
import evalsuite as es

es.kruskal_wallis_test(scores_a, scores_b, scores_c)

References

  1. Kruskal, W. H., & Wallis, W. A. (1952). Use of ranks in one-criterion variance analysis. Journal of the American Statistical Association, 47(260), 583–621.

Implementation status