Statistical tests
Mann–Whitney U test
Implemented
statistics.mann_whitneyDefinition
Rank-based test for whether one of two independent samples tends to have larger values.
Formula
U = R₁ − n₁(n₁ + 1)/2
Inputs and outputs
- sample a
- sample b
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
import evalsuite as es
es.mann_whitney_test(scores_a, scores_b)References
- Mann, H. B., & Whitney, D. R. (1947). On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics, 18(1), 50–60.