Skip to content
EvalSuite
Documentation menu

Effect sizes

Cliff's delta

Implementedeffect-size.cliffs_delta

Definition

Non-parametric effect size: the probability that a value from one group exceeds a value from the other, minus the reverse.

Formula

δ = (#(a_i > b_j) − #(a_i < b_j)) / (n_a · n_b)

Range: [−1, 1]

Inputs and outputs

  • a: first sample
  • b: second sample

Returns: result object

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 v0.1.0
import evalsuite as es

es.cliffs_delta(a, b)

References

  1. Cliff N. Dominance statistics: ordinal analyses to answer ordinal questions. Psychol Bull. 1993;114(3):494-509.

Implementation status