Statistics Glossary → Cohen's h
Cohen's hconcept
An effect size for the difference between two proportions, expressed on the scale of the arcsine-transformed proportions.
It is computed as h = 2·arcsin(√p₁) − 2·arcsin(√p₂). A raw difference p₁ − p₂ means different things depending on where the proportions lie on the scale: going from 0.50 to 0.55 and from 0.01 to 0.06 are both 5-point differences, but the second is a much larger relative change. The arcsine transformation roughly stabilises the variance of a proportion and takes this into account. In the one-sample case, the null proportion p₀ is used instead of p₂.
By Cohen's rough benchmarks, |h| ≈ 0.2 is a small, 0.5 a medium and 0.8 a large effect; these thresholds should not replace subject-matter knowledge. The sign of h shows the direction of the difference. It is the most widely used effect size for proportions in power analysis. For readers, the raw difference, the risk ratio or the odds ratio may be easier to understand, and they can be reported alongside h.
An everyday example
Before a campaign 50% of customers knew a product; a survey after the campaign puts the figure at 65%. Cohen's h = 2·arcsin(√0.65) − 2·arcsin(√0.50) ≈ 0.30, an effect between small and medium.
In MerQur
In the binomial test, the 'Effect size' checkbox gives the raw difference p̂ − p₀ together with Cohen's h = 2·asin(√p̂) − 2·asin(√p₀).
Analyses where this concept appears
Related concepts
Effect sizeConfidence interval for a proportion (Wilson, Clopper-Pearson, Agresti-Coull, Jeffreys)