Effect sizes describe how big a difference or relationship is, and different fields use different ones. The effect size converter takes Cohen’s d, a correlation r, an odds ratio, eta squared or Cohen’s f, and gives all the others, together with Cohen’s U3, the probability of superiority and the overlap of the two distributions.
How the calculation works
- Every conversion goes through Cohen’s d, the difference between two group means in standard deviation units. Choose which measure you have and enter its value.
- r = d ÷ √(d² + 4) (equal group sizes) and its inverse d = 2r ÷ √(1 − r²). η² = r² and f = d ÷ 2.
- The odds ratio is e^(d × π ÷ √3), and the inverse is d = ln(OR) × √3 ÷ π. This assumes a logistic distribution behind the outcome.
- Under normal distributions with equal spreads, U3 = Φ(d), the probability of superiority = Φ(d ÷ √2) and the overlap = 2 × Φ(−|d| ÷ 2), where Φ is the standard normal distribution.
Example
Take a Cohen’s d of 0.5. The correlation is r = 0.5 ÷ √(0.25 + 4) = 0.243, so η² is 0.059 and Cohen’s f is 0.25. The odds ratio is e^(0.5 × π ÷ √3) = 2.477. A typical person in group 1 is above 69.1 % of group 2 (U3), the chance that a random person from group 1 beats one from group 2 is 63.8 %, and the two distributions overlap by 80.3 %.
Why it matters
A study may report a correlation, another an odds ratio and a third a difference of means. Putting them on one scale allows comparison, the planning of a power analysis from published results, and the combination of results in a meta-analysis.
Good to know
The conversions rest on the assumption of normal distributions with equal variances and equal group sizes, so they are approximations for real data. The odds ratio conversion uses the logistic approximation, and the sign of d is kept for r and the odds ratio, while η², f and the overlap are shown without sign.
Related tools
The Cohen’s d calculator computes d from means and standard deviations, and the statistical power calculator links the effect size to sample size.