In independent two-sample design, the most popular effect size measure is defined as

$$ \delta = \frac{\mu_1 - \mu_2}{\sigma} $$

where $\mu_1$ and $\mu_2$ are the population means of the two population and $\sigma$ is the population standard deviation which is assumed to be the same for the two populations.

Although Cohen's d is the most widely used effect size estimator in samples, it is biased. An unbiased estimator, often referred to Hedges' g, was developed by Hedges (1981). It corrects the Cohen's d in the following way:

$$ g = d\times \frac{\Gamma(m/2)}{\sqrt{(m/2)} \Gamma[(m-1)/2]} $$

where $m = n_1 + n_2 - 2$.

Hedges (1981) showed that the variance of $g$ is

$$ var(g) = \frac{m(1+\tilde{n} \delta^2)}{(m-2)\tilde{n}} - \frac{\delta^2}{c^2} $$

where $\tilde{n}=n_1 n_2/(n_1+n_2)$ and

$$c=\frac{\Gamma(m/2)}{\sqrt{(m/2)} \Gamma[(m-1)/2]}$$.

With it, a confidence interval for $g$ can be constructed as

$$g \pm \Phi^{-1}(1-\alpha/2) \sqrt{\widehat{var(g)}}$$

where $\sqrt{\widehat{var(g)}}$ is an estimate of $var(g)$ by replacing $\delta$ with Cohen's d.

• Normally distributed data
• Equal variances

1. Cohen's d
2. Effect size of independent two samples

Hedges, L. V. (1981). Distribution theory for Glass’s estimator of effect size and related estimators. Journal of Educational Statistics, 6(2), 107–128.