Sample size to estimate proportions
The minimum sample size required to estimate the proportions $p_1-p_2$ is
$$ n =\big[p_1*(1-p_1)+p_2*(1-p_2)\big]\bigg(\frac{z}{E}\bigg)^2 $$
where
- $p_1$ is the proportion of successes in first group,
- $p_2$ is the proportion of successes in second group,
- $z$ is the $Z_{\alpha/2}$ and
- $E$ is the margin of error.