Power explorer: two-group comparison

Change the true effect and the sample size, then run some studies.

Sampling distributions of the observed effect size under the null hypothesis and under the true effect, with the critical values and the area that corresponds to statistical power.

Each dot will be one simulated study: blue = significant, grey = not significant.

Power (1 − β)
chance to detect the effect
False positives (α)
if there were no effect
Detectable |d|
smallest significant estimate
n for 80% power
per group, at this d and α

Power by sample size

Two-sided test, normal approximation (very close to pwr::pwr.t.test() except for very small samples).