Work out statistical power for study instantly with clear inputs, formula shown and shareable results.
Power is the probability of detecting a real effect. For d = 0.5 with 64 per group at alpha 0.05, power is about 80% - exactly why 64 per group is the classic textbook figure for a medium effect. The type II error rate of 20% means one study in five would miss a genuine effect of that size.
Power
power = P(|Z| > z(alpha/2) | true effect), with non-centrality d x sqrt(n/2)
Sample for 80% power
n per group = 2((z(alpha/2) + z(0.80)) / d)^2
It balances the cost of a larger sample against the risk of missing an effect. Some fields now expect 90% or 95%.
They miss real effects and, when they do reach significance, overestimate the effect size - a major driver of the replication crisis.