Statistical power for two proportions
Power of a completed or planned two-group comparison of proportions.
Inputs
1 of 4%
%
per group
Alpha (two-sided)
When to use
- Planning, or judging whether a negative trial was large enough to detect a meaningful effect.
Pearls and pitfalls
- Power computed from the observed effect adds nothing to the P value; use the smallest clinically important effect instead.
Why use
Answers 'was this study big enough?' honestly.
z = (|p₁ − p₂| √n − z₁₋α/₂ √(2p̄q̄)) ÷ √(p₁q₁ + p₂q₂); power = Φ(z)