A worked example, rendered from real sample data. Sign in to run the tool on your own input.
0.5
5
80═══ Verdict ═══
✓ 64 per group (128 total) for d = 0.50 at 80% power
ℹ d = 0.50 is a medium effect (0.2 small, 0.5 medium, 0.8 large)
═══ Inputs ═══
Effect size (Cohen's d): 0.500
Alpha (significance): 5.00% two-tailed
Power (1-β): 80.0%
Test: independent two-sample t-test, equal group sizes
═══ Result ═══
Sample size per group: 64
Total sample size: 128
Achieved power at that n: 80.2%
With dropout allowance: 64 per group
Degrees of freedom: 126
Starts from n = 2(z₁₋α/₂ + z₁₋β)²/d² and then steps up until the t-based power reaches the target.
Power uses a normal approximation to the noncentral t distribution (within roughly 1 percentage point of exact for n ≥ 10).
═══ Power Curve ═══
n= 9 │████······················ 15%
n= 18 │████████·················· 30%
n= 27 │███████████··············· 43%
n= 36 │██████████████············ 55%
n= 45 │█████████████████········· 65%
n= 54 │███████████████████······· 73%
n= 63 │█████████████████████····· 80%
n= 64 │█████████████████████····· 80%
n= 72 │██████████████████████···· 85%
n= 81 │███████████████████████··· 89%
n= 90 │████████████████████████·· 92%
n= 99 │████████████████████████·· 94%
n= 108 │█████████████████████████· 96%
n= 117 │█████████████████████████· 97%
n= 126 │█████████████████████████· 98%
Each row is the chance of detecting the effect at that per-group sample size.
═══ What This Buys You ═══
If a true difference of d = 0.50 standard deviations really exists, this sample size gives you a 80% chance of ending up with a significan
…
Calculate required sample size for statistical tests. Part of the DevTools Surf developer suite. Browse more tools in the Statistics collection.