STATISTICSInference & TestsStatistics Calculator
📊

Power Analysis Calculator

Free power analysis calculator. Sample size, power, effect size for t-tests, proportions, ANOVA, cor

Run CalculatorExplore data analysis and statistical calculations

Why This Statistical Analysis Matters

Why: Statistical calculator for analysis.

How: Enter inputs and compute results.

1−β
POWERCohen 1988 • G*Power • NIST • Penn State

Power Analysis — Sample Size & Power

Solve for sample size, power, effect size, or α. Two-sample t, one-sample t, proportion, ANOVA, correlation. Interactive power curves.

Real-World Scenarios — Click to Load

Test Configuration

power_analysis.sh
CALCULATED
$ power_analysis --test="two-sample-t" --solve="sample-size"
Primary Result
63
Required n per group
Details
Total n = 126
α
0.05
Power
66.90%
Critical value z*
±1.9600
Noncentrality δ
2.8062
Actual power at n
66.90%
Share:
Power Analysis Result
Required n per group
63
Total n = 126
numbervibe.com/calculators/statistics/power-analysis-calculator

Calculation Breakdown

INPUTS
z_α/2 (critical value)
1.9600
α=0.05, two-tailed
z_β (power)
0.8416
power=0.8
COMPUTATION
z_α/2 + z_β
2.8016
n per group
63
n = 2((z_α/2 + z_β)/d)² = 2(2.80/0.5)²

Power Curve vs n

Power vs Effect Size (Cohen d)

Sample Size Comparison (α=0.05, power=0.8)

EffectCohen dn per groupTotal n
Small0.2393786
Medium0.563126
Large0.82550

For educational and informational purposes only. Verify with a qualified professional.

Key Takeaways

  • Power = 1 − β = P(reject H₀ | H₁ true). Target typically 0.8 (Cohen 1988).
  • Two-sample t: n per group = 2((z_α/2 + z_β) / d)². d = Cohen's d.
  • One-sample t: n = ((z_α + z_β) / δ)². δ = (μ₁ − μ₀)/σ.
  • Proportion: n = (z_α√(p₀(1−p₀)) + z_β√(p₁(1−p₁)))² / (p₁−p₀)².
  • Correlation: n ≈ ((z_α + z_β) / z_r)² + 3. z_r = Fisher z-transform of r.
  • ANOVA: n per group depends on Cohen f and number of groups.

Power Analysis Workflow

1. Define parameters

α (significance level, usually 0.05), desired power (usually 0.8), effect size (d, r, f, or proportion difference).

2. Choose effect size

Use Cohen's conventions (d: 0.2=small, 0.5=medium, 0.8=large) or prior literature. Never guess optimistically.

3. Solve for unknown

Given two of {n, power, effect size, α}, solve for the third using the appropriate formula.

4. Account for attrition

Increase n by 10–20% if dropout is expected.

5. Pre-register

Document your power analysis plan before data collection.

Cohen Effect Size Guidelines

EffectCohen drCohen f
Small0.20.10.1
Medium0.50.30.25
Large0.80.50.4

Did You Know?

💊Clinical trials typically target 80% power to detect clinically meaningful effects.Source: Clinical Research
📱A/B tests often use proportion tests. Small conversion differences need large samples.Source: Product
📊Cohen d: 0.2=small, 0.5=medium, 0.8=large. Use for sample size planning.Source: Cohen 1988
🔬Underpowered studies risk false negatives. Pre-register power analysis.Source: Methodology
📈Power curves show how power increases with sample size or effect size.Source: Visualization
🗳️Political polls use proportion power analysis for margin of error.Source: Survey Design

Expert Tips

Target 80% power

Most studies aim for 80% power. Higher power needs larger samples.

Use realistic effect sizes

Base d, r, or proportion difference on prior literature or pilot data.

Two-sided tests

Default α/2 for two-tailed. Use one-sided only when justified.

Account for attrition

Increase n by 10–20% if dropout is expected.

Frequently Asked Questions

What is statistical power?

Power = probability of correctly rejecting H₀ when H₁ is true. Low power means high risk of false negatives.

Why target 80% power?

Convention balances cost (sample size) vs risk of missing real effects. Some fields use 90%.

What is Cohen d?

Standardized mean difference: (μ₁ − μ₂)/σ. Measures effect size in standard deviation units.

How do I choose effect size?

Use prior studies, meta-analyses, or define smallest clinically/practically meaningful difference.

What if my study is underpowered?

Increase n, use a larger effect size if justified, or consider a one-sided test when appropriate.

Power Analysis by the Numbers

0.8
Typical power target
0.05
α (significance)
d=0.5
Medium effect
n≈64
Per group (t-test)

Disclaimer: These formulas use normal approximations. For small samples, use t-distribution. Chi-square uses noncentral distribution. Consult G*Power or statistical software for exact values.

👈 START HERE
⬅️Jump in and explore the concept!
AI

Related Calculators