统计功效分析计算器
计算统计检验的功效或确定所需的样本量。
Sample Size per Group
31.40
Sample Size per Group vs Effect Size (Cohen's d)
公式
How to Perform a Power Analysis
Formula
n per group = ((z_alpha/2 + z_beta) / d)^2
Power is the probability of correctly rejecting a false null hypothesis (detecting a real effect). This formula estimates the sample size needed per group for a two-sample test. Higher power, smaller effect sizes, and lower alpha all require larger samples.
计算示例
Detect a medium effect (d = 0.5) with 80% power at alpha = 0.05.
- 01n = ((1.96 + 0.842) / 0.5)^2
- 02= (2.802 / 0.5)^2
- 03= 5.604^2
- 04= 31.4
- 05Round up: 32 per group, 64 total
常见问题
What is an acceptable power level?
80% (0.80) is the conventional minimum. Many researchers aim for 90% power. Lower power means a higher risk of failing to detect a real effect (Type II error).
What is the relationship between alpha, power, and sample size?
They are interconnected: fixing any two determines the third (given the effect size). Decreasing alpha or increasing power both require larger sample sizes.
What if the effect size is unknown?
Use pilot study estimates, published benchmarks (small=0.2, medium=0.5, large=0.8), or determine the minimum clinically important difference and convert it to d by dividing by the expected SD.
学习