Nonparametric tests are said to have what percentage of the power of their parametric counterparts on average?

• A. 95.5%
• B. 100%
• C. 85%
• D. 75%

Answer: (A)

Nonparametric tests rely on ranks rather than raw scores, which makes them robust to non-normal data but slightly less efficient when the data are actually normal. The standard benchmark is the comparison between the Wilcoxon rank-sum test and the two-sample t-test under normal data, where the asymptotic relative efficiency is about 0.955. This means the nonparametric test has roughly 95.5% of the power of its parametric counterpart for the same sample size. In practical terms, you’d need a bit more data to achieve the same power, about 1/0.955 ≈ 1.05 times as many observations. So, on average, nonparametric tests retain most of the power—around 95.5%—while offering robustness to non-normality and outliers.