When might you use a nonparametric test?

• A. When data are ordinal or not normally distributed or the sample size is small.
• B. When data are perfectly normally distributed and continuous.
• C. When you want to compare means across groups with equal variances.
• D. When you know the population distribution.

Answer: (A)

Nonparametric tests are chosen when the usual parametric assumptions about the data can’t be met. They’re especially appropriate when data are ordinal (ranked) rather than truly interval/ratio, when the distribution is not normal or is heavily skewed, or when you have a small sample size where normality is hard to establish. Because nonparametric tests rely on ranks or medians instead of means and standard deviations, they’re more robust to outliers and distributional quirks. If the data are actually normally distributed and measured on an interval/ratio scale with a reasonable sample size, parametric tests generally offer more statistical power. Knowing the population distribution and having a good parametric model available would typically steer you toward those parametric tests rather than a nonparametric one.