StatReasonNonparametric Statistics
Nonparametric methods compare groups or paired measurements using the ranks of your data rather than its raw values and a normal-distribution assumption. They are the safer choice when your data is skewed, contains extreme values, is ordinal rather than truly numeric, or your sample is too small to trust a normality assumption.
Every method here names its exact/asymptotic mode and its tie-handling convention directly, since these choices change the result. A rank-based test on its own describes a difference in distribution, not automatically a difference in mean or median, unless the two groups being compared share a similar shape.
Engines in this category
- Mann-Whitney U Test
Compare two independent groups without assuming a normal distribution.
- Kruskal-Wallis Test
Compare three or more independent groups using ranks, not raw values.
- Wilcoxon Signed-Rank Test
Compare matched pairs without assuming a normal distribution of differences.
- Sign Test
Test direction-only evidence from paired data or a single median comparison.
- Friedman Test
Compare three or more related conditions using ranks within each block.
- Mood's Median Test
Compare group medians with a low-power, low-assumption test.
- Jonckheere-Terpstra Trend Test
Test for an ordered trend across three or more independent groups.
- Dunn Post-Hoc Test
Make adjusted pairwise rank comparisons after a Kruskal-Wallis test.
- Rank-Biserial Correlation Calculator
Report a Mann-Whitney effect size directly from U and group sizes.
- Cliff's Delta Calculator
Report probability-based ordinal dominance between two groups.