Mann–Whitney U
§2.4
Rev 4.6 · reviewed Feb 2026Mann–Whitney U
The nonparametric partner of the independent t-test — use it for skewed data, ordinal scales, or small samples. It compares ranks, not means, so it doesn't assume normality.
Source formula: Mann HB, Whitney DR. Ann Math Stat 1947;18:50–60.
Group 1 values
Group 2 values
| n (group 1 / group 2) | 7 / 7 |
| Rank sum, group 1 | 36.0 |
| U statistic | 8.0 |
| z (normal approx, tie-corrected) | -2.113 |
* Normal approximation with tie correction. Mann HB, Whitney DR. Ann Math Stat 1947;18:50–60.
Two-sided p
0.035▲
copyThe two groups' distributions differ significantly in rank (p 0.035). Mann–Whitney compares whether one group tends to have higher values, not the difference in means — report medians and IQRs alongside it.
Understand what this is measuring