Overview
The Mann-Whitney U test checks whether values in one group tend to be higher than in another, without assuming the data follow a bell curve.
Paste the values for each group. The groups can be different sizes.
Worked examples
| Inputs | Result |
|---|---|
| Group A12, 15, 14, 19, 22, 17, 13, 16Group B21, 24, 18, 26, 23, 25, 20Significance level5% | p-value: 0.00654 (two-tailed)U: 4 (Group A: 4, Group B: 52)z: 2.72Group A: 8 valuesGroup B: 7 valuesSignificant at the 5% level: the groups likely differ.With small groups this normal approximation is rough, so check an exact table for close calls. |
How it works
All the values are ranked together, with tied values sharing their average rank. U counts how often a value in one group beats a value in the other. The p-value comes from a normal approximation with tie and continuity corrections, the same approach many statistics packages use for larger samples.
The formula
| R | Sum of the ranks in a group |
| n | Number of values in that group |
Practical tips
- Use it when your data are skewed, ranked or have outliers that would distort a t-test.
- It compares two separate groups. For before-and-after data on the same people, use a paired test instead.
- With small groups the normal approximation is rough, so check an exact table for close calls.
Frequently asked questions
Is this the same as the Wilcoxon rank-sum test?
Yes, the two tests are equivalent. They use different statistics but give the same p-value.
Does it compare medians?
Only if the two groups have a similar shape. In general it tests whether one group tends to produce higher values.
What does a small p-value mean?
That a difference this large would be unlikely if the groups really came from the same distribution.