Compare two independent groups when your data are not normally distributed or are ordinal. This non-parametric test works on ranks rather than raw means, so it is robust to skew and outliers.
How to use it
Paste the values for each group, separated by commas or spaces. The calculator ranks all the values, computes the U statistic and gives you a z-score and p-value using the normal approximation.
How to read the result
A p-value below 0.05 means the two groups differ significantly in their distributions. This is the go-to test when a t-test’s normality assumption does not hold — for example with small, skewed samples or Likert-type data.
Rather have it done for you?
Send us your data and a qualified statistician will run the analysis and write it up. Fixed price, by email.
Frequently asked questions
When should I use Mann-Whitney instead of a t-test?
Use it when your outcome is ordinal, or when a numeric outcome is clearly non-normal, especially in small samples. It compares ranks rather than means, so outliers and skew matter less.
How many values do I need?
At least a few per group to compute anything, but the normal approximation used here is most accurate when each group has more than about 20 values. For very small samples, exact tables are preferable.
Does it handle tied values?
Yes. Tied values receive average ranks, which is the standard way to handle ties in this test.