1. Definition
- The Mann–Whitney U test (also called the Wilcoxon rank-sum test) is a non-parametric test used to compare two independent groups.
- It answers: “Do these two groups come from the same distribution?”
- Unlike the t-test, it does not assume normality. Instead, it compares the ranks of the data.
2. Assumptions
- The two samples are independent.
- The dependent variable should be ordinal or continuous.
- If distributions are similar in shape, the test is often interpreted as comparing medians.
3. Test Statistic (U)
- Combine data from both groups and assign ranks.
- Compute the sum of ranks for each group ($R_1, R_2$).
- Calculate:
$U_1 = n_1 n_2 + \frac{n_1(n_1+1)}{2} – R_1$
$U_2 = n_1 n_2 + \frac{n_2(n_2+1)}{2} – R_2$
$U = \min(U_1, U_2)$
- Large samples → $U$ is approximated by a normal distribution (z-test).
4. Hypotheses
- Null hypothesis (H₀): The two groups come from the same distribution.
- Alternative hypothesis (H₁): One group tends to have larger (or smaller) values than the other.
5. Example
Suppose we test exam scores:
- Group A: 88, 92, 100, 75, 85
- Group B: 60, 70, 65, 80, 72
We rank all values, compute rank sums, calculate $U$, and check the p-value.
- If p < 0.05 → reject H₀ → groups differ significantly.
6. Connection to ROC-AUC
- The Mann–Whitney U statistic is mathematically equivalent to ROC-AUC.
- Interpretation: the probability that a randomly chosen observation from Group A has a higher score than a randomly chosen one from Group B.
Summary:
The Mann–Whitney U test is a non-parametric alternative to the t-test.
It checks whether two independent groups differ in distribution by comparing ranks.
It’s especially useful when the data are not normally distributed or when you only care about relative ordering.
