A Mann-Whitney test indicates whether values tend to be larger in one of two columns. If the values are drawn from distributions having the same shape but different locations, the Mann-Whitney test can be interpreted as a test of whether the column medians are equal.

To perform a Mann-Whitney test:

- In the Summary view, use the confidence slider to choose a desired confidence level
- Command-click two numeric columns of interest on the left
- Click the button labeled
*Boxes*below the Covariate view - The result of the Mann-Whitney test will appear in the Bottom Line

Note that even if the observed medians are equal across categories, the Mann-Whitney test can indicate that the population medians are not equal. This is most likely due to the distribution of values differing across columns.

To analyze differences in three or more numeric columns’ medians, use a Kruskal-Wallis test.

See also:

- Performing a Mann-Whitney test of a numeric column’s median across two categories
- Performing a Kruskal-Wallis test of three or more numeric columns’ medians
- Performing a Wilcoxon signed-rank test of the median difference between two numeric columns
- Performing an unpaired t-test of two numeric columns’ means
- Performing a Kolmogorov-Smirnov test of two or more numeric columns’ distributions
- Viewing the correlation coefficient between two numeric columns
- Formatting descriptive and inferential statistics
- About the Summary view