Statistics Glossary → Mood's median test
Mood's median testconcept
A nonparametric test of whether two or more independent groups share the same median, based on whether observations fall above or below the overall median.
First the median of all groups combined is found. Then the observations in each group are counted as 'above the overall median' or 'at or below it', giving a 2 × k table. If the groups share the same median, roughly half of each group should fall above and half below; the table is tested with Pearson's chi-square test.
Because it only looks at which side of the median a value falls, it is very robust to extreme values, but since it does not use the full information in the ranks it is usually less powerful than the Kruskal-Wallis test. How values equal to the median are counted can affect the result, especially with many ties. With very small groups the expected frequencies drop and the chi-square approximation weakens.
An everyday example
Rents in three districts are compared. The combined median of the three districts is 15 000. In the first district 70% of homes are above that value, in the other two only 40% and 35%. The median test shows that this imbalance cannot be explained by chance, and a few very expensive villas have no effect on the result.
In MerQur
The 'Extra outputs: Median test' checkbox in the Kruskal-Wallis test; it gives Pearson's χ² without the Yates correction on the 2 × k table of values above versus at or below the overall median.
Analyses where this concept appears
Related concepts
Score and scale tests (Van der Waerden, Savage, Siegel-Tukey, Ansari-Bradley, Mood, Klotz)