Friedman Test
What you'll learn: You'll discover the nonparametric alternative to repeated measures ANOVA when you need to compare multiple related groups without assuming normal distributions.
The Core Idea
Imagine you're testing three different teaching methods on the same group of students, measuring their performance after each method. If your data violates the normality assumptions required for repeated measures ANOVA, the Friedman test comes to the rescue.
The Friedman test is a rank-based procedure that detects differences across three or more related groups (repeated measurements on the same subjects or matched groups). Instead of comparing actual scores, it ranks the observations within each subject across conditions, then tests whether these ranks differ systematically.
How It Works
For each subject (or block), you rank their responses across all conditions. Then you sum the ranks for each condition and compare these sums. If one treatment consistently receives higher or lower ranks across subjects, the test detects this pattern.
The test statistic follows approximately a chi-squared distribution when the null hypothesis (no differences among treatments) is true, allowing you to determine statistical significance.
When to Use It
- You have three or more related groups (repeated measures or matched subjects)
- Your data is ordinal or doesn't meet normality assumptions
- You've already learned the Kruskal-Wallis test (for independent groups) and need its counterpart for dependent samples
- Think of it as the nonparametric cousin of repeated measures ANOVA, just as Wilcoxon signed-rank was to the paired t-test
Key Takeaway: The Friedman test ranks observations within each subject across conditions to detect treatment differences in repeated measures designs without requiring normal distributions—your go-to when repeated measures ANOVA assumptions fail.