Use the Friedman Test for Repeated Measures in a Thesis

Match the test to the design

The Friedman test compares at least three related conditions or time points. The same people or matched blocks contribute observations in every condition. Values are ranked within each block; different blocks should be independent. It is often considered for ordinal outcomes or repeated measures that do not suit a parametric approach.

Example: Twenty-four students rate the same learning app after three teaching formats. Three ratings from one student belong together. Three independent groups of twenty-four students would require another analysis. State a hypothesis about the linked conditions rather than treating repeated ratings as unrelated cases.

Inspect data before calculation

Check that each analysed person has a valid value for every condition. Some implementations drop an entire case when one measurement is missing. Report the actual analysed number and consider why data are absent. Equal values, or ties, are common on rating scales; make sure your software handles them appropriately.

Choose the outcome and conditions before examining p-values. Within-person ranks describe an ordering of ratings, not equal distances between original scale points. Present distributions or medians for each condition and explain the coding so readers can understand the direction of the result.

Report overall and pairwise findings

A significant Friedman result indicates that the conditions do not all behave alike. It does not identify which pair differs. For specific pairwise claims, use suitable matched follow-up tests and control multiple comparisons. Select comparisons from the research question rather than treating every post-hoc contrast as equally planned.

Report the test statistic, degrees of freedom, p-value and analysed sample. Add a relevant effect description, such as Kendall's W, where your software computes it appropriately. Interpret practical meaning in the context of the scale; a small p-value alone does not imply a large difference.

Keep conclusions within the design

The test detects a pattern; it does not reveal its cause. Order, practice and fatigue may influence repeated ratings. State whether condition order was fixed or randomised. In a longitudinal study, a time difference may coincide with other changes that the test cannot separate.

Ensure the table, prose and hypothesis describe the same direction. Say which condition received higher ratings without presenting a rank contrast as an exact difference on the original scale. Keep the analysis decisions and output so the final numbers can be checked.

Frequently asked questions