Three or more measurements on the same people, with an outcome that is ordinal or badly behaved. How the ranking works within each person, why that is the point, and how to report it in APA 7.
The Friedman test compares three or more measurements taken from the same people. Three time points. Four conditions each participant experienced. Five items each judge rated.
The mechanism is what makes it useful. Ranking happens within each person, not across the whole sample. Each participant's three or four scores are ranked among themselves, and the test then asks whether those within-person rankings line up consistently across participants.
That design removes the differences between people entirely. A participant who scores low on everything and one who scores high on everything contribute the same information: the order of their own conditions. It is the rank-based counterpart to repeated-measures ANOVA, and it is why the test works on ordinal outcomes where averaging would be indefensible.
The result header names the design directly, as a count of conditions and subjects, which is the pair of numbers that determines everything else.
Four figures follow: chi-square, df, p, and Kendall's W as the effect size.
Degrees of freedom are the number of conditions minus one. Three conditions gives df = 2 whether you have 20 participants or 250.
Kendall's W is the coefficient of concordance and has a genuinely useful interpretation: it runs from 0 to 1 and measures how much the participants agree in how they ranked the conditions. W near 1 means nearly everyone ordered the conditions the same way. W near 0 means the orderings are effectively random. A result of chi-square = 0.29, df = 2, p = .866, W = 0.001 says there is no consistent ordering at all, which is a clearer statement than the p value alone provides.
A significant Friedman test says the conditions are not all alike. It does not say which pair differs.
Reach for it when the outcome is ordinal, such as ranked preferences or single Likert items, or when the distributions are too skewed for repeated-measures ANOVA to be defensible.
Complete cases only. A participant missing any one of the measurements contributes nothing, because their within-person ranking cannot be formed. In a study with several time points this can remove a substantial share of the sample, and who it removes matters more than how many.
The repeated measurements must be genuinely repeated. Same people, same order of conditions, one row per person.
For exactly two repeated measurements, use Wilcoxon signed-rank. For three or more separate groups rather than repeated measures, use Kruskal-Wallis.
Lay the data out in wide format. One row per person, one column per measurement or condition. Three time points means three columns.
Open Nonparametric & Categorical and select Friedman from the test list, then configure the measurement columns in the Setup panel on the right. Results appear in the center.
Read the missing-cell count with care. On a repeated-measures design, a single missing cell removes that entire participant.
Check the subject count in the result header against the rows you imported. The difference is the part of your sample the result does not describe.
Use the action bar to add the result to a report, copy the table, or copy the APA text.
Report chi-square with its degrees of freedom, the number of complete cases, p, and Kendall's W. Give medians for each condition.
A Friedman test found no difference in ratings across the three conditions, χ²(2) = 0.29, p = .866, Kendall's W = .001, n = 250.
Report Kendall's W. It answers a question the chi-square does not: whether participants agreed on an ordering, regardless of whether the differences reached significance.
State the number of complete cases and how many participants were excluded for incomplete data. On repeated measures this is not a footnote; it is the definition of who your finding is about.
Every result carries a ReliCheck Intelligence card with one action: explain this result in plain language. It runs entirely on your Mac with Apple Intelligence. No upload, no API key, nothing to configure. The model only explains the result Quanta has already computed and validated. It never computes a statistic, and it never applies a decision on your behalf.
From the action bar you can add the result to a report, copy the table, or copy the APA text. Exported reports carry the APA table, its note, and a caution specific to that analysis beneath it.
Quanta's correlation and nonparametric procedures are validated against R's cor and against hand-computed closed-form values, matched exactly. The engine is pure Swift with no R or Python at runtime. Full record on the validation page.
Nonparametric is part of every Quanta subscription, $9.99 a month or $79.99 a year after the 30-day trial. One price for everyone. See pricing.
The analysis, the report builder, and the on-device explanations all work with the network off. Your dataset is never uploaded.
Wilcoxon signed-rank for exactly two repeated measurements · Kruskal-Wallis for three or more separate groups · One-way ANOVA for separate groups with a well-behaved outcome · the complete list on the analyses page.