Page's Trend Test (ordered repeated measures)
Coming soonNonparametric ANOVA
This test is implemented and is currently going through StatMinds’ production verification: every statistic is independently checked against a trusted reference (scipy / R), locked with regression tests, and the screen is exercised across assumption-met/violated and significant/non-significant scenarios before it opens up.
See what’s live nowPaired analogue of the Jonckheere-Terpstra trend test.
Detects a MONOTONE trend across ≥ 3 ordered within-subjects conditions. Computes L = Σ c_j · T_j where c_j = (1, 2, …, k) are the a-priori condition scores and T_j is the column-j rank-sum within subject, then refers a Z-transform to the standard normal. Reports L, Z, p, and per-condition within-subject median ranks. Typically 10-30% more powerful than Friedman when the within-subject ordering is correctly specified; no power against non-monotone within-subject trajectories.
Worked example
Do repeated conditions follow a predicted order?
Page's trend test checks for a pre-specified ordered trend across matched (repeated) conditions — the ranked-trend counterpart to Friedman.
Rankings followed the predicted order, Page's L = 148, p = .006.
Page's trend test confirmed the predicted ordered trend across conditions, L = 148, p = .006.
When to use it
- Longitudinal repeated measures with a priori orderingA clinical study measures depression score on the same 20 patients at baseline, 1 month, 3 months, and 6 months post-treatment.
- Within-subjects ordered dose / intensityPain-tolerance study: 25 subjects exposed to 4 ascending stimulus intensities (counterbalanced order).
When NOT to — use instead
- Independent groupsPage is paired/within-subjects. → Jonckheere-Terpstra (ordered trend)
- Unordered repeated measuresPage requires ORDERING. → Friedman Test (non-parametric RM)
- Only 2 repeated measurementsPage is k ≥ 3. → Wilcoxon Signed-Rank
- Need covariate adjustmentPage is unadjusted. → Linear Mixed Effects (LMM) \u2014 clustered / repeated data (ICC, BLUPs)
Hypotheses
Parameter tested: monotone trend across the k ordered within-subjects conditions
Assumptions (and what to do if they fail)
Check: Tabulate missing-per-condition; report n_dropped.
If violated: Subjects with missing conditions are dropped by listwise deletion, reducing power and potentially biasing if missingness is not random.
Check: Report n_subjects and k.
If violated: Small n makes the asymptotic Z approximation coarse.
Ready to run a Page's Trend Test (ordered repeated measures) on your own data?
Guided setup, automatic assumption checks, effect sizes, figures and an APA write-up.
Run this test →