Jonckheere-Terpstra Test for Ordered Alternatives
Also known as: Jonckheere-Terpstra Testi, JT test, ordered k-sample test, trend test for ordered groups
The Jonckheere-Terpstra test is a nonparametric hypothesis test that detects a monotone trend across k ordered groups — testing whether the outcome rises (or falls) systematically as the group order increases. Developed independently by T. J. Terpstra (1952) and A. R. Jonckheere (1954), it is the directional, ordered-alternative counterpart to the Kruskal-Wallis test.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use the Jonckheere-Terpstra test when you have three or more groups that follow a meaningful natural order (e.g., dose levels, education tiers, severity stages) and you hypothesise a monotone trend in the outcome. The outcome may be continuous or ordinal. The two key assumptions are: (1) observations are independent both within and between groups, and (2) the group ordering reflects a substantive a priori hypothesis — the test exploits this ordering, so it must be specified before inspecting the data. With only two ordered groups the Mann-Whitney U test is equivalent and preferred. When group ordering is irrelevant or unknown, use the Kruskal-Wallis test instead.
Strengths & limitations
- More statistically powerful than the Kruskal-Wallis test when a monotone trend is the true alternative, because it focuses the rejection region on ordered alternatives.
- Distribution-free: requires no normality or equal-variance assumption, making it robust to skewed distributions and outliers.
- Handles both continuous and ordinal outcome variables.
- Accommodates tied observations through the standard tie-correction in the variance formula.
- Only detects monotone (unidirectional) trends; it has no power against non-monotone patterns such as a U-shape or inverted-U.
- The group ordering must be specified a priori; if chosen after inspecting the data, the p-value is not valid.
- Minimum sample size guidance (n ≥ 15 total) is approximate; very small groups reduce the accuracy of the normal approximation.
- Rejects H₀ for the overall trend but does not identify which adjacent group pairs drive the difference — pairwise follow-up is needed.
Frequently asked
How is this different from the Kruskal-Wallis test?
The Kruskal-Wallis test is an omnibus test that detects any difference among k groups, regardless of direction. The Jonckheere-Terpstra test specifically tests for a monotone ordered trend (group 1 ≤ group 2 ≤ ... ≤ group k). When the true pattern is a monotone trend, the Jonckheere-Terpstra test is more powerful; when the pattern is non-monotone or the ordering is unknown, Kruskal-Wallis is the right choice.
Does the group ordering have to be pre-specified?
Yes — this is essential. The Jonckheere-Terpstra test's directional alternative is only valid when the order of groups was defined before the data were examined. Choosing the ordering based on the observed group means inflates the false-positive rate and invalidates the p-value.
What should I do after the test is significant?
The test identifies an overall monotone trend but does not pinpoint where the trend is concentrated. Follow up with pairwise Mann-Whitney U tests between adjacent groups (applying a Bonferroni or Holm correction), or compute Kendall's tau-b between the group index and the outcome as a descriptive measure of trend strength.
Can I use this test with ordinal (Likert-scale) data?
Yes. The Jonckheere-Terpstra test operates on ranks and does not require interval-level measurement, so it is appropriate for ordinal outcome variables such as Likert scales, severity ratings, or ranked scores, provided the groups themselves are also ordered.
Sources
- Jonckheere, A. R. (1954). A distribution-free k-sample test against ordered alternatives. Biometrika, 41(1-2), 133–145. DOI: 10.1093/biomet/41.1-2.133 ↗
- Terpstra, T. J. (1952). The asymptotic normality and consistency of Kendall's test against trend, when ties are present in one ranking. Indagationes Mathematicae, 14, 327–333. link ↗
- Hollander, M., Wolfe, D. A., & Chicken, E. (2014). Nonparametric Statistical Methods (3rd ed.). Wiley. ISBN: 978-0470387375
How to cite this page
ScholarGate. (2026, June 1). Jonckheere-Terpstra Test for Ordered Alternatives. ScholarGate. https://scholargate.app/en/statistics/jonckheere-terpstra
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Friedman testStatistics↔ compare
- Kruskal-Wallis testStatistics↔ compare
- Mann-Whitney U testStatistics↔ compare
- One-way ANOVAStatistics↔ compare
- Spearman CorrelationStatistics↔ compare