A k-sample test for trend that is the analogue of the Mann–Whitney test. The test was suggested independently by Jonckheere (in 1954) and Terpstra (in 1952). The assumption is that all k populations have the same continuous distribution, but that the population means may change (monotonically) during the ordered sequence of samples. Let yij denote the jth observation from the ith population. Let nij be the number of observations in the previous (i-1) samples that are smaller than yij. Define , then the test statistic is J, given by J=2S-M, where M is the maximum possible value for S.
Subjects: Probability and Statistics.