“…1 Bartholomew's trend test Bartholomew's Ē2 k test [Bartholomew, 1959] for the simple ordered alternative H 1 : µ 1 ≤ µ 2 ≤ ... ≤ µ k |µ 1 < µ k in a fully randomised one-way layout with N (µ i , σ 2 ) (i = 1, ..., k) is a modification of the usual ANOVA F-test (which is sensitive to any heterogeneity between qualitative treatment levels), replacing the arithmetic mean estimators xi by the pool-adjacent-violators under order restriction (PAVA) estimates μi . This test belongs to the class of isotonic regression models, but for a single quantitative factor and not for a regression covariate or even dependent time series data (see the discussion for dependent data [Lei et al, 1995]). However, its use is severely limited by the lack of software (a permutative version is available in the CRAN package IsoGene [Lin et al, 2015]).…”