Small Sample Properties of the Wilcoxon Signed RankTest with Discontinuous and Dependent Observations

This Monte-Carlo study investigates sensitivity of the Wilcoxon signed rank test to certainassumption violations in small samples. Emphasis is put on within-sample-dependence,between-sample dependence, and the presence of ties. Our results show that both assumptionviolations induce severe size distortions and entail power losses. Surprisingly, theseconsequences do vary substantially with other properties the data may display. Resultsprovided are particularly relevant for experimental settings where ties and within-sampledependence are frequently observed.