Dropout in crossover designs

Rebecca Lowdwick (University of Portsmouth, UK)

Participant dropout in crossover designs may result in one or more treatment contrasts being non-estimable in the implemented design. Via a simple graphical method for determining whether a two-period crossover design is connected, we introduce a criterion for selecting those crossover designs which are less likely to become disconnected due to participant dropout. Additionally, we explore how methodological aspects of crossover designs are reported in practice in the general scientific literature.