Carry-over effects when using crossover designs

John Stufken, Professor and Head, Univ of Georgia, March 12, 2010

One of the most controversial topics when using a crossover design is how to deal with possible carryover effects. While there is broad agreement that one should attempt to use sufficiently long washout periods, this may not always be possible and it is not always clear what sufficiently long really means. A growing number of researchers recommend that a crossover design should not be used if there is a possibility of carryover effects, with or without a washout period, that persist into the next active treatment period. Others have proposed more realistic models for modeling possible carryover effects. This talk will provide a closer look into these issues and review some of the proposals for alternative models. The talk will also consider efficient designs for different scenarios.