ABSTRACT
This chapter aims to establish some basic tools that will enable the user to analyze any cross-over design for which the effects are estimable. The two major tools are: a method of coding, by which any cross-over design can be prepared for parameter estimation and subsequent statistical analysis and a measure of “separability” of treatment and carryover effects in a model. The chapter deals with a variety of other issues and principles which are necessary in order to analyze cross-over trials in a systematic and correct way, making appropriate use of statistical testing procedures. Cross-over designs are examples of designs in which the effects to be estimated are, in general, not “orthogonal”. There are several major reasons for concern about one’s ability to separate direct treatment effects from carryover effects. Both reasons concern the interpretability of the results of the analysis of variance of a cross-over design.
