Title: Experimental Designs in Cross-over Trials
Abstract: In a parallel group trial, the experimental units are randomized to receive one treatment each. If, instead of individual treatments, sequences of treatments are randomized to be given to experimental units over a number of periods, it is called a cross-over trial. Such trials occur in a number of fields such as pharmaceutical studies for testing new drugs, psychological studies, sensory evaluations, agricultural experimentation and consumer research. The sequences applied to experimental units in these trials vary depending upon the number of treatments to be tested, duration of the extent of the experiment, and other resources available. However, the aim of such trials still remains same – to compare the effects of individual treatments (and sometimes, of the sequences as a whole).
Experimental designs formulated to suit cross-over trials are known as cross-over designs. Their structural formulation makes them advantageous in cases where the experimental units are expected to exhibit large variability, they are either scarce or are supposed to be highly trained, or if the objective is to study some trend in the observations on the units by successive application of a number of treatments. These designs, however, are susceptible to the occurrence of carryover effect(s) of treatment(s) applied in one period, in the period next to it, due to the successive application of treatments on a unit.The aim of this talk is to indulge in certain aspects of cross-over designs in terms of their model, which includes the special feature of carryover effect, and relevant construction strategies, so as to fine-tune them for the difference in areas of their application.