Optimising changeover designs using the average efficiency factors

In a changeover design, each subject receives a sequence of treatments over consecutive time periods. We provide simple recursive formulae for updating the average efficiency factors of both direct and residual treatment effects, after two treatments are interchanged. We have incorporated these formulae into two interchange algorithms. From a comparison with designs generated from other algorithms using surrogate objective functions, we show that these recursive methods generally produce more efficient designs more quickly.