Comparison of Two Exponential Distributions

The problem of comparing two treatments with respect to a continuous response variable, following an exponential distribution, is considered. It is assumed that a relevant comparative clinical trial is planned to either (i) test that the effect of one treatment is larger or not much smaller than that of the other, (ii) test whether the treatments are equivalent, or (iii) select the better treatment when they may not be equivalent and otherwise assert equivalence. Statistical procedures for achieving these goals are proposed and description of the required sample sizes is given with the necessary tables.