Permutation tests - a revival?! I: Optimum properties (Q1381185): Difference between revisions

The authors of this paper argue that in the present era of computer intensive methods permutation tests should not be discarded any more on grounds of the computational effort that is needed for constructing their critical region. In support of their argument they show that permutation tests have convincing optimum properties for continuous distributions as well as for discrete ones. Their considerations are carried out in terms of the two-sample problem of comparing two treatments which are applied to \(n_1\) resp. \(n_2\) homogeneous items. It is assumed that the observations are realizations of independent random variables \(X_{1,1}, \dots, X_{1,n_1}\) and \(X_{2,1}, \dots, X_{2,n_2}\), respective1y, whose distributions are stochastically comparable. General conditions are given that are sufficient for uniformly maximal power on certain subclasses that are also illustrated by a number of examples.