Optimum properties of the Wilcoxon signed rank test under a Lehmann alternative (Q1083148)

Let \(X_ 1,...,X_ n\) be a random sample from an absolutely continuous distribution function \(F(x)\). The problem is to test the null hypothesis H: \(F(x)=G(x)\) where \(G'(x)\) is assumed to be symmetric about zero. We consider the alternative of the contaminated distribution \[ K: F(x)=(1- \theta)G(x) + \theta \{G(x)\}^ 2\text{ for } 0<\theta <1, \] and the sequence of the local alternatives \[ K_ n: F(x)=(1-\Delta /\sqrt{n})G(x)+(\Delta /\sqrt{n})\{G(x)\}^ 2\text{ for } \Delta >0. \] Then we show that the Wilcoxon signed rank test is locally most powerful among all rank tests under K and is asymptotically most powerful under \(K_ n\). Further we compare the Wilcoxon signed rank test with the one- sample t-test by the asymptotic relative efficiency under the local alternatives of general contaminated distributions \[ K_ n': F(x)=(1- \Delta /\sqrt{n})G(x)+(\Delta /\sqrt{n})H(G(x))\quad for\quad \Delta >0. \]