Some general results on least favorable parameter configurations with special reference to equivalence testing and the range statistic (Q1176756)

Given a family of distributions with densities depending on a location parameter \(\delta = (\delta_ 1, \dots, \delta_ k) \in\mathbb{R}^ k\), \(k \geq 2\), the problem of testing \[ H_ 0: \delta \not\in \Omega (\delta_ 0): = \left\{ \delta \in \mathbb{R}^ k : \max_{i<j} | \delta_ i - \delta_ j | \leq \delta_ 0 \right\} \] against \(H_ 1: \delta \in \Omega(\delta_ 0)\) for some preassigned \(\delta_ 0 > 0\) is considered. For nonrandomized tests having invariance properties, general results for the extremal points of the power function over both the hypotheses are obtained. The power problem is completely solved for tests based on the range statistic. The equicorrelated normal case is studied in detail and sample sizes for controlling both error probabilities in equivalence testing are given. Appropriate tables are included.