Testing interval hypotheses for scale parameters in gamma distributions

Abstract

In this paper we derive the likelihood ratio test (LRT) for testing interval hypotheses for several gamma scales with known shapes, and compare its performance with those of ones based on suitable combinations of independent p-values. It turns out that Tippett’s method of combination of p-values performs the best among such combinations, and its performance matches that of the LRT. A real application illustrates the proposed methods.