Tests of a normal mean vector with hypotheses determined by linear inequalities

Abstract

For hypotheses about linear inequalities, we present algorithms to construct a more powerful test from a test with symmetric critical region, e.g., LRT and the tests of Liu and Berger's (1995). The test obtained from our algorithm is permutational invariant and admissible among tests with critical region inside the cone in the sample space

Keywords:

• Conditional Distribution
• Likelihood Ratio Test
• Polyhedral Cone