UMP unbiased tests for multiparameter testing problems with restricted alternatives

Abstract

Applications of the standard theory of UMP unbiased tests depends on conditions which in general are difficult to verify. In the present paper, however, we suggest more simple rules for applying this theory for regular exponential families of distributions. This approach leads to UMP unbiased tests for various multiparameter testing problems with restricted alternatives, and is shown to give justification for conditional tests of various test problems for contingency tables. The derived tests are shown to possess attractive small sample properties. Keywords: Exponential families of distributions, UMP unbiased tests, exact conditional tests, contingency