AN ANCILLARY PARADOX IN TESTING
Abstract: In multiple linear regression with normally distributed errors, it is shown that a
test procedure for a hypothesis about the intercept which is -admissible when
the design matrix is fixed is inadmissible when the design matrix is an ancillary
statistic. The result of this paper is a complementary one to Brown’s paper [2].
2000 AMS Mathematics Subject Classification: Primary: -; Secondary: -;
Key words and phrases: -