Multiple classification rules for signs of parameters (Q1084791)

The Spjøtvoll-Bohrer (SB) procedure [\textit{R. Bohrer}, J. Am. Stat. Assoc. 74, 432-437 (1979; Zbl 0418.62041) and \textit{E. Spøtvoll}, Ann. Math. Stat. 43, 398-411 (1972; Zbl 0241.62045)] for classifying signs of \(k\geq 2\) unknown scalar parameters based on normally distributed estimators with known covariance matrix is considered. Under a particular criterion it is shown that the SB rule is uniformaly dominated by some members of a class of procedures that control the probability of any misclassification at level \(\alpha\). On the other hand it is shown that a generalized SB rule is the only admissible rule in another class of simpler procedures.