ON THE PROPERTIES OF SUBSET SELECTION PROCEDURES.
PURDUE UNIV LAFAYETTE IND DEPT OF STATISTICS
Pagination or Media Count:
Some desirable properties are studied of a selection procedure which selects the normal population with mean m and variance unity i1,2,...,k if the observed sample mean x sub i from p contained in xMax- d, xmax. This rule earlier studied by Gupta 1956, 1965 is compared with the approximate optimal rule D of Seal 1955. It is shown that the rule R is minimax. It is also shown that under the slippage configuration of means given by m,m,...,m delta the expected size of the selected subset using R is smaller than that corresponding to D and that the probability of a correct selection using R is strictly greater than that of D, provided delta satisfies some inequalities. Under a more general linear loss function, the Bayes rule for selecting a subset is also derived. Author