Lognormal Selection With Applications to Lifetime Data

Abstract

We consider selection problems on k lognormal (/spl mu//sub i/,/spl sigma//sub i//sup 2/) populations when the /spl sigma//sub i//sup 2/ are unequal, under both known & unknown cases. Selection based on linear functions of /spl mu/ & /spl sigma//sup 2/ are considered. For the unequal & unknown /spl sigma//sub i//sup 2/ case, a two-stage Rinott-type asymptotic procedure is proposed. In consideration to reliability & survival analysis applications, an asymptotic procedure is also proposed for the parametric selection based on the /spl alpha/-quantiles. Simulation studies to evaluate the procedures & to compare the latter procedure to the nonparametric procedure based on the /spl alpha/-quantiles are presented. The procedures are demonstrated using examples from reliability & quality control.