Probability Optimal Supersaturated Designs

Abstract

Abstract:

In this paper we have argued that the probability of correct searching is a more acceptable criterion for comparing two competitive supersaturated designs. We have derived an expression for the maximum value of the probability of correct searching for a supersaturated design for a given run size, assuming that there is only one active factor present in the model. We have also provided a result pertaining to the maximum number of factors that can be accommodated with a given value of probability of correct searching and with a particular run size. Finally an algorithm for constructing probability optimal supersaturated designs is presented. The performance of the designs constructed are also evaluated, in terms of the probability of correct search in stituations involving the presence of two active factors, using a simulation study.

AMS (2000) Subject Classification: 62K05, 62K15.