ISBIS4 Abstract

Contact Author's Name: Randy Sitter
Title of Abstract: Fast Algorithms for Finding Experiment Designs in Large-Scale
Author(s): M. Hamada, R. Sitter, D. Bingham, L. Moore and J. Wendelberger
Affiliation: Los Alamos National Laboratories and Simon Fraser University

Orthogonal arrays (OA) are often used in industrial and scientific settings where investigators wish to run experiments with factors that have different levels and there is a limit on the number of runs relative to the number of factors. They are also used in computer experiments as the basis for constructing Latin hypercubes with good space-filling and projection properties. In either context, the combination of run-size restrictions and level combinations may create a situation where no OA exists. In these cases one would like to obtain a nearly OA. Though there can be many definitions of what ``nearly\'\' might mean, the general idea is fairly clear. We will consider algorithms for obtaining NOA quickly.