Abstract and subjects
Screening designs for identifying active main effects among a group of factors are frequently used by experimenters. Methods for efficiently designing experiments that estimate main effects are needed, particularly if the experiment involves categorical factors at more than two levels and requires a nonstandard run size. We propose a new and powerful algorithm based on a sequential element-wise-column-wise strategy that focuses on maintaining balance within and between columns. We compare the new algorithm with a column-wise algorithm for finding some selected orthogonal and near-orthogonal arrays. We examine the performance of both algorithms using different criteria and provide a discussion of general strategies in searching for designs.