An Approach to Developing Multi-Level Fractional Factorial Designs

Abstract

Two-level factorial designed experiments are popular in industry due to their simplicity, efficiency, ease of interpretation, and flexibility in sequential experimentation. However, experimenters are often frustrated by a few factors that require more than two test conditions among many that do not, forcing the use of unnecessarily complex designs, or compromising the experimental objectives. This paper discusses some straightforward extensions to the two-level factorial approach which allow a “mostly” two-level design with some three or four level exceptions with no loss of orthogonality or design resolution.