A screening design for factorial experiments with interactions

Abstract

This paper describes a new screening design which isolates the most important factors in experiments involving a large number of significant factors and interactions. The design is called the systematic fractional replicate design. For an n-factor experiment 2n + 2 units are required, and the results indicate whether each factor has odd and even order effects. No prior assumptions are made about the nature or absence of interactions. The design can be arranged in blocks of any size without altering the analysis. © 1979 Biometrika Trust.