A Class of Three-Level Designs for Definitive Screening in the Presence of Second-Order Effects
A Class of Three-Level Designs for Definitive Screening in the Presence of Second-Order Effects
- Publication:
- Journal of Quality Technology
- Date:
- January 2011
- Issue:
- Volume 43 Issue 1
- Pages:
- pp. 1-15
- Author(s):
- Jones, Bradley , Nachtsheim, Christopher J.
- Organization(s):
- SAS Institute, Cary, NC, University of Minnesota, Minneapolis, MN
Abstract
[This abstract is based on the authors' abstract.]Screening designs help assess the relative impact of a large number of factors. Experimenters often prefer quantitative factors to have three levels rather than two, but common screening designs use only two factors. This article proposes a new class of three-level designs that is unbiased by second order effects, require only one more than twice as many runs as factors and avoid confounding any pair of second order effects. The new class of designs also allows for efficient estimation of the full quadratic model for any three factors if six or more factors exist, potentially rendering followup experiments unnecessary. The article provides an algorithm for use in design construction.