A Comparison of Dispersion Effect Identification Methods for Unreplicated Two-Level Factorials

Abstract

Identifying factors that affect the variability of a process has become an important step in improving quality. Over the years many methods based on unreplicated two-level factorial experiments have been proposed for identifying such factors. In this article we present a comparison of the statistical power associated with several alternative methods. To this end we also present a method of operationalizing the Half-normal probability plot which can be used to assess the relative "power" associated with it.