Dispersion Effects from Fractional Factorials in Taguchi's Method of Quality Design
Journal of the Royal Statistical Society: Series B (Statistical Methodology) · 1994 · Volume 56 · Issue 4
DOI: 10.1111/j.2517-6161.1994.tb02005.x
Abstract
Taguchi proposed direct product designs for studying the influence of control factors on the dispersion introduced by error factors. The current paper presents several general results concerning the use of orthogonal array designs to estimate the effects of control factors on the dispersive effects of error factors. Sufficient conditions are given such that dispersion estimates are ‘undistorted’, in that they estimate the same quantity in fractional factorials as they would in complete factorials. The conditions concern the strength of the orthogonal array and the complexity of the linear model. Taguchi's designs are ‘reliable’ in that they prevent interactions among control factors from being mistaken for dispersion effects involving error effects. There are other reliable orthogonal array designs that are, in some ways, preferable to Taguchi's direct product designs in that they have greater strength or resolution with the same number of observations.
Keywords
experimental design, fractional factorial design, orthogonal array, quality improvement
Enable JavaScript for citation tools and additional paper interactions.