Generalized Linear Models as an Alternative Approach to the Taguchi's two-step procedure

The aim of this paper is to show the efficiency and the validity of the statistical models as an alternative tools to the Taguchi's technique, an innovative methodology of the off-line Quality Control. More precisely, we focalize our attention to Generalized Linear Models (GLMs) and to the Taguchi's two-step procedure, fully identified in the Nominal The Best situation. The performance measure of this experimental situation is replaced by means of two distinct but, at the same time, joined GLMs: one for mean and one for dispersion. The structure of the variance function of the dependent variable allows us to express in articulate manner the relationship between mean and dispersion, so common in Quality Control. We illustrate our approach through an application on data obtained by an experimental planning according to Taguchi's parameter design, showing the improving achieved through GLM's and the dual approach.