An Analysis of Taguchi's Method of Confirmatory Trials

Abstract

Taguchi has suggested a new method of confirmatory trials that is intended to test for possible interaction effects. The method is being promoted by many recent authors. A careful analysis of this method, however, shows that confirmatory trials are ineffective in detecting the presence of interaction effects. In fact in many cases, it is shown that the probability of detecting an interaction effect decreases rather than increases as the size of the interaction effect increases. Thus, we advise practitioners not to adopt this new method as a test for interactions.