Discussion of Boh's 1987 Article in Technometrics

摘要

Professor Box is to be congratulated for his timely and informative article. Along with Box and Fung (1983, 1986), Box and Meyer (1986a,b; 1987) and Leon Shoemaker, and Kacker (1987), this article provides the statistics community with a clearer sense of the meaning of the term 'Taguchi's method' and how to improve it. Although we do not work in the area of quality control, we do have experience in areas that overlap in terms of statistical analysis. The overlap occurs after a screening analysis has been undertaken to identify those factors that affect the response. Some of our experience is related in Carroll and Ruppert (1987, 1988) and Davidian and Carroll (1987). One interesting component of Box's article is the idea that a transformation of the data can be found that induces not homoscedasticity, but a partial separation of the factors affecting mean and variance. (mgm)