Industrial experiments often result in the production of various data, both continuous data as well as discrete data. Most of the time, practitioners use a normal analysis for continuous data and a logistic regression analysis for binary data. Taguchi (System of Experirnental Design, Kraus International, New York, 1986) suggested that one use various signal-to-noise analyses when encountering different types of data. Engineers do not usually encounter Poisson data; however, in this paper a set of Poisson data is presented with three approaches to analyse the data set. Their comparison is also presented. Copyright (C) 1999 John Wiley & Sons, Ltd.
