Exploring Statistical Design of Experiments

This chapter shows how to reduce the number of trials while extracting as much information as possible out of an experiment. This requires planning experiments statistically, to maximize the information derived and minimize the work required. Novice experimental plans often run “one factor at a time,” stepping across the range of each variable in turn while holding all others constant. The chapter overviews full-factorial designs and fractional-factorial designs. It aims to set up a well-designed 12-run Plackett–Burman (PB) screening design in order to identify the three or four most important factors from a list of up to 11 candidate factors. In order to lay a foundation for PB, the chapter introduces concepts of statistical experiment design via a two-level factorial design. It explains how to analyze full-factorial and PB results by free open-source software.