Classical Design and Analysis of Experiments

Abstract

Experiments are used in industry to improve productivity, reduce variability, enhance quality and obtain robust products and manufacturing processes. In this chapter we study how to design and analyze experiments which are aimed at testing scientific or technological hypotheses. These hypotheses are concerned with the effects of procedures or treatments on quality and productivity; or the general relationship between variables. Designed experiments help determine the conditions under which a production process yields maximum output or other optimum results, etc. The chapter presents the classical methods of design of experiments. It starts with an introductory section with examples and discusses guiding principles in designing experiments. The chapter covers the range of classical experimental designs including complete block designs, Latin squares, full and fractional factorial designs with factors at two and three levels. The basic approach to the analysis is through modeling the response variable and computing ANOVA tables. Particular attention is given to the generation of designs using Python.