ABSTRACT
Factorial designs with each factor at two levels, referred to as low and high, are introduced with examples that show the importance of including interaction terms. There are two limitations of such experiments: the number of runs required if there are many factors, and the implicit assumption that the effect of factors defined over a continuous scale are linear. The first limitation is answered by introducing fractional factorial designs. Central composite designs are an answer to the second limitation. Case studies that use central composite designs in evolutionary operation (EVOP) for optimising the yield of a chemical process, and for designing an expert system for the operation of a cement kiln are described. All the analyses are performed using multiple regression. The experiments “Factorial experiments - reaction times by distraction, dexterity, and distinctness” and “Where is the summit” relate to a factorial experiment with three binary factors and EVOP respectively. The chapter ends with: a summary of: notation used; the main results, MATLAB and R syntax; and exercises.
