Experiments with Mixtures; Designs, Models, and the Analysis of Mixture Data. 2nd Edition

John A Cornell (1990) Experiments with Mixtures; Designs, Models, and the Analysis of Mixture Data. 2nd Edition John Wiley & Sons, Inc.

A mixture experiment might be run to see the affect of varying the percentages of two ingredients blended together. If one is to be set at 35% then the other must be at 65%. As it is raised to 45% the other percentage will become 55%. The individual percentages cannot be varied independently and so cannot be considered individual factors. The usual fractional factorial experiment cannot be run. Cornell's comprehensive book describes experimenting in these situations and must be considered the definitive work. The initial chapter is a useful overview of mixture experiments. The fundamentals of setting up and analyzing (including reading plots) mixture experiments are presented. The material has examples with data that mostly from the chemical industry. As well as being an introduction (although one that includes matrix algebra), the material covers particular cases and is apparently very thorough. review by Cooper