An Application of Model-Based Reasoning in Experiment Design

Scientists and engineers in diverse fields such as manufacturing, medicine, and design use experiments to learn about processes and the behavior of systems. Experiments study how the settings of a series of factors affect one or more response variables. For example, an engineer trying to develop a reliable painting process for automobile components might set up an experiment to study how paint viscosity and temperature (two factors) affect a numeric measure of paint surface quality (a response variable). Because an experiment can require significant resources, the experimenter often must make trade-offs between the number of experimental trials, the order of these trials, and the expected amount and type of information gained as a result of running the experiment. Design of experiments is the field of statistics that addresses the problem of creating layouts (ordered lists of factor combinations , or trials, to be tested) that will provide the experimenter with statistically sound results yet account for the constraints under which the experiment must be run. The use of experimentation in industry has grown as companies at