ABSTRACT
This chapter explores the modeling of numeric variables by factors, variables that take on one of a specified set of levels. Such data are often the result of a designed experiment, in which observations of one or more responses are made for changing values of several factors. In a designed experiment, some of those variables will, in principle, have values from a finite set of levels specified by the experimenter in advance. In classical experimental design, these variables are categorical, and are usually called the factors in the design. The objective of the experiment was to investigate various factors that might affect the quality of the fabrication process, with the goal of controlling both the average level and the variability of the quality. Some factors in experiments have the property that the levels chosen are not so much interesting in themselves but instead are examples of "typical" values for the underlying variable.
