Representative design and the general linear model

This paper argues that traditional experimental methodology contributes to violating “representative validity,” defined herein as the degree to which “actual” generalizable behavior is produced in an experimental context. The importance of representative validity is discussed, and the use of the general linear model is advocated as a means of maximizing representativeness in experimental design. Six issues are discussed in connection with representative design, and the general linear model is presented as a solution to various problems which adversely affect representative validity. These issues are statistical power analysis, the use of continuous variables, unbalanced designs, non‐linearity, interactions, and random and mixed‐effects models. It is suggested that the common use of linear model techniques provides the potential for increased flexibility and encourages creative design of experimental research.