Additivity Tests

Abstract Additive linear models are ubiquitous in applied areas of behavioral statistics, and so it is often important to test whether data fit the assumption of additivity. In a test of additivity, the null hypothesis is that a dependent variable is an additive, noninteractive function of two (or more) independent variables, while the alternative hypothesis of nonadditivity is characterized by one or more interactions between the independent variables. It is possible to test for additivity in the context of Analysis of Variance (ANOVA), while the theory of Additive Conjoint Measurement provides a more general basis for testing additivity.