ABSTRACT
Interaction between xi predictor variables is a common phenomenon in multiple regression practices. Technically, a regression model contains only
independent xi variables and is concerned with the predicted additive effects of each variable. For example, for the model, y^¼ b0 þ b1x1 þ b2x2 þ b3x3 þ b4x4, the predictor xi components that make up the SSR are additive if one can add the SSR values for the separate individual regression models
(y^¼ b0 þ b1x1; y^¼ b0 þ b2x2; y^¼ b0 þ b3x3; y^¼ b0 þ b4x4), and their sum equals the SSR of the full model. This condition rarely occurs in practice, so it
is important to add interaction terms to ‘‘check’’ for significant interaction
effects. Those interaction terms that are not significant can be removed.
