Practical Predictive Analytics

Interactions

We have seen that regression is additive, that is, the prediction is mostly calculated via adding crossproducts of the coefficients and independent single variables together. However, often single variables by themselves are not enough to adequately predict an outcome in a simple regression model. Often considering how say, predictor variable A affects predictor variable B can improve the outcome. This also goes a long way in terms of explaining the rationale behind using certain variables. The way two or more variables influence each other is referred to as interaction, and using an interaction in a model can produce an effect which is greater than the sum of their individual effects. Usually, two or more interaction variables ...