Higher-order terms refer to polynomials constructed from a numeric variable by raising that variable to one or more powers. For example, the higher-order terms of a variable called age would be (age)2, (age)3, etc. Because polynomial powers can form a large number of smooth shapes, including higher-order terms in a model can provide a better model fit and account for some non-linearity in the relationship between an explanatory variable and the outcome variables. However, including higher-order terms without a prior justification may add unnecessary complexity to the model and ultimately reduce its predictive power by increasing the number of model coefficients relative to the number of observations.
To include higher-order terms in a model:
Up to four higher-order terms (up to a fifth power) may be included. Predictions will automatically take the included higher-order terms into account.