Point-of-expansion, structure, and selection in multivariable polynomial regression

Abstract

This article discusses three issues important in multiple linear regression analysis when a polynomial response function is used. These are (1) the role of the point of expansion, about which the response function is formulated, (2) the effect of polynomial structure among the predictor variables, and (3) the appropriate use of methods for selecting predictors. Two examples are included for illustration.

Keywords:

• centering
• poly-subsets regression
• subpoly-subsets regression