A Graphical Method for Selecting the Best Sub-Set Regression Model

Abstract

The purpose of the present paper is to provide an empirical example of a graphical procedure that may be used for parameter selection in multiple linear regression. An all possible sub-sets approach is utilized in order to accomplish this objective. Elements related to the model construction process, such as parsimony, and explanation value are explored in the context of this approach. A graphical presentation of the findings is discussed as a way of facilitating the understanding of how R^2, adjusted R^2 and the number of parameters in a model are related to one another.