ABSTRACT
In this chapter, the authors begin by briefly considering these similarities and differences with respect to the model, the new problem of redundancy among the predictors, and statistical inference. They consider an extended example designed to illustrate the interpretation of the parameters in the multiple regression model. Multiple regression is likely the most widely used statistical procedure in the social sciences. As a consequence, it is also the most frequently abused procedure. Many regression programs have procedures for automatically building models. The most commonly used such procedure is stepwise regression, which, as its name implies, builds a model step by step. Stepwise regression and its variants present the alluring promise of finding the "best" model with virtually no effort by the researcher or data analyst. The discussion of statistical power in the context of simple regression, using a single predictor variable, generalizes quite readily to the multiple regression case, with multiple predictors.
