ABSTRACT
Interpretation of the ordinary regression model—the most popular technique for making predictions of a single continuous variable Y—focuses on the model's coefficients with the aid of three concepts: the statistical p-value, variables held constant, and the standardized regression coefficient. This chapter demonstrates that the statistical p-value as the sole measure for declaring predictor variable X important is sometimes problematic. The concept of variables held constant is critical for reliable assessment of how predictor variable X affects the prediction of Y. And, the standardized regression coefficient provides the correct ranking of variables in order of predictive importance—only under special circumstances. The ordinary regression model, formally known as the ordinary least squares multiple linear regression model, is the most popular technique for making predictions of a single continuous variable. The regression model illustrates the difficulty in relying on the regression coefficient for ranking predictor variables.
