Understanding Multiple Linear Regression Models: Quantifying Predictor Importance and Residual Variance in Predictive Accuracy