Linear Regression Analysis, testing for serial correlation (Durbin Watson Test) based on Linear regression model for a linear line established by X (Independent Variable) & Y (Dependent Variable), checking for Description, positive vs. negative serial correlation,If Residual (Deviations) are Serial Correlated (+/-), 1-Estimated Regression Coefficients Not Minimized, 2-MSE (Mean Square) may Underestimate Variance Error, 3-Computed Standard Error (Parameter) Underestimated, detailed example by Allen Mursau