A statistical procedure that estimated regression equation coefficients that produce the lowest sum of squared differences between the actual and predicted values of the dependent variable is called .