This paper analyzes linear models. It investigates the difference between the sum of squares of the residuals and the sum of squares of the prediction errors when the parameter is estimated consecutively. In case the regressors are “fractionally integrated” (in a very broad sense) it is...