of the variation in the data, and we would disagree that the straight line is “obviously bad” or “not a good description of the data.” We do not advocate elimination of the term “test for the significance of the regression,” but we would like to see the term “goodness of fit” associated with F = MSifMS and dissociated from F = MS101/MSpe. The larger the value of F = MSiact / MS, the greater is ...
