You can also try out heading in the opposite course: Identify the 'outliers' & see how they differ from the rest of your information. Checking out the residuals vs fitted plot, I see that a great deal of the 'outliers' have predicted values in the center. W/o figuring out add