As expected, the more outliers that are removed from the equation, the more it lends stability to the end equation. This reduced the number of outliers from three to one unusual observation noted. The equation changed slightly as well as other computations. The R-square increased from 11.6% up to 20.1%! With this new development, we decided to take a one step further and conduct one more regression analysis to see what would happen if that one outlier was omitted. We predicted that the R-square and F-value would continue to increase whereas the P-value would become smaller and smaller. We also predicted that Minitab would no longer find anymore unusual observations. The results are shown in the following table:
Regression Analysis: Spending per pupil ($) versus Composite Score
(chart)