The change in the degree of multicollinearity provides a clue to the reasons for the apparent outlier. A scatterplot of x2 and x3 (not reproduce here) shows that these two variables are strongly related except for observed combinations of x2 and x3 values. The conclusion with reached to the observed combinations probably produces a production rate that does not conform to the model that is valid for the rage of data represented by the remainder of the observed sample,or it could be an erroneously value