
182 BUSINESS APPLICATIONS OF MULTIPLE REGRESSION
or logs. These transformations are particularly effective when the data
show high variability on one side of the residual plot and low variabil-
ity on the other. However, a nonlinear transformation introduces all the
theoretical interpretation concerns raised earlier.
A third approach is to simply realize that heteroscedasticity exists in
the multiple regression model but to otherwise ignore it in the calcula-
tion of the model. Given the minor impact of heteroscedasticity, this is
the approach often taken by researchers.
Summary
In this chapter we looked at model building, incorporating ...