SECTION 11.3 Summary

• Start the model building process by performing data analysis for multiple linear regression. Examine the distribution of the variables and the form of relationships among them.
• Note any categorical explanatory variables that have very few cases for some values. If it is reasonable, combine these values with other values. If not, delete these cases and examine them separately.
• For curved relationships, consider transformations or additional explanatory variables that will account for the curvature. Sometimes adding a quadratic term improves the fit.
• Examine the possibility that the effect of one explanatory variable depends upon the value of another explanatory variable. Interactions can be used to model this situation.
• Examine software outputs from modern variable selection methods to see what models of each subset size have the highest .