Chapter 9 Multiple Regression 375You might wonder why the variable Age Hired is used instead of em-
ployee age. The problem with using employee age is one of collinearity.
Collinearity means that one or more of the predictor variables are highly
correlated with each other. In this case, the age of the employee is highly
correlated with the number of years employed because there is some over-
lap between the two. (People who have been employed more years are likely
to be older.) This means that the information those two variables provide
is somewhat redundant. However, you can tell from Figure 9-13 that the
relationship between years employed and age when hired is negligible be-
cause the p value is .681 (cell E13). Using the variable age hired instead of
age gives the advantage of having two nearly uncorrelated predictors in the
model. When predictors are only weakly correlated, it is much easier to in-
terpret the results of a multiple regression.
The correlations for Salary show a strong relationship to the number of
years employed and some relationship to age when hired, but there is little
relationship to a person’s degree. This is in agreement with the SPLOM in
Figure 9-12.Figure 9-13
Correlation
matrix for
male faculty
salary data
4 Click the Output button, click the New Sheet option button, and
type Male Corr Matrix in the New Sheet text box; then click OK
twice. The resulting correlation matrix appears on its own sheet, as
shown in Figure 9-13.