On some aspects of variable importance in regression models

Variable importance evaluation functions can be separated into two groups: those that use the model information and those that do not. The advantage of using a model based approach is that is more closely tied to the model performance and that it may be able to incorporate the correlation structure between the predictors into the importance calculation. Regardless of how the importance is calculated: Of the variables chosen as predictors of salary, only degree had a negative effect on the dependent variable; that is, as level of education increased, salary tended to decrease for career services directors at both institutional types. This negative effect signified a high probability of the presence of salary compression. Further, correlation matrices verified that degree was acting as a suppressor variable within the regression models by denoting that it had a very low, close to zero correlation with salary and a correlation with another independent variable (e.g., age). The negative suppressor designation of degree was applicable by observing its positive correlation with salary and its negative beta weights within the regression equations.