Looking to Benchmark Pay?
SUMMARY
Regression is a powerful tool and can be very useful in the analysis of compensation. But before you perform regression analysis, you must first understand distributions, standard error, and standard deviation. You must also be able to use correlation, since a strong correlation between the dependent variable and the independent variable will give you a good regression model. On the other hand, strong correlation between the independent variables will lead to collinearity and a poor regression model.
Compensation Models
Regression analyses lets compensation analysts consider multiple factors when setting pay levels. Linear regression allows you to make predictions on any dependent variable, based on any independent variable. Multiple regression lets you make predictions of any dependent variable, based on several independent variables.
Judging Regression Equations
When evaluating regression outputs, the R2 and F statistic will give you an idea as to the goodness of the regression model. The t statistic and standard error will give you an idea as to the fitness of the variables used in the regression model. Finally, outliers can give you poor regression models. Therefore, you should examine the spread of the data using histograms before formulating a regression model.