What is Regression Analysis?
Multiple regression is a technique used to help evaluate the relationship between measurable data factors. Specifically, it assists in determining the extent that one or more factors (tenure, performance appraisals, education, etc.) impact a quantity of interest, like compensation. A factor in a regression is said to be significant if it helps explain variations in the quantity of interest to a large degree. If demographic characteristics of employees appear to be significant factors in a regression analysis of compensation, this may be evidence that discrimination is present in employer compensation decisions.
PayEval makes conducting a regression analysis of employer decisions simple.
The detailed report will include a statistical test of the factors used in the regression model. For this analysis, a table is included with the number of observations used in the analysis, the degrees of freedom for error and the percent of the variation explained by the model. For the overall analysis of gender impacts on compensation, the selected model with the included factors explains 86.5% of the variation in compensation. In the subsequent table, note that service, EEO job code and job title have a significant impact on compensation at a 95% confidence level, as indicated by the highlighted standard deviation values of 5.20, 3.56, and 7.30.
Another option is to display only significant regression results. This would only include the items from the standard analysis that produced significant results. This also provides information such as reference groups and controls used in the regression model. Note that all the resulting estimated impact for this specified report would be highlighted in red or bolded since only results considered significant based on typical OFCCP guidelines are included. For this regression analysis, the “Black” employees appears to have a pay rate that is 71.35% less than the reference race within the “white” job group. However, note that none of these are in qualified comparison groups since there are only 20 employees in the group (based on OFCCP’s 30/5 rule).
