In hiring, termination and promotion cases, the analysis examines each step of the decision making process to determine who should have been equally likely to be selected for the particular employment event. For example, in hiring at each stage you look at who remained eligible to progress to the next stage. Ultimately, the investigation should arrive at the final stage where selection rates of the remaining eligible individuals (the refined pools) are examined to see if the selection rates indicate any statistically significant disparity. If they do, discrimination can be inferred.
Compensation analysis is handled differently. Compensation analysis begins with snap shot data that is examined for disparities of a certain magnitude in pay that correlate with protected class status. Certain factors, such as education, skills and experience, believed to impact pay are then examined in a regression analysis and any unexplained residual disparity is attributed to discrimination. The regression analysis will also include items identified by the contractor as possible explanations for how pay is set.