EnPact, a company which performs environmental impact studies, is a medium-sized company. Currently, they are being audited by the Equal Opportunity Employment Agency for possible gender discrimination. Our firm has been brought in to conduct a preliminary analysis. A database of employee information is available in the attachment below. These data include employee salaries, genders, education, job level, experience, and age.

First, I want you to construct a full regression model for these data. Next, you should work toward the best possible model by dropping insignificant variables, one at a time according to the following rules:

1. Always drop the least significant variables first because this may change the significance of the remaining explanatory variables.
2. If you decide to drop a category of a categorical variable from the model, you must drop all the other categories of that categorical variable as well. This is an all-or-nothing proposition for categorical variables at this stage of our analysis.
3. Only drop a single numerical variable or a group of related dummy variables at each stage of the model-building process.
4. Any variables whose significance is questionable (that are close to the border, p = 0.05) should be kept, but noted for further investigation in your report.
5. Furthermore, you may detect outliers in the residual plots. At this stage of our analysis, do not delete them; further investigations may determine that these should be kept in the data. However, notes should be made in your report to identify any outliers.

Your final report on these data must discuss what your model tells you about the significant influences on the salaries at EnPact and should explain how gender might be implicated in the salary structure.

Attachment: Data file C09 EnPact.xls [.rda]