The use of stagewise regression was investigated in a university study of salary equity for full-time staff-exempt personnel. The commonly used method of regression substitution was modified in three ways: (1) shifting the emphasis away from estimating inequity and toward describing the impact of men's and women's population differences on salaries; (2) using stagewise regression instead of multiple regression; and (3) adding a graphical procedure (plots) to the variable selection process. The basic idea in stagewise regression is that at each step the effects of the most important remaining variables are estimated and removed. Study variables included: age when hired, years of service, years in rank, educational level attained, staff grade, and method of job entry (e.g., new hire, promotion). The stagewise regression method allowed for nearly parallel models for men and women to be developed. It also provided a different perspective from least squares by attributing more of the variation in the response variable to those variables that are judged to be more important. In addition, the agreement, or lack of it, between stagewise and least squares results can be an indicator of how balanced or unbalanced the data are. (SW)