TWO PITFALLS IN ASSESSING FAIRNESS OF SELECTION TESTS USING THE REGRESSION MODEL

This note examines two potential pitfalls in applying the Cleary or regression model of test fairness. The first lies in a misinterpretation of significance tests on intercept differences which can result when the researcher is unaware of the properties of analysis of covariance tests for intercept differences and relies on computer printouts of regression equations. The second lies in the dependence of some tests for intercept differences on predictor scaling. Once aware of them, the researcher can avoid both these pitfalls.