18 Pages

Module 17. Combining Predictors Using Multiple Regression

In Module 8, we said that evidence of criterion-related validity is demonstrated by correlating test scores with corresponding criterion scores. If the test is sufficiently related to the criterion of interest, regression can later be used to predict criterion scores in a sample of people for whom we have no actual criterion values, as long as the sample is drawn from the same population used in the original validation study. Obviously, the stronger the relationship between our test and the criterion, the more accurate will be our predicted criterion score. If a test is used in this way to predict a criterion score, the test is often referred to as a predictor . In most cases, we could increase the accuracy of predicting our criterion if we expanded the number of predictors beyond one.