Abstract
Tests for bias under the regression definition were carried out for seven major tests used to predict first-grade performance on the Metropolitan Achievement Tests, over a 12-month interval, using Potthoff's (1966) method. Analyses were con ducted to evaluate bias in prediction for 322 children across sex and race (black-white). In the 112 tests for bias, only 13 significant differences in prediction occurred; in each case intercept bias was present that resulted in the overprediction of minority performance on the criterion measures. Correlation coefficients between pre- and post-tests were also compared across groups. For the 112 comparisons of validity coefficients, only two showed any race or sex differences, a chance finding at p < .05.
Get full access to this article
View all access options for this article.
