Detecting Predictive Bias: The WISC-R vs. Achievement Scores of Mexican-American and Non-Minority Students

Abstract

This study takes issue with the conventional application of regression analysis to the problem of detecting predictive bias of IQ tests toward minorities. Jensen (1968, 1980) and Reynolds (1982) rejected the hypothesis that IQ tests are unfair. However, the statistical techniques employed by these and other researchers in the analysis of this problem are of questionable sensitivity to predictive bias. In the present study, the predictive relationship between the WISC-R and PIAT and WRAT achievement scores of 125 Mexican-American and 125 non-minority students was analyzed using four separate methods of analysis. Conventional methods were found to be insensitive to and inappropriate for the detection of predictive bias. An alternative means of detecting predictive bias is proposed.