OFF-QUADRANT COMMENT

Points out a serious methodological problem of quadrant analysis (QA), a method for identifying correlates of individual differences in predictability. QA involves division of the predictor-criterion scatterplot into 4 cells by cutting at the predictor and criterion medians. For cases below the predictor median, and separately, for cases above the predictor median, a comparison is made between the test responses and scores of those above the criterion median with those below. The distinguishing responses and scores are combined into moderator variables, 1 for each predictor group. To the extent of the relationship between the predictor and criterion, cases above and below the criterion median within each predictor category are not equivalent on predictor scores. Because of these differences, items or scores found to differentiate between comparison groups may merely reflect the predictor composite differences. Moderators developed through QA may therefore be predictive only of the composite predictor itself. (15 ref.) (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1969 American Psychological Association.