Appropriate and Inappropriate Use of Estimated True Scores for Normative Comparisons

Abstract

Some commentators have argued that test users should employ estimated true scores in place of the usual observed scores for many of the interpretations of standardized test results. We agree with this position for some of the score interpretations described here. However, for normative interpretations of individual scores and class averages, the use of estimated true scores or average estimated true scores is not recommended. No change in status would result from the use of estimated true values, since regressed scores preserve the rank ordering of groups and individuals when appropriate norms are used. Since test norms are always stated in terms of observed scores, such norms would be inappropriate and misleading if used to interpret estimated true values of local pupils or groups.