What Difference Does it Make? Implications of the Size of the Difference between the Means of Two Groups

Abstract

This study employed a model with two normal distributions of scores to study group differences and similarities as a function of the distance between group means. The mean of one distribution was systematically moved further and further away from the mean of the other. A table was developed of associated measures of group overlap and separation, e.g., r, eta, percent correct classification, percentage of overlap, for several levels of d' (difference between means expressed in standard deviation units). The implications of various d' values for the selection of extreme groups, e.g., top 1% of scorers, were discussed. It was noted that even small differences between means could result in an extremely disproportionate inclusion of members from the two original groups in a high-scoring subgroup.