Suppressor Variables Revisited

A scheme for classifying suppressor variables is proposed, within which mathematical models are described and statistical conditions derived for distinguishing among several kinds of suppressors. In addition, some general criticisms are made with respect to the inadequacies of the previous use of the suppression technique. In a supplementary note in the well-known 1941 SSRC Bulletin on the prediction of personal adjustment, Horst (1941) pointed out that if a predictor having zero validity but high correlation with a valid predictor was included in the regression equation along with the latter, the multiple correlation could be significantly increased. He proposed to call such an invalid predictor a 'suppressor' variable.