Regression Effects When the Assumption of Rectilinearity is Not Tenable

Abstract

When analyzing data which deals with repeated testing, one may find that extreme scores are regressing away from the mean, contrary to what one would expect based on the regression effect. This paper discusses the regression effect and presents the argument that when these contrary results occur, they are indicative of the violation of the underlying assumption of rectilinearity. One should be required to look for non-linear relationships when interpreting such data. In addition, three methods for determining whether or not non-linear relationships exist in data are suggested and briefly discussed.