Regression toward the mean associated with measurement error and the identification of improvement and deterioration in psychotherapy.

Abstract

Several factors can cause posttherapy scores to regress (relative to pretherapy scores) toward a group mean. This article focuses on regression effects that are due to measurement error, and on methods used to evaluate pre- to posttherapy score changes that attempt to take these regression effects into account. Statistical and practical problems associated with these methods are discussed. In particular, a statistical problem is identified in the most recent article that advocates use of these methods. Consequences of ignoring this problem and a way of correcting it are presented. Five practical problems encountered in using a correct classical true-score theory method of identifying reliable changes (based on derivations of H. Gulliksen [1950] and F. M. Lord and M. R. Novick [1968]) are then presented. Finally, objectives of these methods are distinguished from objectives of other strategies used to evaluate changes in psychotherapy.