The analysis of change: issues, fallacies, and new ideas

Abstract

In this special section of The Journal of Consulting and Clinical Psychology, new ideas about how to analyze change are presented in a format that is accessible to clinicians and clinical researchers. Rogosa's (1988) myths of longitudinal research are reviewed in an attempt to familiarize psychologists with the dangers of assuming (a) that regression toward the mean is unavoidable, (b) that difference scores are unreliable, (c) that analysis of covariance is the way to analyze change, (d) that two points are adequate to measure change, and (e) that the correlation between change and initial level is always negative. An overview of the articles emphasizes what is new and improved in the design and analysis of change. The articles are preceded with a conceptual discussion of how to measure change over time when the stability of the criterion construct is high and there is little variance to predict. Other articles discuss the form of change over time and how this can be an important tool in testing specific hypotheses. Individual change over time can be described with short time-series analysis or sequential analysis of continuous data. Individual and group change over time can be described in survival analyses or cohort-sequential designs. Some articles minimize the problems of cohort-sequential designs by including cohorts of overlapping ages and comparing hierarchical models of change. Discrete-time survival analyses have intuitive appeal, can include several types of predictors in the models, and are relatively simple to compute.