Structural models for studying correlates and predictors of change

This paper is concerned with the theoretically and empirically important issue of identifying correlates and predictors of change in the context of a repeated measures design. Two traditional approaches to change measurement — the residualised observed difference score and the base-free measurement of change (Cronbach Tuckert, Damarin, & Messick, 1966) — are initially discussed. Within the structural equation modelling methodology, the multiple-indicator extension of the original base-free measurement of change approach is then focused on. A structural model is next described that allows estimating the extent of covariation between true change and other measures. The two models permit consistent and efficient estimation of the degree of interrelationship between true change in longitudinally assessed psychological constructs and other variables, such as studied/presumed correlates and predictors of that change. These structural models are useful when it is of interest to identify variables that are correlated with, and can be used to predict, growth or decline in repeatedly assessed latent dimensions. The described modelling approach is illustrated using data from a cognitive intervention study of aged adults (Baltes, Dittmann-Kohli, & Kliegl, 1986).