A Configural Approach to Predicting the Shape of Trajectories

Abstract

Many developmental pathways and trajectories are far from universal. In this article, configural methods are proposed that allow one to test hypotheses about differential developmental trajectories. Trends in categorical variables are discussed. A simulation study is presented in which it is shown that (1) all conceivable patterns of trajectories can be characterized configurally, and (2) even when the null hypothesis of independence of repeated observations holds, descriptors of growth trends can be correlated. Implications of these results for the specification of base models are discussed. Models with single and multiple categorical predictors are discussed, as are models that include multiple descriptors for the shape of trajectories. A data example is given in which trajectories of depression are predicted for women who did versus did not experience intimate partner violence. The data are analyzed from both person- and variable-oriented perspectives.