FOCUS ARTICLE: Separating Stable From Variable Individuals in Longitudinal Studies by Mixture Distribution Models

Publisher Website
Google Scholar
Abstract
After a short overview of the metatrait approach in personality psychology and of several related concepts (intraindividual variability, individual reactivity, hardiness, resilience), the consequences of these approaches for the measurement of individuals over time are discussed. Based on the implications of the metatrait approach for psychometric modeling, several extensions of a latent state-trait model for categorical response variables (Eid & Langeheine, 1999) are described. These extensions are mixture distributions variants of the basic model suitable for separating stable (traited) from variable (untraited, less traited) individuals in longitudinal studies. An empirical application of the models to the measurement of momentary negative mood on 4 occasions of measurement (N = 494) revealed that a model with 2 latent chains fits the data well. The first chain (size: 24%) consists of individuals that are perfectly stable (traited) on the latent level, whereas for the second chain (size: 76 %), a latent state-trait model with Markov structure (less traited individuals) explains variability and stability appropriately. The consequences of the results of the study for the measurement of individuals in contexts as well as the psychometric modeling of change processes are discussed.