Advancing the dynamic characterization of psychological development through latent change score models

• Autores: Pablo Fernández Cáncer
• Directores de la Tesis: Eduardo Estrada Alonso (dir. tes.)
• Lectura: En la Universidad Autónoma de Madrid ( España ) en 2023
• Idioma: inglés
• Número de páginas: 122
• Títulos paralelos:
  • Avances en la caracterización dinámica del desarrollo psicológico mediante modelos de puntuación de cambio latente
• Enlaces
  • Tesis en acceso abierto en: Biblos-e Archivo
• Resumen

  • Most developmental theories describe processes that are cumulative, multivariate, and dynamic, inasmuch they build on previous states and involve multiple interrelated parts that unfold together over time. Latent Change Score (LCS) models are a powerful framework for examining such processes. Due to their flexibility to characterize change, dynamics, and leadlag relations between latent processes, these models have become very frequent in developmental research. Notwithstanding this popularity, there are several aspects regarding the interpretation, effectiveness, and application LCS models that have not been properly explored. First, the interpretation of the LCS model parameters and their connection with substantive mechanisms of change has not been comprehensively explained anywhere in the psychological literature. In particular, several key aspects of these models remain misunderstood to many behavioral, health, and social researchers. Second, the ability of LCS models to recover bivariate dynamics has not been sufficiently investigated. Consequently, researchers lack clear guidelines on the minimum sampling requirements for obtaining quality estimates. Third, the vast majority of LCS model specifications in the psychology literature are deterministic. However, it is unknown whether this model is capable of recovering bivariate dynamics when the trajectories are affected by stochastic innovations. Fourth, LCS models are an effective tool for the study of dynamics in accelerated longitudinal designs (ALDs), but they lead to unreliable parameter estimates when the trajectories are affected by cohort effects. Finding ways of controlling for cohort effects is crucial for a correct characterization of developmental processes in the context of ALDs. The present dissertation aims to address these gaps in three separate studies. Study 1 is a conceptual analysis and tutorial on the substantive meaning and interpretation of the LCS model parameters, as well as their connection with theoretical mechanisms of change. Study 2 is a systematic evaluation of the effectiveness of deterministic and stochastic LCS models to recover bivariate dynamics under a broad set of empirically relevant conditions. Factors such as sample size, number of repeated measures, and amount of stochasticity due to innovations are considered in a Monte Carlo study. Study 3 presents a novel extension of LCS models that allows controlling for cohort effects in the speed of maturation of developmental processes. The performance of the proposed model is evaluated through a Monte Carlo study under various combinations of sample size, number of repeated measures, and sampling schedules. Based on the findings derived from these studies, we provide practical guidelines and recommendations for the specification, estimation, and application of LCS models, as well as for the implementation of ALDs