Score-Guided Recursive Partitioning of Continuous-Time Structural Equation Models

Model-based recursive partitioning is a powerful approach to analyzing heterogeneity between subjects. In the past decade, the semtree software package has established itself as one of the primary tools for the recursive partitioning of structural equation models (SEM). The resulting SEM trees parti...

Descripción completa

Detalles Bibliográficos
Autores: Arnold, Manuel, Cáncer, Pablo F., Estrada Alonso, Eduardo, Voelkle, Manuel C.
Tipo de recurso: capítulo de libro
Fecha de publicación:2024
País:España
Institución:Universidad Autónoma de Madrid
Repositorio:Biblos-e Archivo. Repositorio Institucional de la UAM
Idioma:inglés
OAI Identifier:oai:repositorio.uam.es:10486/742460
Acceso en línea:https://hdl.handle.net/10486/742460
https://dx.doi.org/10.1007/978-3-031-56318-8_3
Access Level:acceso abierto
Palabra clave:continuous time
decision trees
longitudinal data
model-based recursive partitioning
score-based tests
stochastic differential equation
Psicología
Descripción
Sumario:Model-based recursive partitioning is a powerful approach to analyzing heterogeneity between subjects. In the past decade, the semtree software package has established itself as one of the primary tools for the recursive partitioning of structural equation models (SEM). The resulting SEM trees partition the sample into groups of similar individuals while identifying the most important predictors of group differences in the process. However, until recently, an ad hoc covariate testing procedure that was computationally demanding and favored the selection of certain covariates over others hindered the partitioning of complex SEMs. These hurdles have been overcome by selecting covariates utilizing score-based tests, which offer unbiased covariate selection and drastically reduce the runtime of trees. In this chapter, we show how semtree can be used to uncover heterogeneity in dynamic structural equation models for longitudinal data, focusing on continuous-time (CT) models. Unlike the more widely used discrete-time (DT) models, CT models do not require the time intervals between measurements to be equal and, therefore, can adapt effortlessly to irregular sampling schemes. Thus, our resulting approach, which we call score-based CTSEM trees, is well suited to deal with heterogeneity between individuals and measurement occasions. We illustrate the approach with empirical data from the Survey of Health, Ageing and Retirement in Europe