Multilevel Conference 2026 - Modelling intra-individual differences in variability using 4-level hierarchical dynamic structural equation models (DSEM): How to optimally trade-off fixed and random effects with respect to model complexity?
Date:
Within the CODEC project (Coolen et al., 2024) children’s performance is measured on five different cognitive tasks in 15 sessions over an in-class burst week. We hope to estimate children’s variability in reaction time at the trial, session, and day timescale, investigating individual differences and their interrelatedness. To do so, we expand proprietary, 2-level hierarchical dynamic structural equation models to open-source (R/Rstan), 4-level (and up) models. Several challenges remain in model development and parameter estimation. How can we best investigate variability at the population and individual level, accounting for nesting, the trade-off between fixed and random effects, and model convergence?