Finding Signals in the Noise: Modeling Intra-Individual Variability with Dynamic and Multilevel Approaches

Abstract: 

Intra-individual variability (IIV) is often treated as noise, yet it can provide meaningful insights into psychological processes. This talk focuses on dynamic models that capture time-varying variance and covariance structures, moving beyond traditional static approaches.

First, we will discuss Multivariate GARCH (MGARCH) models, specifically BEKK and Dynamic Conditional Correlation (DCC), which decompose residual variance into stable, autoregressive, and shock-driven components. These models allow for dynamic estimation of intra-individual fluctuations, potentially revealing patterns that are overlooked in standard models. Building on this, I will introduce a Multilevel VAR-DCC framework that extends dynamic modeling to hierarchical data, capturing individual differences in time-varying covariance structures.

Applications to intensive longitudinal data, including dyadic affect dynamics and ecological momentary assessments (EMA), illustrate the potential of these models to uncover meaningful signals in within-person variability. By treating heteroskedasticity as an informative process, this vein of models offers a new perspective on psychological dynamics.