Current Methods for Capturing Heterogeneity in Dynamic Psychological Processes with Multivariate Intensive Longitudinal Data

The collection of intensive longitudinal data (ILD; i.e., time series data from a sample ofindividuals) has gained momentum in the behavioral sciences due to an increased interest instudying within-person dynamic processes. Because of the lack of ergodicity in psychologicalprocesses, it is often invalid to generalize results from the population level to the individuallevel, and vice versa. However, establishing conditional ergodicity may allow for such cross-level generalizations, thus bridging the gap between idiographic and nomothetic analyses (Adolf& Fried, 2019; Beltz et al., 2016; Voelkle et al., 2014). This review paper focuses on the firstpremise of conditional ergodicity: accounting for heterogeneity in within-person dynamics. Wefirst elucidate the taxonomy of sample heterogeneity in within-person dynamics, then reviewvarious dynamic subgrouping methods to capture sample heterogeneity with multivariate timesseries data. We summarize the promises and pitfalls of each method, grouping them into twocategories: distance-based approaches and distributional approaches. In the discussion, weprovide cautious notes and highlight decision points to help researchers and practitioners choosethe methods that best match their research goals and theoretical concerns.