Disentangling Individual Dynamics: An Adaptive Dynamic Clustering Model for Longitudinal Data

Standard multi-subject time series models assume all individuals in a gathered sample to exhibit equal dynamic processes, resulting thus in an aggregation that may not apply at the individual level. As a result, an easily applicable method is needed to identify groups of individuals who exhibit distinct dynamic processes based on intensive longitudinal data. We propose an adaptive clustering procedure which can bridge the gap between ideographic and nomothetic time series analysis approaches by identifying groups of individuals exhibiting different dynamics. Accounting for individual differences in time series can enable accurate descriptions of dynamic processes at the individual and the population level.