Dynamic clustering: classifying people through ecological momentary assessment

Nowadays, smartphones and tablets are enabling researchers to collect time-intensive data through ecological momentary assessment. During ecological momentary assessment people are asked to report on their feelings and experiences multiple times a day, for one or several weeks. The availability of the resulting intensive longitudinal data has brought about a shift towards studying within-individual dynamics in the social sciences. Often, researchers are interested in summarizing the within-individual dynamics of several individuals into a common longitudinal model. As dynamics can be rather heterogeneous across individuals, one needs sophisticated tools to express the essential similarities and differences across individuals. A way to proceed is to identify subgroups of individuals who are characterized by distinct differences in their dynamics. The aim of this dissertation is to develop novel dynamic clustering procedures that account for between-individual differences in within-individual dynamics. This goal is achieved by proposing dynamic clustering procedures that uncover clusters of individuals who exhibit qualitatively different dynamic processes in intensive longitudinal data. Thereby unknown subgroups of individuals with similar dynamics are identified. Dynamic clustering allows the information of several individuals to be pooled, while accounting for qualitative between-individual heterogeneity in the underlying dynamics.