This talk will have two parts, both pertaining to finding latent groups with similar characteristics that evolve over time. First I’ll discuss longitudinal clustering with multivariate time series of varying lengths. The proposed model can be viewed as an extension of the normal mixture model to longitudinal data. We relax the assumption of local independence, and use covariates to predict the clustering assignments and transitions. Recursive relationships are derived that allow the computational cost of a generalized EM algorithm to grow linearly in time rather than having polynomial growth. In the second part of my talk, I’ll discuss community detection in dynamic networks. This is done by embedding the actors of the network in a latent space and performing longitudinal clustering within this space. The proposed approach captures the temporal nature of the data, can model directed or undirected edges, incorporate transitivity, and accounts for individual propensities to send and receive edges.