Group-dependent finite mixture model

We present a Bayesian nonparametric group-dependent mixture model for clustering. This is achieved by building a hierarchical structure, where the discreteness of the shared base measure is exploited to cluster the data, between and within groups. We study the properties of the group-dependent clustering structure based on the latent parameters of the model. Furthermore, we obtain the joint distribution of the clustering induced by the hierarchical mixture model and define the complete posterior characterization of interest. We construct a Gibbs sampler to perform Bayesian inference and measure performances on simulated and a real data.