Model-based clustering of mixed data with sparse dependence

Mixed data refers to a mixture of continuous and categorical variables. The clustering problem with mixed is long-standing statistical problem. latent Gaussian model, model-based approach for such problem, has received attention owing its simplicity interpretability. However, these approaches are prone dimensionality problems. Specifically, parameters must be estimated each group, the number covariance quadratic in To address this, we propose “regClustMD,” novel method that can sparse dependence among We consider assuming precision matrix between variables nonzero elements. maximizing penalized complete log-likelihood using Monte Carlo expectation-maximization (MCEM) algorithm. demonstrate our through simulation study real-world examples.