Model selection for maternal hypertensive disorders with symmetric hierarchical Dirichlet processes

Hypertensive disorders of pregnancy occur in about 10% pregnant women around the world. Though there is evidence that hypertension impacts maternal cardiac functions, relation between and dysfunctions only partially understood. The study this relationship can be framed as a joint inferential problem on multiple populations, each corresponding to different hypertensive disorder diagnosis, combines multivariate information provided by collection function indexes. A Bayesian nonparametric approach seems particularly suited for setup, we demonstrate it dataset consisting transthoracic echocardiography results cohort Indian women. We are able perform model selection, provide density estimates indexes latent clustering patients: these readily interpretable outputs allow single out modified functions patients, compared healthy subjects, progressively increased alterations with severity disorder. analysis based relies novel hierarchical structure, called symmetric Dirichlet process. This suitably designed so mean parameters identified used selection across penalization multiplicity enforced, presence unobserved relevant factors investigated through subjects. Posterior inference suitable Markov chain Monte Carlo algorithm, behaviour also showcased simulated data.