Two-Level Clustering of Patients and Hospitals via Thinned Dependent Dirichlet Process Mixtures
摘要
The Collaborative Perinatal Project is a large study aimed at identifying the risks and complications of pregnancies. Data about the pregnancy and the birth outcome were collected on a large number of women from 12 different hospitals over several years. We focus on the duration of the pregnancy, and we investigate the heterogeneity of the gestational age among women and between hospitals. We apply a hierarchical mixture model based on the thinned dependent Dirichlet process of [2], and we show how this approach is able to simultaneously convey a clustering of patients and hospitals.