A Bayesian latent class model for integrating multi-source longitudinal data: application to the CHILD cohort study
针对多源纵向数据，提出一个贝叶斯潜类模型，允许每个数据源有独立的聚类结构，再整合为全局聚类，并通过CHILD队列研究数据和模拟研究验证其有效性。
Abstract Multi-source longitudinal data have become increasingly common. This type of data refers to longitudinal datasets collected from multiple sources describing the same set of individuals. Representing distinct features of the individuals, each data source may consist of multiple longitudinal markers of distinct types and measurement frequencies. Motivated by the CHILD cohort study, we develop a model for joint clustering multi-source longitudinal data. The proposed model allows each data source to follow source-specific clustering, and they are aggregated to yield a global clustering. The proposed model is demonstrated through real-data analysis and simulation study.