多施設青年期脳および認知発達研究のベイズ統合混合モデリングフレームワーク

Integrating high-dimensional, heterogeneous data from multi-site cohort studies with complex hierarchical structures poses significant variable selection and prediction challenges. We extend the Bayesian Integrative Analysis and Prediction (BIP) framework to enable simultaneous variable selection and outcome modeling in data of a multi-view nested hierarchical structure. We apply the proposed Bayesian Integrative Mixed Modeling (BIPmixed) framework to the Adolescent Brain Cognitive Development (ABCD) Study, leveraging multi-view data, including structural and functional MRI and early life adversity (ELA) metrics, to identify relevant variables and predict the behavioral outcome. BIPmixed incorporates 2-level nested random effects to enhance interpretability and make predictions in hierarchical data settings. Simulation studies illustrate BIPmixed's robustness in distinct random effect settings, highlighting its use for complex study designs. Our findings suggest that BIPmixed effectively integrates multi-view data while accounting for nested sampling, making it a valuable tool for analyzing large-scale studies with hierarchical data.