Statistics Seminar

Abstract: Traditional regression models typically assume linear relationships among predictors and response variables, often neglecting complex dependencies and variations across different observations or variables. However, in numerous real-world scenarios, predictors exhibit structured interactions, and their effects on an outcome variable may vary depending on additional covariates. These types of scenarios commonly arise in fields like genomics, neuroimaging, healthcare, and psychology, where understanding relationships among variables is critical. I will introduce novel Bayesian network-guided sparse regression models with varying-coefficients, that simultaneously achieves variable selection and infer a network among predictors. I will consider continuous and mixed-type predictors, as well continuous and binary responses. Using simulation studies, I will show that integrating network information into feature selection improves power to detect the true predictors, also outperforming regularization. I will show applications to genomic data as well as to clinical data involving adolescents diagnosed with eating disorders.