B1374
Title: Forward selection in ultra-high dimensional functional concurrent models with applications to functional GWAS
Authors: Lily Wang - George Mason University (United States) [presenting]
Shan Yu - Iowa State University (United States)
Rodrigo Plazola-Ortiz - Iowa State University (United States)
Yehua Li - University of California at Riverside (United States)
Abstract: In a functional genome-wide association study (fGWAS) dataset from the Alzheimer's Disease Neuroimage Initiative (ADNI), the longitudinal Alzheimer Disease (AD) biomarkers are modeled by a class of concurrent functional linear models. The model includes the functional effects of environmental covariates, ultra-high dimensional genetic covariates and their interactions. We approximate the coefficient functions using B-splines, and select important main effects and interactions using a forward selection procedure based on a functional Bayesian Information Criterion (fBIC). The proposed fBIC is adaptive to both sparse and dense functional data, leads to a consistent variable selection procedure when the dimension of covariates is fixed; enjoys the sure screening property and leads to less false positive than existing methods when the covariate dimensionality is ultra-high. Simulation studies confirm the properties and the analysis of the ADNI data leads to new findings on AD-related environmental, genetic and interaction effects.
