Stat Colloquium [Virtual]: Dr. Honglang Wang
Indiana University-Purdue University Indianapolis
Title: Empirical Likelihood Based Efficient Inference for Sparse Functional Data Accounting for Within-Subject Correlation
Abstract: We enhance the functional mean estimator and the semiparametric profile estimator for analyzing sparse functional data, specifically addressing the challenge of within-subject correlation. Our refined estimators significantly boost the efficiency compared to the traditional local kernel smoothing estimator, which operates under the assumption of an independent correlation structure. The empirical likelihood (EL) based inference is proposed for the functional mean curve as well as for the parameters of interest in the semiparametric model. The proposed methods perform favorably in finite sample applications from our simulation studies as well as real data application to the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study and the Genetic Analysis Workshop 18 (GAW18) study.