Research Abstracts Online
2008 - March 2009
University of Minnesota Twin Cities
College of Liberal Arts
School of Statistics
PI: Lan Wang
High-dimensional Variable Selection for Correlated Data
This project investigates the problem of variable selection in statistical regression models when the number of covariates is large. Correlated data frequently occurs in many fields, such as biomedical and health sciences, economics, social sciences, and environmental studies. Dependence structure arises naturally when repeated measurements are taken on the same subjects over time or when observations are made on each individual within a cluster. Correlated data structure allows investigation of events that occur in time (such as aging) and is essential in studying the temporal patterns of response to treatments. Most of the vast existing literature on statistical model selection focuses on the independent data case and largely neglects the important area of correlated data. This project will develop a set of semi-parametric and nonparametric tools for model selection for high-dimensional correlated data. In particular, the research focuses on the important and popular semi-parametric marginal regression analysis approach based on the generalized estimating equations.