Regularized Robust Regression for Quantitative Genetic Traits

• Activity Number: 139
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Chad He*+ and Yanhua Wang and Linglong Kong and Sijian Wang
• Companies: Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center and University of Alberta and University of Wisconsin
• Keywords: Quantitative Traits ; Regularized Regression ; Quantile Regression ; Robust Estimation ; Genomic Features

Abstract

Genetic association studies often involve quantitative traits, such as Body Mass Index, blood pressure, or lipids level. To identify the genetic variants underlying these traits can help uncover the etiology of a number of complex diseases. Currently, many methods are available for the genetic association analysis of quantitative traits, but the majority of them is based-on least square estimates and may not be robust to outliers due to biological variation or inaccurate measurement. High-dimensional genomic features associated with these studies further complicate the analysis. Quantile regression represents a robust estimation method and is able to describe the underlying structure in a more complete manner. Here, we introduce a regularized quantile regression method that is able to characterize the underlying genetic structure in a more integrative manner, and at the same time is able to account for the potential genetic heterogeneity. We investigate the theoretical property of our method, and examine its performance through a series of simulation studies.