Integrating M-Regression with False Discovery Rates for Outlier Detection in Genetic Association Studies of Quantitative Traits

• Activity Number: 240
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Vanda Lourenco*+ and Ana Maria Pires
• Companies: CMA, FCT - Universidade Nova de Lisboa and CEMAT, IST - Universidade Técnica de Lisboa
• Keywords: Robust regression ; Robust outlier test ; False discovery rate ; Single nucleotide polymorphism ; Genetic association studies

Abstract

Outlier detection is a widely discussed issue in the literature in many areas of research one of which is the field of genetic association studies. Here, as far as we have seen, the motivation behind is data cleaning so that classical models may be used in the analysis. Knowledge of these observations however, may be important in the assessment of the underlying mechanisms of that data, since outliers are not always a result of measurement error. Robust multiple linear regression methods have been shown to be a valuable asset in genetic association studies, allowing us not to be concerned with the eventuality of outliers in the data disrupting our analysis results. These methods also provide an adequate setting for searching for outliers since the residuals from robust models are not usually affected by outlying observations. To this respect, we propose and discuss a robust outlier test together with an appropriate robust estimate of scale and an adequate FDR correction measure, to be used in the context of a robust multiple-regression model. We illustrate the good performance of the robust outlier test through an application that uses a real genetic data set from the literature.