Variable Selection in Measurement Error Models via Least Squares Approximation

• Activity Number: 319
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Guangning Xu*+ and Len Stefanski
• Companies: North Carolina State University and North Carolina State University
• Keywords: Least Squares Approximation (LSA) ; Measurement Error Model ; Variable Selection

Abstract

A fundamental problem in biomedical research is identifying key risk factors and determining their impact on health outcomes via statistical modeling. Due to device error and within-subject variation, some risk factors are measured with error. Ignoring measurement error can bias the estimates and adversely impact variable selection.. We propose a new method for variable selection in measurement error models by integrating well-established measurement error modeling methods with the least squares approximation (LSA) variable selection method of Wang and Leng (2007). The resulting estimators are consistent and asymptotically normal in the usual case that the measurement error corrected estimator is root-n consistent. The method inherits the oracle property when an adaptive penalty is used and the tuning parameter is well selected. The key advantage of our new method is that it provides a unified solution to the variable selection in measurement error models and greatly eases computing by using existing algorithms. Compared to existing methods for doing variable selection in the presence of measurement error, our approach is conceptually and computationally simpler.