Variable Selection with Estimating Equations Using Decomposable Norm-Based Regularizers

• Activity Number: 562
• Type: Topic Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Yi Li*+ and Dave Zhao
• Companies: University of Michigan and Harvard University
• Keywords: estimating equations ; variable selection ; survival analysis ; risk assessment

Abstract

We propose a new estimating equation-based regularization method for simultane- ous estimation and variable selection. Our method can be used even when the number of covariates exceeds the number of samples, and can be implemented using well-studied algorithms from the non-linear constrained optimization literature. Furthermore, for a certain class of estimating equations and a certain class of regularizers, which includes the lasso and group lasso, we prove a finite-sample probability bound on the accuracy of our estimator. Our research was motivated by practical problems, from a genome-wide association study of non-small-cell lung cancer patients and a clinical trial of therapies for head and neck cancer, that are difficult to analyze under the likelihood setting. In simulations we show that our procedure outperforms competing methods, and we use it to analyze the aforementioned stud