Multiple Imputation for High-Dimensional Data

• Activity Number: 159
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Yize Zhao*+ and Qi Long
• Companies: Emory University and Emory University
• Keywords: Missing Data ; Multiple Imputation ; Regularized Regression

Abstract

In biomedical and social studies missing data are often encountered, presenting challenges in data analysis. Multiple imputation (MI) methods are widely-used tools for missing data problems, replacing each missing value with a set of plausible values to allow for subsequent standard statistical analysis as well as account for the uncertainty about missing values. However, in the presence of high-dimensional data, existing multiple imputation methods often are not applicable or do not perform well. In this research, we consider a case of missing at random (MAR) and investigate new multiple imputation approaches that can handle high-dimensional data. Simulation studies are conducted under different scenarios to evaluate the performance of the proposed methods and compare them with existing MI methods. The methods are further illustrated using gene expression data collected from a cancer study.