Reducing Parameter Estimation Bias for Data with Missing Values Using Simulation Extrapolation

• Activity Number: 670
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): YU-YI Hsu*+ and Yongming Qu and Alicia Carriquiry
• Companies: Iowa State University and Eli Lilly and Company and Iowa State University
• Keywords: Simulation Extrapolation ; Missing Data

Abstract

Missing data is a common problem in data analysis, and has been studied extensively. We propose a general simulation-based approach to adjust bias that relies on a correctly specified model for the missing mechanism. The simulation extrapolation (SIMEX) approach was originally proposed by Cook and Stefanski (1994) in the context of measurement error problems. The SIMEX method includes simulation steps that use information from the missing mechanism and an extrapolation step to adjust bias. While EM and multiple imputation methods rely on the correct assumptions about the conditional distribution of missing data, the proposed SIMEX method assumes that the missingness model is correct. Hance, SIMEX is more robust to mis-specifications of the distribution of unobserved data. We discuss the properties of the SIMEX estimator and compare its performance to existing methods through simulation.