Bayesian Model Averaging in Multiple Imputation Under Informative Sampling

• Activity Number: 638 - Recent Development of Bayesian Methods in Survey Sampling
• Type: Invited
• Date/Time: Thursday, August 3, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Gyuhyeong Goh* and Jae-kwang Kim
• Companies: Kansas State University and Iowa State University
• Keywords: Bayesian model averaging ; Complex sampling design ; Multiple imputation ; Spike-and-slab prior

Abstract

In survey sampling, multiple imputation provides an effective way to handle missing data. When a large number of possible models are under consideration for the data, the multiple imputation is typically performed under a single selected model from the large pool of candidates. This model selection approach ignores the model uncertainty and so leads to underestimation of the variance of multiple imputation estimator. In this paper, we propose a new multiple imputation procedure for accounting for the model uncertainty under informative sampling design. We use the Bayesian model averaging (BMA) to incorporate all possible models into the imputation procedure. Some theoretical properties of the proposed method are investigated. A simulation study demonstrates that our model averaging approach provides better estimation performance than a single model selection approach under complex sampling design.