Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 285 - New Advances in Sample Design and Adjusting for Survey Nonresponse
Type: Contributed
Date/Time: Wednesday, August 11, 2021 : 1:30 PM to 3:20 PM
Sponsor: Survey Research Methods Section
Abstract #317830
Title: Variational Bayesian Multiple Imputation for Clustered High-Dimensional Data
Author(s): Qiushuang Li* and Recai Yucel
Companies: University at Albany SUNY and Temple University
Keywords: Clustered data; variational inference; multiple imputation; sequential hierarchical regression imputation; calibration-based imputation; spike-and-slab variable selection

Multiple imputations have become one of the standard methods in drawing inferences in many in- complete data applications. Applications of multiple imputations in relatively more complex settings with clustered high-dimensional data require specialized methods to overcome the computational burden. Using mixed-effects models, we develop methods that can be applied to continuous, binary, or categorical incomplete data. These methods are computationally feasible as they rely on variational Bayesian inference for sampling the posterior predictive distribution missing data. These methods specifically target high-dimensional covariates and work with spike-and-slab priors. The individual regression computation is then incorporated in a sequential hierarchical regression imputation. To further improve the computational efficiency for the categorical data, we apply the calibration-based imputation methods using the approximating variational distribution. We present a simulation study to validate the usefulness of the proposed approach.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2021 program