Reducing parameter estimation bias for data with missing values using simulation extrapolation
Date
Authors
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Altmetrics
Authors
Research Projects
Organizational Units
Journal Issue
Is Version Of
Versions
Series
Department
Abstract
Missing data is a common problem in data analysis, and has been studied extensively. We propose using simulation extrapolation(SIMEX), a general simulation-based approach to adjust the bias in the estimator due to missing values assuming the model for missingness is known. The SIMEX approach was originally proposed for measurement error models. The SIMEX method includes simulation steps that use information from the missing mechanism and an extrapolation step to adjust the bias. While EM and multiple imputation methods rely on the correct assumptions on the conditional distribution of missing data given observed data, the proposed SIMEX method assumes the correct model for the missingness. Therefore, SIMEX is more robust on an incorrect specification of the probability model for the unobserved data. We discuss the properties of the SIMEX estimator and compare this method with existing methods using
simulation. The advantages and limitations of our approach are also discussed.