A note on the properties of estimators in missing data analysis

Abstract

In the missing mechanism, missing at random (MAR) is sometimes assumed when data has missing values. When MAR holds and the true distribution belongs to the assumed statistical model, the maximum likelihood estimator based on the observed data has consistency. Based on a weaker condition than MAR, this study investigates the properties of the estimators obtained by applying the maximum likelihood method and the Bayesian method when the true distribution does not belong to the statistical model.

Keywords:

• Bayes estimator
• consistency
• maximum likelihood estimator
• missing at random
• model misspecification