Likelihood-based inference for singly and multiply imputed synthetic data under a normal model

摘要

Likelihood-based inference for both singly and multiply imputed synthetic data is developed in this paper under a univariate normal model and two distinct data generation scenarios, namely, posterior predictive sampling and plug-in sampling. We show that valid and exact inference can be drawn in both scenarios. Some theoretical issues of multiply imputed datasets under posterior predictive sampling are also pointed out. Published by Elsevier B.V.