A Method to Identify Missing Data Mechanism in Incomplete Dataset

摘要

This paper proposes a sequential method of identifying missing data mechanism in incomplete dataset. A simulation study in the context of GNH dataset is designed to validate the proposed method. The dataset is explored for latent factors and a model is constructed from the discovery. Using this model, pseudorandom data are generated for a simulation study. Specific missingness probability is invoked in pseudorandom data to create respective MCAR, MAR and MNAR data. The proposed method is validated and tested for robustness by employing a specific statistical test for each of MCAR, MAR and MNAR data.