ABSTRACT
The traditional methods of missing value processing mainly include deletion, imputation, and likelihood maximum (Ye, 2014). Imputation of missing values include imputation mean, random imputation method, and Multiple Imputation (MI) (Zhao, 2013; Yozgatligil, 2013). Multiple imputation method is widely applied in psychology, medicine, finance, climatology, pharmaceutics (Ji, 2013; Twisk, 2013; Donneau, 2015). Loss mechanism of missing data includes Missing Completely At Random (MCAR), Missing At Random (MAR), and Missing Not At Random (NMAR). Ye (2014) provides an inspection and identification method. Yang (2012) analyzed the effect of the full Bayesian and partial Bayesian methods on parameter estimation with different missing ratios. However, these studies do not deal with missing data based on the classification of data.
