Direct and Unbiased Multiple Imputation Methods for Missing Values of Categorical Variables

Missing data is a common problem in statistical analyses. To make use of information with incomplete observation, missing values can be imputed so that standard methods used to analyze the data. Variables are often categorical and miss ing pattern may not monotone. Currently, commonly imputation for non-monotone do allow di rect inclusion variables. Categorical variables converted numerical before imputation. For many applications, those must then back values. However, this conversion introduces bias which seriously affect subsequent In paper, we propose two direct pattern: approach incorporated expectation maximization algorithm new algorithm: imputation-maximization algorithm. Simulation studies show both perform better than method using vari able conversion. An application real provided compare variable