Investigation of multiple imputation procedures in the presence of missing quantitative and categorical variables

The presence of missing or incomplete data is a ubiquitous problem in real world datasets. In the thesis, we apply multiple imputation procedures to analyze incomplete multivariate datasets. We consider a dataset that contains both continuous and categorical variables, all with some missing values. While investigating three other imputation methods, we propose a two-part combination model, which melds the general linear model and the logistic model together, to predict and impute the missing values. Based on R2 and half-bound criteria, we analyze the different effects on variability due to the proportion of data missing, due to the association structure of the missing data, and due to the imputation procedure used.