Online Program

Handling Data with Three Types of Missing Values

*Jennifer Boyko, University of Connecticut 
Ofer Harel, University of Connecticut 

Keywords: missing data, multiple imputation

Incomplete data is a common obstacle to data analysis in a variety of fields. Values in a data set can be missing for several different reasons including failure to answer a survey question, dropout, planned missing values, intermittent missed measurements, latent variables, and equipment malfunction. In fact, many studies will have more than just one type of missing value. Appropriately handling missing values is critical in the inference for a parameter of interest. Many methods of handling missing values fail to account for uncertainty due to the missing values which can lead to biased estimates and over-confident inferences.

One area which is still unexplored is the situation where there are three types of missing values in a study and the differences between the three types are of interest. The development of a three stage multiple imputation approach would be beneficial in analyzing studies with several types of missing values. Three stage multiple imputation would also extend the benefits of standard multiple imputation and two stage multiple imputation, namely the quantification of the variability attributable to each type of missing value and the flexibility for greater specificity regarding missing data assumptions.