Restricted Latent Class Multiple Imputation Method of Categorical Missing Data

• Activity Number: 444
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Qiao Ma*+
• Companies: University of Nebraska - Lincoln
• Keywords: Categorical data ; missing data ; multiple imputation ; latent class model ; simulation study ; survey nonresponse

Abstract

Multiple imputation is a commonly used method to deal with incomplete data sets and is used by researchers on many different analytical levels. Imputation substitutes missing data with some values instead of discarding the entire case from the analysis. While dealing with large data sets with more than a few incomplete categorical variables, it is not possible to apply log-linear modeling due to limitations of sparseness. It is because we are not able to set up and process the full multi-way cross-tabulation required for the log-linear analysis. The latent class model is a plausible multiple imputation tool to solve this problem (Vermunt 2008). Another possible solution of a limited number of categorical variables associated with the log-linear method is to use hot-deck imputation (Rubin 1987). In this study, several multiple imputation methods for large categorical datasets will be tested. An advanced restricted latent class model-based multiple imputation method is proposed to be a better, more representative approach than the unrestricted latent class model since it specifies equality and inequality constraints on sums of conditional response probabilities.