Model-Based Methods for Missing Data in Surveys with Post-Stratification Information

• 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): Sahar Zangeneh*+ and Roderick J. Little
• Companies: Fred Hutchinson Cancer Research Center and University of Michigan
• Keywords: Post-Stratification ; Maximum Likelihood ; Unit nonresponse ; MNAR ; Model-based ; Missing-data

Abstract

We study maximum likelihood estimation of the population mean for a survey experiencing unit nonresponse, i.e., when a sampled unit does not respond to the entire survey. We consider situations where post-stratification information is externally available for the population. Without external information, unit nonresponse, may lead to missing-data mechanisms that are missing not at random (MNAR), which generally require a model for the missing-data mechanism. We develop a new model-based approach to weaken the missing at random (MAR) assumption by inclusion of external information for situations where the data are MNAR in the classical sense defined by Rubin (1976), but post-stratification information is externally available. This framework is then extended to also incorporate covariate information that is fully observed for the sampled units. We compare and contrast the proposed model-based method to existing design-based methods empirically for incomplete categorical data.