A Study of Data-Collection Rules Involving Real-Time Imputation for Adaptive Survey Design

• Activity Number: 633
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Social Statistics Section
• Author(s): Darcy Steeg Morris*+ and Yves Thibaudeau
• Companies: U.S. Census Bureau and U.S. Census Bureau
• Keywords: Adaptive Design ; Imputation Methods ; Decision Theory ; Survey Data Collection ; Missing Data

Abstract

Adaptive design is a means to reduce survey costs by taking focus away from maximizing response rate exclusively. An optimal adaptive procedure would minimize factors such as cost and respondent burden, while maintaining or improving the quality of the estimates. This work theoretically and empirically assesses the quality of key estimates from the National Health Interview Survey (NHIS) under various adaptive sampling data collection rules. We examine simple data collection decision rules that involve real-time imputation for non-interviews and rely on estimates of errors accounting for non-response in survey indicators. We present the decision-theoretical framework (DeGroot, 1970) connecting together the decision rules, the imputation methods, and the error estimation. We run simulations to explore the complications of the situations motivating the rules, as well as more elaborate rules that use multiple imputation to deal with the added complexities of the missing data mechanism (Little and Rubin, 2002).