Decision-Tree Models for Survey Nonresponse

• Activity Number: 190
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Government Statistics
• Author(s): Amanda K. McCracken*+
• Companies: The University of Alabama
• Keywords: survey nonresponse ; establishment survey ; classification tree

Abstract

Nearly all surveys are affected by nonresponse to some degree. Predicting survey nonresponse can help minimize its impact on estimates obtained from surveys, allowing survey organizations to focus on obtaining information from expected nonresponders or to adjust for this nonresponse. Unfortunately, modeling efforts are often limited by the unavailability of information about nonrespondents. However, in some cases, much information is known about sample members prior to conducting a survey. Many prior nonresponse modeling efforts have utilized regression models and focused on household surveys. This poster examines nonresponse in an establishment survey rather than a household survey and takes a different approach to modeling nonresponse-- the classification tree-which has several advantages over regression modeling.