WITHDRAWN Changes at the Margins: Evaluating MRP Performance for Data Sets with Clustered Sparsity

• Activity Number: 352 - Small Area Estimation, Analysis of Complex Sample Survey Data, and New Advances for Health Surveys
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Survey Research Methods Section
• Author(s): Robert Petrin and Kevin Fenton and Meng Li
• Companies: Ipsos Public Affairs and Ipsos Public Affairs and Ipsos Public Affairs
• Keywords: MRP; Sparsity; Design; Health; Polling

Abstract

Multilevel regression with poststratification (MRP; MRP; e.g., Gelman, 2007; Ghitza and Gelman, 2013) and its extensions have made important contributions to addressing common problems in survey research, most notably frame error and sparse data. Despite its tremendous flexibility, researchers as yet have limited guidance on how studies can be designed to optimize MRP performance when they have at least some degree of control over the data collection.

The purpose of this paper is to understand MRP’s performance in scenarios common in public health, behavioral, and opinion research with highly clustered sparsity. A particular focus of this paper will be determining optimal sample sizes required along poststratification table margins; another will be quantifying the impact reasonable auxiliary variables have on data requirement. Results will be based on a comprehensive and realistic empirical simulation study derived from an application of MRP to understanding family migration at the county level in the United States.