The Interplay Between the Practice and Theory of Survey Sampling: Past, Present, and Future

• Activity Number: 482 - Recent Developments in Survey Sampling: Session in Honor of J.N.K. Rao's 80th Birthday
• Type: Invited
• Date/Time: Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
• Sponsor: SSC
• Author(s): Graham Kalton*
• Companies: Westat
• Keywords: systematic sampling ; stratified sampling ; multi-stage sampling ; two-phase sampling ; imputation ; calibration

Abstract

This paper reviews the development of survey sampling practice and theory from the beginning of the twentieth century. The theoretical foundations of survey sampling design-based inference were presented in Neyman's 1934 landmark paper. This design-based approach is the norm for large-scale surveys; model-assisted methods are widely used in design and analysis but inference does not depend on the validity of statistical models. Since then, a variety of sampling methods has been devised and the associated theory developed. The need to estimate the precision of a survey's estimates from the selected sample has led to the development of theory for variance estimation for complex sample designs, including various replication methods. Over the years the issue of design-based versus model-based inference has often been debated. Even with the design-based approach, survey estimates are becoming more model-dependent because of rising nonresponse rates. Greater reliance on model-dependent inference has also occurred because of user demands for small area estimates, because of the desire to study hard-to-survey populations, and through the use of nonprobability methods for web surveys.