A Multivariate Stopping Rule for Survey Data Collection
Xinyu Zhang is a PhD candidate studying survey and data science at the University of Michigan. He is primarily interested in responsive survey designs, survey nonresponse, and machine learning techniques. His dissertation topic is using models to inform responsive survey designs.
Abstract
Surveys are experiencing declining response rates. With more and more effort expended to combat these declining response rates, the cost of large-scale surveys has continued to rise. Recent technological developments in survey data collection have allowed the survey designer to make near-real-time intervention decisions. Stopping rules are one of the interventions often considered to improve the efficiency of data collection. Stopping some cases essentially reallocates effort from stopped cases to others, but most previously proposed stopping rules have only considered single estimates. In multipurpose surveys, there may be data quality objectives that must be met for multiple estimates with constraints on costs. We introduce a stopping rule that accounts for the cost and the quality of one or more estimates. The proposed stopping rule is illustrated via simulation using data from the Health and Retirement Study.