Stephanie Coffey is a senior mathematical statistician in the Center for Optimization and Data Science at the U.S. Census Bureau. Her research is focused on the use of survey data, paradata, and other external data sources to improve the quality and cost properties of Census-conducted surveys, often through adaptive and responsive survey designs. In this work, Stephanie has collaborated with several federal agencies, including the National Center for Science and Engineering Statistics, the National Center for Education Statistics, and the National Center for Health Statistics. Stephanie is the co-chair of the Federal Committee on Statistical Methodology (FCSM) Adaptive Design Interest Group, and she recently received her doctoral degree from the Joint Program in Survey Methodology from the University of Maryland.
Optimizing the Cost-Quality Tradeoff in a Responsive Design Setting
Adaptive and responsive survey designs rely on estimates of survey data collection parameters (SDCPs), such as response propensity, to make intervention decisions during data collection. These interventions are made with some data collection goal in mind, such as maximizing data quality for a fixed cost or minimizing costs for a fixed measure of data quality. As a result, the quality of the estimates of the SDCPs influences which cases are identified for interventions and the ultimate effectiveness of intervention decisions. Higher quality estimates of SDCPs lead to more optimal intervention decisions, allowing survey managers to betrer balance quality goals with external constraints, such as cost.
This talk discusses an experiment where a Bayesian framework was employed to improve estimates of SDCPs during data collection, which were then used in real-time simulations to identify cases for intervention. The use of Bayesian methods introduced modest improvements in the predictions of SDCPs, especially early in data collection, when interventions would have the largest effect on survey outcomes. Additionally, the experiment resulted in significant data collection cost savings without having a significant effect on a key survey estimate. While there are many areas for future research in this area, this experiment suggests that Bayesian methods can improve predictions of SDCPs that are critical for adaptive and responsive data collection interventions.
To register for this event, please contact Lucy Robles at lrobles1 [at] umd [dot] edu