Using Partially Synthetic Frames to Evaluate Alternative Sample Designs for Estimating a Rare Business Characteristic

Katherine Jenny Thompson, U.S. Census Bureau 

Hang Joon Kim (University of Cincinnati) 

Stephen Kaputa (U.S. Census Bureau)

In the “traditional'” finite population sampling framework, the sample designer has a complete list (frame) of eligible units with classification information and auxiliary variables related to surveyed characteristics. In our setting, the frame auxiliary variables are weakly related to the survey characteristic, which is not present for most units. Hence, using frame auxiliary variables to assess survey design efficacy can be misleading. Instead, we propose generating multiple partially synthetic frames, modeling characteristic values for each unit on the frame, then drawing repeated samples from each synthetic frame using the candidate sample design(s) to assess finite sample performance for each design within and between the synthetic frames. Focusing on establishment survey data, we illustrate our proposed approach on a subset of industries surveyed annually by the Business Enterprise Research and Development Survey.

Katherine Jenny Thompson is the Senior Mathematical Statistician in the Economic Directorate of the Census Bureau. Jenny holds a masters of science degree in Applied Statistics from the George Washington University and an bachelor or arts degree in Mathematics from Oberlin College. She is an American Statistical Association (ASA) Fellow, an elected member of the International Statistics Institute, and the Vice President Elect of the ASA. She is the Survey Statistics Editor-in-Chief of the Journal of Survey Statistics and Methodology and an Associate Editor for the Journal of Official Statistics. She has published papers on a variety of topics related to complex surveys in several journals, including the Journal of Official Statistics, Journal of the Royal Statistical Society (Series A), Survey Methodology, Annals of Applied Statistics, International Statistical Review, Journal of Survey Sampling and Methodology, and Public Opinion Quarterly.