25 – Recent Research on Current Population Survey
Domain Estimation and Successive Difference Replication
Tim Trudell
U.S. Census Bureau
Khoa Dong
U.S. Census Bureau
Yang Cheng
U.S. Census Bureau
Eric Slud
U.S. Census Bureau, Center for Statistical Research & Methodology
Successive Difference Replication (SDR) is a variance estimation method originally used for systematic samples, which is applied to complex surveys at the U.S. Census Bureau, including the Current Population Survey. Often we are interested in estimates for domains or subpopulations. In practice, replicate factors are assigned to the full sample. For SDR variance estimation of an estimated domain total, we use a subset of the full replicate factors. The subsetting of the domain on the sorted full data translates to a skip pattern. The effects of different skip patterns on the SDR variance error are examined in this paper via ideal superpopulation models and a simulation study.