Abstract:
|
In designing the 2020 Post-Enumeration Survey sample, we conduct a simulation study by selecting repeated samples from a research frame of basic collection units. The post-enumeration survey uses two different housing unit counts - initial counts from the U.S. Census Bureau Master Address File and updated counts from listing - to construct measure of sizes for stratification and sample allocation. The initial housing unit count is available for every basic collection unit on the research frame, but not the listing count. It is too costly to list every basic collection unit on the research frame. In this paper, we compare two imputation methods to simulate the housing unit counts from listing. In the first method, we use sample and census data to model the listing housing unit counts. In the second method, we use sample data from the 2010 post-enumeration survey as donors to impute listing housing unit counts. We compare the results of the two imputation methods, and decide to use the imputation method that uses donors from the 2010 post-enumeration survey to create listing housing unit counts for the 2020 Post-Enumeration Survey research frame.
|