Coordinated Sample Design for Multiple Surveys
Conference
64th ISI World Statistics Congress
Format: IPS Abstract
Session: IPS 109 - Data Integration in Survey Sampling
Thursday 20 July 2 p.m. - 3:40 p.m. (Canada/Eastern)
Abstract
We consider six exemplar surveys that use the Common Core Data (CCD) as the sampling frame of U.S. public schools. Taking into account the survey stratification and eligibility restriction across surveys, we focus on the joint select probability estimation for each school to be selected for all surveys subject to the response burden. We have developed four strategies for a coordinated sampling process to potentially reduce response burdens: 1) Independently select schools for each survey, compute the burden for each selected school, and randomly substitute schools from the same stratum. 2) Independently select schools for each survey, compute the burden for each selected school, and reject samples that exceed the burden, 3) Sequentially sample schools based on a random survey order and decrease the selection probability for schools selected in previous surveys, and 4) Use matrix sampling of to assign surveys to schools using a probabilistic mechanism, i.e., create replicates. None of these methods will work without the “centralization” of sampling activities. We conduct simulation studies to compare different strategies for the improvement of data quality and implementation feasibility. This is joint work with Mike Elliott, Roderick Little and Trivellore Raghunathan at University of Michigan.