376 – Contributed Oral Poster Presentations: Section on Statistical Computing
Targeting Minorities Using Address-Based Sampling: A Simulation Study
Pedro Saavedra
ICF International
Francine Barrington
ICF International
Address Based Sampling (ABS) has the advantage of allowing the use of census files to oversample minorities. This study uses a file listing zip code areas and their population for three ethnic groups to identify the optimal strategy for oversampling minorities to maximize the minimum effective sample size for each minority group. Several different sampling strategies were simulated, assuming the ability to randomly sample from zip code areas or combinations of zip code areas. One of the strategies involved sampling zip code areas with replacement using various measures of size that took into account the number of residents in each of the three groups (African-American, Hispanic and other) and then randomly sampling a resident for each time a zip code was sampled. A second set of approaches created strata based on the prevailing minority and the proportion of that minority in the zip code area, and then sampled strata with high minority residence at a higher rate. A third set of approaches used a self-weighting sample and combined or cumulated it with a second sample that used one of the strategies mentioned above. The strategies were evaluated not just in terms of the proportion or number of minorities in each sample, but of the effective sample size due to weighting for each minority.