This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognizing you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.
Westat statisticians in the Statistics and Data Science group have published articles in the December 2023 issue of the journal, Survey Methodology. Andreea L. Erciulescu, senior statistician, and her co-authors developed a small area methodology for constrained estimation, and J. Michael Brick, Senior Vice President, and Jill M. DeMatteis, Vice President, examined new sample designs for multimode household surveys. More information on the research and links to the articles (or abstracts) follow.
Bayesian Small Area Models for Constrained Estimation
Benchmark county-level estimates of crop area totals must meet state preset totals and are subject to constraints and uncertainty. In a recent study, “Bayesian small area models under inequality constraints with benchmarking and double shrinkage,” (PDF) researchers developed a methodology to solve this problem. Westat’s Andreea L. Erciulescu, PhD, was among the co-authors of this study.
With a full Bayesian analysis of the Fay-Herriot model with benchmarking and inequality constraints, researchers studied two model approaches: 1) single-shrinkage model, assuming known variances, and 2) double-shrinkage model, assuming unknown random variances. Both come with computational challenges. However, they expect the second model should perform better in terms of goodness of fit (reliability) and possibly precision.
Alternative Survey Designs to Improve Response Rates
Survey response rates are declining, and costs are increasing, pushing researchers towards multimode data collection. In a new study, “Sample designs and estimators for multimode surveys with face-to-face data collection,” (PDF) Westat’s J. Michael Brick, PhD, and Jill M. DeMatteis, PhD, address this problem by developing two new sample designs for multimode household surveys.
The researchers propose two alternatives to the traditional design: 1) subsampling primary sampling units (PSUs) and 2) a hybrid design that combines a clustered (two-stage) sample and an unclustered sample. Using a simulation, they demonstrate that the hybrid design presents considerable advantages compared to the traditional approach.