Using a model-aided sampling paradigm instead of a traditional sampling paradigm in a nationally representative establishment survey

By Marcus Berzofsky, Brandon Welch, Rick Williams, Paul Biemer

We compare traditional survey inference, which is based on probability sample selection and weighting, with a model-based approach based on sampling quotas and model-based weighting. Compared with the traditional approach, the model-based approach more efficiently controls subgroup sample sizes when a large number of rare subgroups are studied. Using data from a national survey of US businesses, we simulated a model-based paradigm and compared estimates with those under the traditional paradigm. In this study, the findings suggest that the model-based approach offers advantages over the traditional sampling approach; however, a hybrid approach capturing the advantages of both paradigms proved best.


Berzofsky, M., Welch, B., Williams, R., & Biemer, P. (2008). Using a model-aided sampling paradigm instead of a traditional sampling paradigm in a nationally representative establishment survey. (RTI Press Publication No. MR-0004-0802). Research Triangle Park, NC: RTI Press.

© 2019 RTI International. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.


Marcus BerzofskyMarcus Berzofsky, MS, is a statistician at RTI International in Research Triangle Park, NC.

Brandon WelchBrandon Welch, MS, was formerly a statistician at RTI; he is now at Rho.

Rick WilliamsRick L. Williams, PhD, has more than 30 years of experience in public health research as a statistician and is a fellow of the American Statistical Association. He designs population surveys, observational studies, and randomized clinical trials, and develops and applies analysis methods for correlated and sample survey data. Dr. Williams has extensive experience with health studies of women, infants, and children. He has experience in preclinical and early-phase drug development studies, such as population pharmacokinetic studies, and in late-phase pharmaceutical outcomes research. He frequently conducts workshops and training classes on the use of appropriate statistical analysis methods for cluster-correlated, longitudinal, or repeated measures data, such as generalized estimating equations for marginal models, hierarchical linear and nonlinear mixed models, and multilevel models. He was selected by a peer-review committee to serve as a fellow for one year with U.S. Department of Agriculture.

Paul BiemerPaul Biemer, PhD, is a Distinguished Fellow in Statistics at RTI International in Research Triangle Park, NC. Dr. Biemer has more than 30 years of postdoctoral experience in survey methods and statistics. He joined RTI in 1991, serving as director of the survey methods program until 1994 and of the Center for Survey Methods and Research from 1994 to 2000. Dr. Biemer's scientific contributions to survey methodology and statistics include developing methodologies for using computer audio-recorded interviewing, using latent class analysis as a survey error evaluation tool, and applying continuous quality improvement to the coding of industry and occupation question responses. He holds a joint appointment with the Odum Institute for Research in Social Sciences at the University of North Carolina at Chapel Hill where he is associate director for survey research and director of the certificate program in survey methodology. He has written five books, 35 peer-reviewed publications, 17 book chapters, and numerous papers and presentations.

Contact RTI Press

To contact an author, request an exam or review copy, or seek permission to use copyrighted content, contact our editorial team.