Statistical Resources

Our statistical staff offers unparalleled expertise in constructing sophisticated, cost-effective samples and maintaining their integrity. Survey samples at Mathematica are designed to meet your research objectives, provide high-quality survey estimates, and minimize costs. Our staff is highly experienced with area probability samples; random-digit-dial samples of households; and list frame samples of establishments, professions, and program participants. Our designs incorporate stratification, clustering, and multiple stages or multiple phases of selection, with oversampling of policy-relevant subgroups and use of multiple frames when appropriate.

Our staff includes statisticians with many years of experience who offer:

• Sample design and survey estimation
• Optimal sample size allocations given competing objectives
• Classical experimental design methods, hypothesis testing, and statistical modeling
• Biostatistics and various other specialty areas

These individuals also specialize in power analysis, complex sample survey design, design of complex experiments within surveys, propensity score modeling, and other techniques. Our staff provides the full spectrum of statistical services, including random assignment, calculation of weighted and unweighted response rates, replication and management of sample releases, and clearly written technical documentation.

Innovation is an essential part of our work. We have developed sample designs with allocations that use complex mathematical modeling to minimize costs and satisfy multiple precision objectives. Our detailed weighting techniques account for the probabilities of selection from the simplest to the most sophisticated designs. To adjust for nonresponse, we use the latest and most efficient techniques appropriate for the design of the study, including weighting class methods, response propensity modeling, raking and other post-stratification techniques, and trimming of outlier weights.

We also provide parameters for variance estimation that accurately account for sample design complexities in surveys using specialized software. Our statistical services staff has experience in methods of variance estimation based on linearization, replication/sampling, generalized variance functions, and other methods of variance estimation for complex survey data. We are also experienced in statistical programming, especially in SAS and SUDAAN, imputation techniques, simulation, and data quality analysis. To support client needs, we also conduct workshops and training.

Some examples of our work include the following:

• The National Beneficiary Survey for our Ticket to Work study includes a large-scale survey involving frame development, sample design, weighting, variance estimation, and imputation.
• Our survey of SSI children and their families involved designing a complex experiment embedded within the survey to test response incentives.
• For our SESTAT project, we use statistical techniques to integrate data collected through three national sample surveys fielded every two or three years—the National Survey of College Graduates, the Survey of Doctorate Recipients, and the National Survey of Recent College Graduates.
• Our Community Tracking Survey of 60 communities across the nation involved frame development, sample design, weighting, and variance estimation.

Our staff present their work at professional meetings, and many also hold leadership positions in the statistical community as officers of professional organizations. We also serve as national experts on issues such as establishment survey samples and the impact of technology on survey research.

Back to Top