Sample Design and Estimation

Sampling statisticians at Abt Associates have extensive experience in the design of complex samples and in all aspects of estimation.  We work closely with other team members on survey data collection projects to integrate the sampling and estimation with other aspects of the project.  We pay close attention to the primary descriptive and analytic goals of the project to ensure that the design and estimation procedures will meet those objectives.

Our sampling expertise encompasses many different types of populations, including households, business establishments, schools, childcare centers, and medical providers.  In the area of estimation from samples, we provide technical knowledge and practical experience in imputation, weighting, nonresponse adjustments, coverage adjustments, and poststratification to control totals.  We also offer expertise in the area of variance estimation from complex samples.

The design of samples is a key component of survey research projects because the design serves as the basis for inference from the sample to the population.  Sampling statisticians at Abt Associates have many years of experience in all aspects of sample design: planning, use of prior studies and pilot studies to establish design components (e.g., expected design effects), determining sample size, selecting the sample, and implementing the sample.  We use a wide variety of sampling techniques to customize the design to the descriptive and analytic objectives of the project.  Key considerations in most designs include tailoring the sample selection methods to take into account cost considerations, while controlling respondent burden and reducing nonresponse to the survey.  In the area of telephone surveys, sampling statisticians at Abt Associates have extensive knowledge in the design and proper implementation of random-digit-dialing samples.

Estimation work at Abt Associates is the cornerstone for inferences from the sample to the population.  We apply imputation techniques such as hot-deck and regression imputation for item nonresponse.  We analyze the components of nonresponse experienced in a survey to develop the best possible set of unit nonresponse adjustments.  We have been a leader in the development of coverage adjustment methods, particularly to compensate for the fact that random-digit-dialing surveys do not cover households without telephones.  We also have a great deal of experience in the use of poststratification adjustment techniques such as raking.  Application of variance estimation methods for complex sample designs has included Taylor-series approximations, the jackknife, and the bootstrap.

Sampling statisticians at Abt Associates always work closely with subject matter experts to coordinate the design activity.  This collaboration usually entails defining the key units of analysis, identifying key survey outcome measures, and determining the appropriate level of precision for point estimates or the minimum detectable difference for comparisons.

Survey Clients

Clients for Abt Associates’ survey work are in both the government and business sectors.  Our government clients in the United States include virtually every federal government agency — including the Centers for Disease Control and Prevention, the Food and Nutrition Service of the Department of Agriculture, the Centers for Medicare and Medicaid Services, the National Center for Education Statistics, the National Center for Health Statistics, the National Institutes of Health, and the U.S. Bureau of the Census — and many state governments.  In our international survey research, we provide both design services and data analysis for data collection work that is generally conducted by in-country firms.  Our private clients include major money-center and regional banks, insurance companies, health industry companies, foundations, not-for-profit organizations, and professional and trade associations.

Our survey capability has been employed on a wide variety of registries, outcomes research, physician belief and practice studies, and other health-related projects.