Keywords: survey, address-based sampling, coverage, bias
Surveys have played significant roles in agricultural research, industrial quality control, marketing, politics, and social science research. Survey research has evolved over time with innovations in theory and methods and changes in technology and public attitudes about participation. The evolution is reflected in the research and methods that a survey statistician encounters. In this panel presentation, one senior statistician reflects on the trends in survey statistics experienced in her career. The panelists then discuss their recent collaborative research on address-based sampling.
Statistical literature has no standard method for estimating how well address frames cover the population of housing units, especially for subnational areas. We tested various estimates of coverage to determine the best methods for mail and in-person studies.
Undercoverage is a problem only if uncovered housing units have different attributes than covered units. We assessed how different the uncovered units would need to be to significantly change the estimates, what geographies and types of variables have higher risks of coverage bias, and whether weighting techniques correct the bias.