|
Abstract:
|
Address-Based Sampling frames contain auxiliary variables from the U.S. Postal Service with characteristics of addresses, which can be used to restrict the frame. ABS frames can be enhanced with many other auxiliary variables. Geocodes assign addresses to specific geographic areas. Then Census Bureau demographic variables at area levels can be appended for stratification and disproportionate sampling. Household and person-level marketing variables can reduce screening costs or influence planned contact attempts. Auxiliary variables may also be useful for weighting, imputation, or estimation.
Two quality factors affect the usefulness of an auxiliary variable. First is its completeness. Marketing variables, for example, are not available for all addresses. The second factor is accuracy. Area variables are not accurate for all households in an area, and household or person variables may be incorrect. In this paper, completeness of some variables is evaluated using an Enhanced ABS Frame. Accuracy is more difficult; here we compare to survey responses or to area totals from other reliable sources. The completeness and accuracy inform the appropriate uses of an auxiliary variable.
|
