|
Abstract:
|
In 2016, Westat and Census Bureau conducted concurrent versions of an on-going survey of fishing, hunting and wildlife-associated recreation (FHWAR). The surveys applied different modes and estimation approaches, and the resulting estimates were substantially different in a number of variables. In order to preserve the time series of estimates, calibration of the two surveys was therefore required. We developed a survey calibration procedure consisting of hybrid design-based and model-based adjustments, which was necessitated by the large number of variables involved. A key step in the procedure involved aligning the demographic characteristics of both samples. Sample-based raking was used to create calibrated weights for the Westat survey. Construction of associated variance estimates that reflect the uncertainty of the two surveys and the calibration was complicated by the fact that the surveys used incompatible methods of replication variance estimation. This presentation describes the method we developed to solve this problem, and provides recommendations on how to generalize our approach to other sample-based calibration settings.
|
