|
Abstract:
|
The USPS Computerized Delivery Sequence file (CDS) is often used as the foundation from which to draw an address-based sample for field surveys. Researchers often use a hybrid frame because the CDS does not provide complete coverage of all US housing units, introducing the risk of coverage bias. The CDS is used as the sampling frame in areas where coverage is estimated to be sufficient, and field enumeration is used in segments where coverage is low. However, the coverage rate may not be the best indicator of the presence of coverage bias. Coverage bias is a function of both the coverage rate and the difference between uncovered and covered housing units. For example, a low coverage segment could have no bias if the uncovered units are similar to the covered units. We develop and evaluate an alternative method, the Coverage Bias Risk Index, to identify segments with the highest risk of coverage bias. This model-based method compares the estimated coverage of multiple subgroups to known controls and combines the differences into one metric. Additionally, we add a weighting step to correct for coverage bias at the sampling stage rather than only at the post-survey adjustment stage.
|
