Using Calibration Weighting to Adjust for Nonresponse under a Plausible Model (with Full Appendices)

摘要

Calibration forces the weighted estimates of certain variables to match known or alternatively estimated population totals called benchmarks. It can be used to correct for sample-survey nonresponse or for coverage error resulting from frame undercoverage or unit duplication. The quasi-randomization theory supporting its use in nonresponse adjustment treats response as an additional phase of random sampling. The functional form of a quasi-random response model is assumed to be known, its parameter values estimated implicitly through the creation of calibration weights. Unfortunately, calibration depends upon known benchmark totals while the variables in a plausible model for survey response are not necessarily the same as the benchmark variables. Moreover, it may be prudent to keep the number of explanatory variables in a response model small. We will address using calibration to adjust for nonresponse when the explanatory model variables and benchmark variables are allowed to differ as long as the number of benchmark variables is at least as great as the number of model variables. Data from National Agricultural Statistical Services 2002 Census of Agriculture and simulations based upon that data will be used to illustrate alternative adjustments for nonresponse. The paper concludes with some remarks about extension of the methodology to adjustment for coverage error.