Using Calibration to Fit Nonresponse and Undercoverage Models

• Activity Number: 478
• Type: Topic Contributed
• Date/Time: Thursday, August 11, 2005 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Theodore Chang*+ and Phillip S. Kott
• Companies: University of Virginia and National Agricultural Statistics Service
• Keywords: Benchmarks ; Weight adjustments ; Explanatory variables ; Quasi-random model

Abstract

Calibration can be used to correct for sample nonresponse and frame undercoverage and assure that weighted estimates of the calibration variables match known or alternatively estimated population totals, called "benchmarks." The quasi-randomization theory supporting this use treats response or coverage as an additional phase of random sampling (one that takes place before the sample is drawn in the case of undercoverage or after in the case of nonresponse). he functional form of a quasi-random response or coverage model is assumed to be known, while its parameter values are estimated implicitly through calibration. Unfortunately, the variables in a reasonable quasi-random model are not necessarily the same as the calibration variables for which benchmark totals are available. Moreover, it often is prudent to keep the number of explanatory variables in a model small. We will address using calibration to adjust for nonresponse or undercoverage when the number of calibration variables exceeds the total of explanatory model variables. Data from National Agricultural Statistical Service' s area sample will be used.