Domain estimators calibrated on information from another survey

Abstract

We examine calibration estimation in a setting where two surveys are conducted on the same finite population. Some variables of study are common to the two surveys, but the second one requires greater detail in the statistics produced than the already published first one. More specifically, we require estimates for sub-populations, called domains, that are identified only in the second survey to add up consistently to known or estimated totals published for the common variables in the first survey. We outline and study several options for deriving calibration estimators for the domains identified in the second survey. We obtain explicit expressions showing how the calibration weights are related in the different approaches. The concluding section presents the results of a simulation study, comparing the precisions attained in the different options.