Adjusting for Misclassification: A Three-Phase Sampling Approach

Abstract The United States Department of Agriculture’s National Agricultural Statistics Service (NASS) conducts the June Agricultural Survey (JAS) annually. Substantial misclassification occurs during the prescreening process and from field-estimating farm status for nonresponse and inaccessible records, resulting in a biased estimate of the number of US farms from the JAS. Here, the Annual Land Utilization Survey (ALUS) is proposed as a follow-on survey to the JAS to adjust the estimates of the number of US farms and other important variables. A three-phase survey design-based estimator is developed for the JAS-ALUS with nonresponse adjustment for the second phase (ALUS). A design-unbiased estimator of the variance is provided in explicit form.