Building a Better Delete-a-Group Jackknife for a Calibration Estimator (Like That Based on Data from the ARMS III)

This note summarizes much of the theory behind the use of the Delete-a- Group (DAG) jackknife with calibrated survey data like that coming from the third phase of the Agricultural Resources and Management Survey (ARMS III). A DAG jackknife employing 15 sets of replicate weights is used for variance estimation with ARMS-III data. Many analyses of this data are conducted by the Economics Research Service and by other agricultural economists outside of the US Department of Agriculture, making a portable variance-estimation technique like the DAG jackknife especially desirable. Some suggestions are offered to improve the present methodology.