[영문]Calibration makes the generalized regression (GREG) estimator under Poisson sampling practical when there are a number of items of interest. The question becomes how to estimate mean squared errors in an equally practical manner. When all the selection probabilities are small, and the GREG is expressible in projection form, an appropriately defined delete-a-group jackknife variance estimator can have desirable asymptotic properties making it a useful measure in many applications