Efficient Estimation of the Means of Subpopulations by Inverse Sampling

For estimating the average of a characteristic of the units in some subpopulations of a finite population it is under certain conditions more efficient to sample from the entire population rather than from the subpopulation. Such conditions are investigated. The method involves quotas for the subpopulations. In contrast to traditional optimization investigations, which are in terms of expected total costs if these are random, we use a very high quantile of the cost.