On Unequal Probability Sampling: Sufficient Conditions for the Superiority of Sampling Without Replacement
比较了Horvitz-Thompson估计量与Hansen-Hurwitz估计量的方差，推导出无放回抽样方差更小的充分条件，并给出实例验证。
A challenging theoretical problem in unequal probability sampling is the question which design, sampling with or without replacement, should be implemented. In this context we compare the Horvitz-Thompson estimator with the Hansen-Hurwitz estimator and derive conditions under which the variance of the Horvitz-Thompson estimator is smaller than the variance of the Hansen-Hurwitz estimator. Examples are given showing the validity of these conditions.