Simplified Unbiased Variance Estimation for Multistage Designs
该文证明，每层抽取两个一阶段单元的不等概率选择方法均可使用分组工具，从而在多阶段抽样中得到简化的无偏方差估计量，并避免高概率计算一阶段内成分。
All unequal probability selection methods of two first-stage units per stratum are shown to permit the use of a grouping device, leading to simplified unbiased variance estimators in multistage sampling. These variance estimators avoid the calculation of within first-stage components with a high probability. Several desirable properties of the grouping method are established. Empirical results on efficiencies of ‘grouping’ estimators and stabilities of corresponding variance estimators for unistage sampling are also presented, employing real populations.