Behavior of Statistical Estimators in Multilocation Audit Sampling.
通过模拟64个研究总体，比较了多地点两阶段抽样中几种统计估计量的表现，发现全局Stringer界在所有情况下都达到名义置信水平，但常导致误拒合格总体。
Abstract The performance of several statistical estimators in two-stage sampling was examined for accounting populations dispersed over multiple locations. The simulation study involved 64 study populations. The global Stringer bound with cell sampling of dollar units was found to achieve the nominal confidence level in all cases. This bound also tended to be tighter than the mean-per-unit bound with stratified sampling of line items, which statistically achieved the nominal confidence level for only half of the study populations. The global Stringer bound has the disadvantage of frequently leading to rejection of a satisfactory population.