基于样本均值的最优无重叠k-序对排序集抽样

本文提出了无重叠κ-序对排序集抽样方法,即在每个排序集中对κ-序对个体进行观测,并且不同的排序集的κ-序对之间没有任何重复.我们首先探究了此抽样方法得到的样本均值的有效性随每个排序集中κ-序对个体间的相关性变化的趋势.κ-序对个体间的相关性越强,样本均值的有效性损失越大.本文的目的是找到无重叠κ-序对排序集抽样方法中κ-序对分配的最优方案从而使样本均值的有效性损失最小,并证明了最优的无重叠κ-序对排序集抽样比广义排序集抽样以及简单随机抽样更有效.尽管无重叠κ-序对排序集抽样方法的统计效率低于经典的排序集抽样,但是在成本模型下,最优的无重叠κ-序对排序集抽样方法可以比经典的排序集抽样更有效.

Optimal Allocation for the Mean of the No Overlap k-tuple Ranked Set Samples

We propose an alternative ranked set sampling scheme-no overlap k-tuple ranked set sample, in which k units are quantified in each set and the units in different sets have sample mean and show that the correlations between the k-units and show that the correlations between the k-units in each set decreases the efficiency of the sample mean. The stronger the correlation structure is, the more the efficiency of the sample mean decreases. The purpose of this work is to find the optimal k-tuple ranked set sample and prove that the no overlap k-tuple ranked set sample with the optimal allocation is more efficient than generalized k-tuple ranked set sample and simple sample and simple random sample. Although the statistical efficiency of no overlap k-tuple ranked set sample is not as much as the classic ranked set sample, under the consider of the cost for sampling and ranking, the no overlap k-tuple ranked set sample is more efficient than the classic ranked set sample.