ABSTRACT

The presence of extreme values in a data set reduces the efficiency of variance estimators. L-moments are based on the ordered form of a random variable to estimate the variance of the population. The two variance estimators are used for calibration to a stratified random sampling design and relying on an auxiliary variable. The proposed estimators use the properties of L-moments, such as the L-mean, also called L-location, the L-standard deviation, also called L-scaling, and the L-coefficient of variation, which is a measure of variation. The use of these properties allows for providing better estimators. A simulation proves the better efficiency of these estimators.

摘要

数据集中存在极值会降低方差估计器的效率。L-矩基于随机变量的有序形式来估计总体的方差。这两个方差估计量用于校准分层随机抽样设计并依赖于辅助变量。所提出的估计器使用 L 矩的属性，例如 L 均值（也称为 L 位置）、L 标准差（也称为 L 缩放）和 L 变异系数，它是变异的度量. 使用这些属性可以提供更好的估计器。模拟证明了这些估计器的效率更高。