Inferences on the Correlation Coefficient in Bivariate Normal Populations From Ranked Set Samples
将排序集抽样方法推广到双变量正态分布，研究仅基于极端Y值及其伴随X值的最大似然估计，发现当视觉排序优于随机时，估计量的渐近方差比随机样本更小。
Abstract Ranked set sampling was first used to obtain an improved estimate of a population mean. This technique is useful when a small random sample can be visually ordered easily and fairly accurately, but exact measurement of an observation is difficult or expensive. The sampling method is generalized and applied to estimation of the correlation coefficient ρ of a bivariate normal random vector (X, Y). The asymptotic variance of the maximum likelihood estimator of ρ based only on the extreme Y's and their concomitant X's is reduced over that from a random sample if the visual ordering is any better than random.