Ranked-set sampling with regression-type estimators

Ranked set sampling (RSS) is a sampling scheme to reduce cost and increase efficiency in situations where the measurement of a survey variable is costly and/or time-consuming but ranking of sampled items relating to the survey variable can be easily done by certain other means. When a concomitant variable is readily available, the concomitant variable can be employed to aid in both sampling and estimation. Regression-type estimators making use of concomitant variables have been proposed in the literature. In this article, we study further the properties of the regression-type estimators and propose a modified RSS regression estimator which improves the available estimators. Comparison among the proposed and available estimators are made both theoretically and by simulation. Asymptotic distribution of the regression-type estimators are established and hence construction of confidence intervals and hypothesis testing based on these estimators are made possible.