L-Estimation of Population Quantiles Using Ranked Set Sampling

TL;DR

This paper introduces two new RSS-based L-estimators for population quantiles, improving efficiency over traditional methods, especially in biomedical applications, by leveraging ranked set sampling techniques.

Contribution

It extends Stigler-type and Harrell--Davis estimators to the RSS framework, providing scalable and efficient quantile estimation methods with theoretical and practical validation.

Findings

RSS estimators outperform empirical quantiles under various distributions.

Harrell--Davis RSS estimator performs especially well for moderate and upper quantiles.

Simulation and real data demonstrate practical relevance in biomedical settings.

Abstract

Quantile estimation is central when interest lies in thresholds or tail behavior rather than the mean. When exact measurement is costly but units can be ranked cheaply, ranked set sampling (RSS) provides an attractive alternative to simple random sampling (SRS). We develop two families of RSS-based L-estimators for population quantiles that extend Stigler-type and Harrell--Davis estimators to the RSS framework. The first applies weighted-order-statistic estimation directly to the pooled ordered RSS sample and serves primarily as an exact conceptual benchmark, since its computational burden increases rapidly with the set size. The second exploits a decomposition induced by the RSS design that constructs $k$ pooled transformed-scale component estimators indexed by rank stratum and leads to a computationally scalable procedure. We derive large-sample results for these component estimators under regularity conditions; these results provide a principled first-order motivation for the combined estimators employed in practice. Simulation results across several distributions, quantile levels, and ranking qualities show consistent efficiency gains over empirical quantile estimators under both SRS and RSS, with the RSS Harrell--Davis version performing especially well for moderate and upper quantiles. Beyond the simulation study, we demonstrate the practical relevance of the proposed estimators through an application to NHANES transient elastography data, highlighting their usefulness for estimating clinically meaningful quantiles in a biomedical setting