Wilcoxon-Mann-Whitney Effects for Clustered Data: Informative Cluster Size

TL;DR

This paper develops a new nonparametric inferential method for the Wilcoxon-Mann-Whitney effect in clustered data, accounting for informative cluster sizes, and demonstrates its advantages through simulations and real data applications.

Contribution

It introduces an unbiased estimator and asymptotic inference procedure for the Wilcoxon-Mann-Whitney effect that considers informative cluster sizes, improving over existing methods.

Findings

Method accounts for informative cluster sizes in inference.

Proposed estimator outperforms existing methods in simulations.

Applications demonstrate practical utility in longitudinal and periodontal studies.

Abstract

In clustered data setting, informative cluster size has been a focus of recent research. In the nonparametric context, the problem has been considered mainly for testing equality of distribution functions. The aim in this paper is to develop inferential procedure for the Wilcoxon-Mann-Whintey effect (also known as the nonprametric relative effect). Unbiased estimator is provided and its asymptotic properties are investigated. The asymptotic theory is employed to develop inferential methods. While the proposed method takes information in the cluster sizes into consideration when constructing the estimator, it is equally applicable for ignorable cluster size situation. Simulation results show that our method appropriately accounts for informative cluster size and it generally outperforms existing methods, especially those designed under ignorable cluster sizes. The applications of the method is illustrated using data from a longitudinal study of alcohol use and a periodontal study.