generalRSS: Sampling and Inference for Balanced and Unbalanced Ranked Set Sampling in R
By: Chul Moon, Soohyun Ahn
Potential Business Impact:
Helps doctors pick the best patients for studies.
Ranked set sampling (RSS) is a stratified sampling method that improves efficiency over simple random sampling (SRS) by utilizing auxiliary information for ranking and stratification. While balanced RSS (BRSS) assumes equal allocation across strata, unbalanced RSS (URSS) allows unequal allocation, making it particularly effective for skewed distributions. The generalRSS package provides extensive tools for both BRSS and URSS, addressing limitations in existing RSS software that primarily focus on balanced designs. It supports RSS data generation, efficient sample allocation strategies for URSS, and statistical inference for both balanced and unbalanced designs. This paper presents the RSS methodology and demonstrates the utility of generalRSS through two medical data applications: a one-sample mean inference and a two-sample area under the curve (AUC) comparison using NHANES datasets. These applications illustrate the practical implementation of URSS and show how generalRSS facilitates ranked set sampling and inference in real-world data analysis.
Similar Papers
Ranked Set Sampling-Based Multilayer Perceptron: Improving Generalization via Variance-Based Bounds
Machine Learning (CS)
Makes computer learning more accurate by organizing data.
Relation-Stratified Sampling for Shapley Values Estimation in Relational Databases
Databases
Helps databases show who helped get answers.
Scalable and Efficient Multiple Imputation for Case-Cohort Studies via Influence Function-Based Supersampling
Methodology
Makes expensive health tests faster and cheaper.