Statistics Colloquium : Dr. Jesse Frey
Villanova University
Title: Testing Equivalence with Perfect Rankings in Ranked-Set Sampling
Abstract: In ranked-set sampling (RSS), how well a statistical inference procedure performs often varies depending on how good the rankings are. We propose a new population parameter that measures how good the rankings are, and we show how to obtain confidence intervals for this parameter using RSS data. The confidence intervals can be used to test perfect rankings or, more helpfully, to test equivalence with perfect rankings. Testing equivalence with perfect rankings allows one to obtain positive evidence that the rankings are sufficiently close to perfect that perfect rankings can be safely assumed. We illustrate the testing method using data sets and inference procedures from the literature.
Authors: Jesse Frey and Yimin Zhang, Villanova University