A0416
Title: Homogeneity pursuit in ranking inferences based on pairwise comparison data
Authors: Yuxin Tao - Southern University of Science and Technology (China) [presenting]
Tracy Ke - Harvard University (United States)
Abstract: The Bradley-Terry-Luce (BTL) model is one of the most celebrated models for ranking inferences based on pairwise comparison data, which associates individuals with latent preference scores and produces ranks. An important question that arises is the uncertainty quantification for ranks. It is natural to think that ranks for two individuals are not trustworthy if there is only a subtle difference in their preference scores. The homogeneity of scores in the BTL model is explored, which assumes that individuals cluster into groups with the same preference scores. The clustering algorithm in regression is introduced via data-driven segmentation (CARDS) penalty into the likelihood function, which can rigorously and automatically separate parameters and uncover group structure. Statistical properties of two versions of CARDS are analyzed. As a result, a faster convergence rate and sharper confidence intervals were achieved for the maximum likelihood estimation of preference scores, providing insight into the power of exploring low-dimensional structures in a high-dimensional setting. Real data examples are analyzed, including NBA basketball ranking and Netflix movie ranking, to highlight the method's improved prediction performance and interpretation ability.
