Bayesian linear models for cardinal paired comparison data
本文开发了在参数受限于线性子空间时进行贝叶斯更新的方法，应用于配对比较数据中能力参数和排名的后验分布计算，并用模拟数据和NBA球队排名等实例展示。
This paper develops a methodology for Bayesian updating in normal linear models in situations where the parameter of interest is restricted to a linear subspace. The methodology is motivated by and applied to the calculation of posterior distributions for the merit parameters and ranks arising in paired comparison data. The Bayesian paradigm is found to be ideal for assessing and quantifying the uncertainty in ranking procedures. The methodology is illustrated using simulated data and applied to two data sets: a network meta–analysis example and to the ranking of teams in the National Basketball Association (NBA).