报告人：王海斌

工作单位: 厦门大学

报告时间： 2018年11月7日 8:30-9:30

报告地点： 数学与统计学院一楼报告厅

报告内容摘要：The rapid development of the Internet and crowdsourcing platforms has greatly facilitated the collection of preference or ranking data. Many studies now involve assessment of a long list of objects based on the partial or top rankings provided by individuals on selected subsets of the entire set of objects. One typical example in higher education is the use of peer assessment in which each student in a class is asked to evaluate his or her peers' work. However, the lack of expertise and the presence of biases result in significant noise in these assessments, and the number of objects that one is willing or able to evaluate is usually relatively small when compared with the long list of objects. To aggregate a reliable global ranking for the entire set of objects from these partial rankings in which information may be weak, we propose a latent asymmetric Laplace model under the latent utility framework. To make the proposed model more general and to enhance its applicability, we further treat the quantile in the asymmetric Laplace distribution as an unknown parameter and estimate it from the data. The simulation results provide evidence that even if the latent random utility deviates from the asymmetric Laplace distribution, the proposed method can efficiently recover the true ranking. We also illustrate the proposed method with a real data set based on peer assessment.

报告人简介：王海斌，厦门大学数学科学学院教授、博士生导师，中国现场统计研究会理事、中国现场统计研究会高维数据统计分会理事。主要从事潜在变量模型、非/半参数统计模型及时间序列分析的研究工作，多次应邀赴香港中文大学统计系进行合作研究。已在国内外数学、概率、统计、计量心理学等学术期刊上发表学术论文30余篇。目前正在主持一项国家自然科学基金（面上项目）、参与一项国家自然科学基金（地区项目）。