理学院学术报告系列
发布时间: 2018-10-12  浏览次数: 10

报告题目:    A  Bayesian analysis of the multinomialprobit model under symmetric identification

报告人:         潘茂林   助理研究员

报告人单位:   南京大学数学系  

报告日期:     2018年10月15日星期一

报告时间:     上午10:00-11:30

报告地点:     江宁校区励学楼B 219

邀请人:        杨永富

欢迎广大师生参加。    

摘要: Previous studies on Bayesian multinomial probit model mainly select one of the alternatives as a base category to identify the model. Some recent studies showed that the posterior predictions are sensitive to the choice of the alternative acting as the base category. We give a symmetric identification method to fully identify the model and develop a Bayesian algorithm to analyze the model with a symmetric prior. It is shown that the predictions generated by this new method are independent of the labeling of alternatives.                                   

    报告人简介: 潘茂林2013年7月毕业于香港中文大学统计系,获博士学位。2013年9月至今在南京大学数学系工作。主要从事统计计算,计量经济方面的研究工作。发表SCI论文3篇,分别发表在Insurance: Mathematics and Economics, Discrete Dynamic in Nature and Society, Communications in Statistics:Theory and Methods.