数学科学学院学术报告[2024]106号

(高水平大学建设系列报告986号)

报告题目：New Models for Analyzing Positively Correlated Multivariate Continuous Proportional Data

报告人：田国梁 教授（南方科技大学）

报告时间：2024年11月6日下午15:00—16:00

报告地点：深圳大学粤海校区汇星楼一号教室

报告内容: To analyze positively correlated continuous proportional data, we propose a general method via stochastic representation (SR) to construct a random vector with positively correlated components. This paper first proposes two new multivariate distributions (i.e., the multivariate proportional inverse Gaussian (PIG) and multivariate proportional gamma (PGA) distributions), and then develops a normalized expectation–maximization (N-EM) algorithm to calculate the maximum likelihood estimates (MLEs) of parameters of interest. Two mean regression models are studied and simulations are also conducted. We investigated cortical thickness in schizophrenia by modeling selected local brain regions with this approach for the first time. The suspected compensatory changes in patients’ cortical thickness profile were verified quantitatively. We also modelled the mean cortical thickness with covariates such as age and sex. Based on the results fitted using the new models, we propose that the compensatory changes in some brain regions may not be related to the severity of symptoms in schizophrenia but positively related to the duration of the illness. [This is a joint work with Dr. Yuan SUN, Dr. Peng-Yi LIU, Prof. Shui-Xia GUO and Prof. Lena Palaniyappan]

报告人简历：田国梁博士曾在美国马里兰大学从事医学统计研究六年, 在香港大学统计与精算学系任副教授八年, 从2016年6月至今在南方科技大学统计与数据科学系任教授、博士生导师、副系主任。他目前的研究方向为EM/MM/US算法在统计中的应用、(0, 1) 区间上连续比例数据以及多元连续比例数据的统计分析、多元零膨胀计次数据分析, 在国外发表150余篇SCI论文、出版3本英文专著、在科学出版社出版英文教材2本。他是四个国际统计期刊的副主编。主持国家自然科学基金面上项目二项、主持深圳市稳定支持面上项目一项、参加国家自然科学基金重点项目一项。

欢迎师生参加！

邀请人：周彦

　　　　　　　　　　　　　　　　　　　　　　　　　 数学科学学院

　　　　　　　　　　　　　　　　　　　　　2024年10月28日