报告题目：Nonlinear Spatio-TemporalData Analysis: A personalreview and outlook in the digital age
报告人：卢祖帝教授(香港城市大学)
报告时间：2025年12月14日(18:00-19:00)
报告地点：腾讯会议号:764-662-495
报告人简介：卢祖帝教授观任香港城市大学生物统计系教授，曾任英国南普顿大学统计学讲席教授(2013-2024)，并于澳大利亚阿德菜德大学、科廷大学、伦敦政治经济学院、中国科学院等国内外知名高校与研究机构任职;1996年获中国科学院博士学位。他长期从事统计推断与计算、非线性时空大数据建模、金融计里及医学统计等领的研究，以第一、通讯作者发表多篇JASA、JRSSB、Ann.Statist和 Biometrics 等权威刊物。卢教授曾茶奥人利亚研究理事会未来学者奖、欧盟马丽·居里学者集成基金奖，并当选国际统计学会会员，亦曾获中国料学院院长父等多项荣誉。
报告摘要：Nonlinear dynamic modelling of spatio-temporal data is often a challenge, especiallydue to irregularly observed locations and location-wide non-stationarity, This talk is based onseveral of my recent publications, focusing on a paper recently published, We propose as semiparametric family of Dynamic Functional-coefficient Autoregressive Spatio-Temporal(DyFAST) models to address the difficulties, We specify the autoregressive smoothingcoefficients depending dynamically on both a concerned regime and location so that themodels can characterise not only the dynamic regime-switching nature but also the locationwide non-stationarity in real data, Different smoothing schemes are then proposed to modelthe dynamic neighbouring-time interaction effects with irregular locations incorporated bi(spatial) weight matrices. The first scheme popular in econometrics supposes that the weightmatrix is pre-specified, We show that a greedy idea for locally optimal modelling popular inmachine learning should be cautiously applied, Moreover, many weight matrices can begenerated differently by data location features, Model selection is popular but may sufferfrom loss of different candidate features, Our second scheme is thus to suggest a weightmatrix fusion to let data combine or select the candidates with estimation done simultaneously both theoretical properties and Monte Carlo simulations are investigated, The empiricalapplication to an EU energy market dataset further demonstrates the usefulness of ourDyFAST models, An outlook on spatial time series data analysis in the digital age will bediscussed.
