格致经济论坛第109期

发布者:经济学系     时间:2023-07-03     阅读次数:465

报告题目:在连续时间分析中使用函数数据分析方法的拟合和预测

报告人:田人方(西安大略大学国王大学学院)

报告时间:2023年7月3日(星期一)上午10:30—11:30

报告地点:bat365中文官方网站大楼318会议室

邀请部门:经济学系


报告人简介:

田人方,加拿大滑铁卢大学经济学博士,现任西安大略大学国王大学学院MEM经济学助理教授,研究方向为计量经济学和函数数据分析(functional data analysis)。近期开始研究计量经济学和统计学方法在健康学习中的推广和应用。


报告摘要:

Parametric continuous-time analysis for stochastic processes often entails the generalization of a predefined discrete formulation to a continuous-time limit. However, unknown convergence rates of the frequency-dependent parameters can destabilize the continuous-time generalization and cause modeling discrepancy, which in turn leads to unreliable estimation and forecast. To circumvent this discrepancy, we propose a simple solution for fitting and forecasting in continuous-time analysis based on functional data analysis and truncated Taylor series expansions. It is demonstrated through a simulation that our proposed method is superior in both fitting and forecasting continuous-time stochastic processes compared with traditional parametric methods that encounter troubles uncovering the true underlying processes.


×请先登录

账  号

密  码