旧站回顾 加入收藏 English
当前位置: 首页 >> 通知公告 >> 正文 通知公告
中央财经大学统计与数学学院双周学术报告会(三十五)
来源:  点击次数: 次 发布时间:2019-07-05   编辑:统计数学学院

 

时间:20190710(星期14:00-15:00

地点:沙河校区,主教楼 207

报告题目:Estimation and Inference of AHeteroskedasticity Model with Latent Semiparametric Factors for Panel DataAnalysis

报告人:周文,美国Colorado State University 统计系助理教授。

报告摘要:We considerestimation and inference of a flexible subject-specific heteroskedasticitymodel for analyzing large scale panel data, which employs latent semiparametricfactor structure to simultaneously account for the heteroskedasticity acrosssubjects and contemporaneous correlations. Specifically, the heteroskedasticityacross subjects is modeled by the product of unobserved stationary process offactors and subject-specific covariate effect. Serving as the loading, thecovariate effect is further modeled through the additive model. We propose atwo-step procedure for estimation. First, the latent factor process andnonparametric loading are estimated via projection-based methods. Theestimation of regression coefficients is further conducted through thegeneralized least squares type approach. Theoretical validity of the two-stepprocedure is carefully documented. By scrupulously examining the non-asymptoticrates for recovering the latent factor process and its loading, we furtherstudy the properties of the estimated regression coefficients. In particular,we establish the asymptotic normality of the proposed two-step estimate ofregression coefficients. The proposed regression coefficient estimator is alsoshown to be asymptotically efficient. This leads to a more efficient confidenceset of the regression coefficients. Using a comprehensive simulation study, wedemonstrate the finite sample performance of the proposed procedure, andnumerical results corroborate our theoretical findings. Finally, we apply ourproposed method to a data set of air quality and energy consumption collectedat 129 monitoring sites in the United States in 2015.

告人简介:周文,美国Colorado State University 统计系助理教授。美国Iowa State University的统计博士和应用数学博士。他的主要研究方向是High dimensional data inference, graphical modelingStatistical machine learningStatistical genomics and bioinformatics, system biology;Optimization and game theory。周博士的论文发表于Biometrika, Biometrics, Plos One, Journal of MultivariateAnalysis,Nature Biotechnology, Journal of Biological Dynamics, MathematicalBiosciences and Engineering, Vaccine等等。

 

本次活动受中央财经大学2019专题学术讲座项目资助。

首页

          版权所有:中央财经大学统计与数学学院  
          地址:北京市海淀区学院南路39号图配楼5层 邮政编码:100081 电 话:(010)62288522