统计与大数据学术讨论会
(第二场)
统计与大数据理论与方法小型研讨会
演讲人: Zongwu Cai
堪萨斯大学The Charles Oswald 教授、经济学、经济计量学教授
厦门大学王亚南经济研究院教授
The Econometric Theory Multa Scripsit Award
美国统计学会Fellow
《Journal of Business and Economic Statistics》副主编
《Econometric Theory》副主编
《Econometrics》副主编
《African Finance Journal》副主编
演讲题目: A New Semiparametric Quantile Panel Data Model: Theory and Applications
摘要: In this paper, we propose a new semiparametric quantile panel data model with correlated random effects in which some of the coefficients are allowed to depend on some smooth economic variables while other coefficients remain constant. A three-stage estimation procedure is proposed to estimate both constant and functional coefficients based on the integrated quasi-likelihood approach and their asymptotic properties are investigated. We showthat the estimator of constant coefficients is root-N consistent and the estimator of varying coefficients converges in a nonparametric rate. A Monte Carlo simulation is conducted to examine the finite sample performance of the proposed estimators. Finally, the proposed semiparametric quantile panel data model is applied to estimating the impact of foreign direct investment (FDI) on economic growth using the cross-country data from 1970 to 1999.
This is a join work with Dr. Linna Chen and Dr. Ying Fang.
时间: 2015年10月30日13:30-14:30
地点: 明德主楼1030会议室
演讲人: 刘卫东
上海交通大学数学系、自然科学研究院，统计学教授
2010年全国百篇优秀博士学位论文奖
2010年新世界数学奖
2011年教育部新世纪优秀人才计划
2011、2015年上海市“东方学者”高校特聘教授（跟踪计划）
2013国家自然科学优秀青年基金
2014国家自然科学基金重点项目子课题负责人
2013上海市“浦江人才计划”
2014上海市“曙光人才计划”
《Journal of Statistical Planning and Inference》副主编
演讲题目: Two-sample multiple tests on correlation changes with uncorrelated screening
摘要: Detecting correlation changes between two multivariate normal vectors has important applications in construction of gene differential co-expression network and testing interactions between multivariate normal covariates with a binary response. Multiple testing with FDR control provides an efficient way to detect correlation changes. In this paper, we investigate the problem of two-sample multiple tests on correlation changes. The existing multiple testing procedures typically use only p-values/test statistics of all hypotheses. However, we show that, for two-sample multiple correlation tests, the information of supports of correlation matrices can lead to a significant improvement on statistical power. To explore the support information of correlation matrices, we propose screening statistics which are asymptotically uncorrelated with two sample correlation test statistics. A new multiple testing procedure, which combines the uncorrelated screening statistics and correlation test statistics, is developed to control the FDR. The method shares some oracle properties and improves the statistical power significantly, comparing to those procedures with p-values/test statistics only. Our study reveals an interesting phenomenon: for two-sample multiple testing problems, procedures using only p-values/test statistics can be not optimal on statistical power.
时间: 2015年10月30日14:30-15:30
地点: 明德主楼1030会议室
演讲人: Wei Lin
Ohio University数学系副教授（Tenured）
美国统计学会会员
《Advanced Teacher Capacity：Addressing the Ohio Core and the Common Core for Mathematics》项目负责人：2012-2015
演讲题目: Semiparametric Methods for Dimension Reduction under Linearity Conditions
摘要: Dimension reduction has been one of the most popular topics in regression analysis in the past two decades. It sees much progress with the introduction of the sliced inverse regression (SIR, Li 1991) technique and since then many inference methods have been proposed in the literature. While there are nonparametric alternatives (Xia, 2002), vast majority of these methods center around the idea of inverse regression and assume the so-called linearity condition. In the past few years, however, semiparametric methods have brought much development into the field (Ma and Zhu, 2012) without assuming the linearity condition. In this talk, we will introduce how semiparametric methods work in the field of dimension reduction, and show how the linearity condition affects the corresponding results.
时间: 2015年10月30日15:40-16:40
地点: 明德主楼1030会议室
演讲人: 冯兴东
上海财经大学统计与管理学院教授
国际数理统计学会会员
美国统计学会会员
2013上海市“浦江人才”计划
演讲题目: Copula-based quantile regression for longitudinal data
摘要: Inference and prediction in quantile regression for longitudinal data are challenging without parametric distributional assumptions. We propose a new semi-parametric approach that uses copula to account for intra-subject dependence and approximates the marginal distributions of longitudinal measurements, given covariates, through regression of quantiles. The proposed method is flexible, and it can provide not only efficient estimation of quantile regression coefficients but also prediction intervals for a new subject given the prior measurements and covariates. The properties of the proposed estimator and prediction are established theoretically, and assessed numerically through a simulation study and the analysis of a Pennsylvania nursing home data.
时间: 2015年10月30日16:40-17:40
地点: 明德主楼1030会议室