时间：2007年5月24日晚6：30-8：30
地点：紫金港校区体育外围平台APP702会议室
报告人及题目：
Professor John Stufken:
How to Write A Professional Paper to A Referred Journal
Professor Xiangrong Yin:
Dimension Reduction in Time Series（时间序列分析中的减维方法）
主持人：陈占平教授
欢迎广大师生积极参加。
报告人简介：
Professor John Stufken：Professor and Head of Statistics University of Georgia. Former editor of J of Stat. Plann. and Inference. Research interests includes Design of Experiments, Linear Models, Survey Sampling.
Professor Xiangrong Yin: Associate Professor of University of Georgia. Graduated from Hangzhou University. Research interests includes Dimension reduction in regression analysis, Regression graphics and data visualization, Classification and discriminant analysis, Pattern recognition such as face and expression recognition, Data analysis for experimental designs, Microarray data analysis, Data mining, and variable selections, Time series data analysis, Application of multivariate methods in biology, climate, etc., Fast computing algorithms
报告内容摘要：
1. How to Write A Professional Paper to A Referred Journal
Writing of papers, books, reports, and other documents will become an essential part of the tasks of most PhD graduates who take a job in academia, industry or government. Yet there is little formal training for this in most Statistics PhD programs. Those who are lucky work with a major professor who provides many useful tips, but others are left to their own devices. In this talk I will present some thoughts on writing, in particular for papers and reports, as well as on the review process used by most professional journals.
2. Dimension Reduction in Time Series 时间序列分析中的减维方法
Weather influence Wolf Yearly Sunspot Data; seasonally adjusted quarterly growth rates of US Gross National Product (GNP) data; and U.S. beer production data are all time series related data sets. How one can build better models for such kind of data sets? Traditionally, time series analysis involves building an appropriate model and using either parametric or nonparametric methods to make inference about the model parameters. Motivated by recent developments in dimension reduction theory in regression, we develop a similar theory for time series which does not require specification of a model but seeks to find sufficient dimensions, resulting in no loss of information about the conditional distribution of the series. To this end, we define a notion of minimum dimension reduction subspace, called time series central subspace, and estimate it using a recent method based on Kullback-Leibler distance. The estimator is shown to be consistent. Better models for the aforementioned data sets are established. Through simulations, we believe that the methods presented here offer a new approach to analyzing time series data.
传统的时序分析中，通常要建立合适的模型，要么首先假设特定的时序分布，要么用非参数估算模型来估计参数。根据新的回归分析中减维理论的发展，我们创立了时序分析中新的减维理论。它不需要通过假设特定的时序分布而达到减维的目的――通过最小的维数来保证时序中的条件分布和减维前的条件分布不变。我们定义了新的时序最小空间和中心空间，利用“柯伯克―李伯来距离”来估算这些空间，证明了估算参数的收敛性。通过模型数据和实际数据，我们认为它提供了时序分析的新方法。
应用范围：数据分析中建立模型本身就是一种减维过程。减维方法和理论是非常广泛的一种统计理论。准确地说，只要有数据分析，尤其是大型的现代数据分析，减维就是必不可少的。它不仅在统计学中获得了巨大发展，不断完善，而且在其他领域中也有极其广泛的应用。比如在计算机领域中的图像处理（Pattern recognition），区别与分类（Discriminant and classification analysis）；生物中的基因分析（Microarray data analysis）；医学中的相关分析（Survival data analysis）；商业中的市场分析（Marketing data analysis）；经济中的数据分析（Econometrics），以及环境科学中的数据分析（Environmental data analysis）等等。
我们相信越来越多地科学交叉，将会更加推动减维理论和方法的更新和完善。
管理科学与工程学系
