大数据下基于体积抽样的异常点诊断及估计问题

处理大规模数据集时,抽样是一种很受欢迎的有效方法。体积抽样作为一种联合抽样的方法,它是按照与矩阵平方的行列式成比例进行抽样。该方法在线性回归模型背景下能得到参数的无偏估计。然而也容易受到异常点的影响,本文感兴趣的是体积抽样受异常点影响的程度。基于数据删除模型和均值漂移模型构建统计量进行异常点诊断,结果发现体积抽样方法在某些情况下极易受异常点影响。但是在给定损失的条件下,比独立同分布抽样所需的子样本量更小,在此基础上,提出样本量的自适应选择方法。作为体积抽样的扩展,杠杆值体积抽样同样可以得到普通最小二乘线性模型参数的无偏估计,一个有趣的发现是使用杠杆值体积抽样,等权最小二乘估计结果比非等权最小二乘估计效果好。     

Logistic回归模型的统计诊断

统计诊断的主要任务就是通过诊断统计量检测已知观测数据在用既定模型拟合时的合理性,主要是找出数据当中的异常点或强影响点。本文主要研究Logostic回归模型的诊断统计量和诊断统计图。用牛顿迭代法给出Logistic回归模型的极大似然估计值,根据扰动模型得到传统的诊断统计量,结合残差、杠杆值和系数变化三者构造新的诊断统计量,绘制新的诊断统计图,通过模拟研究说明新的诊断统计量的有效性,最后用一个实际案例说明新的诊断方法的应用并进一步验证其优越性。     

非线性测量误差模型的影响分析

本文对非线性测量误差模型给出了统一的诊断方法,并证明了数据删除模型与均值漂移模型的等价性,由此出发得到了Cook距离、残差、杠杆值等诊断统计量.本文还讨论了非线性测量误差模型的局部影响分析,并给出了一个具体应用实例.推广了Zhao & Lee(1995)的结果.     

空间变系数模型的统计诊断

空间变系数模型作为一类有效的空间数据分析方法已经得到了广泛的应用.本文主要研究该模型的统计诊断与影响分析方法。首先我们基于数据删除模型定义了Cook统计量,其次我们基于均值漂移模型讨论了异常点的检验问题。     

约束线性模型异常值检验

本文讨论了带约束线性模型的数据删除模型和均值漂移模型，得到了带约束情况下上述两个模型的统计量之间的关系，建立了相应的异常值检验统计量及性质。     

线性回归诊断的若干问题

本文对于线性回归诊断提出了几种新的模型和方法。我们首次研究了方差加权和均值漂移的混合模型,得到了相应的诊断统计量。本文还引入了罚函数方法,并以此为工具,讨论了若干有偏估计的影响度量,最后,本文提出了基于重心的诊断统计量,对于识别异常点有较好的效果。     

半参数广义线性随机效应模型的影响分析

该文系统研究了半参数广义线性随机效应模型的统计诊断与影响分析方法, 证明了数据删除模型和均值漂移模型的等价性定理, 给出了广义Cook距离等诊断统计量及异常点的Score检验统计量并研究了该模型的局部影响分析,分别对加权扰动模型, 响应变量扰动模型得到了影响距阵的计算公式, 最后通过一个实例验证了文中给出诊断方法的有效性.     

单纯形分布模型的统计诊断

该文讨论了单纯形分布广义线性模型和广义非线性模型的影响分析问题, 得到了若干有用的诊断统计量;证明了数据删除模型和均值漂移模型的的等价性定理.同时还研究了该模型的变离差检验,得到了Score检验统计量.最后给出了两个实例, 说明该文方法的应用价值.     

一种改进的异常值检测算法与稳健估计

对模型精度与稳健性的要求使得异常值检测与稳健估计在模型构建中变得日益重要.本文首先利用基于边际相关系数构造的高维影响度量指标(HIM)与基于距离相关系数构造的高维数据异常值判别方法(HDC)分别对数据中的异常值进行初步检测,将数据集中的点分为正常点与异常点两类,然后在初始正常点集的基础上利用稳健的参数估计方法和残差空间超椭球等高面的概念构造了对初始正常点集中误判点的纠正方法,并对初始异常点集中各点的异常值概率重新进行计算,以进一步纠正误判入异常点集的正常点,最终对异常值检测的准确率进行进一步的提升.通过对两种数据结构下三种不同类型异常数据的模拟,证明了所提方法的有效性,并通过实例进行验证与分析.     

纵向数据下半参数联合均值协方差模型的贝叶斯分析

基于改进的Cholesky分解,研究分析了纵向数据下半参数联合均值协方差模型的贝叶斯估计和贝叶斯统计诊断,其中非参数部分采用B样条逼近.主要通过应用Gibbs抽样和Metropolis-Hastings算法相结合的混合算法获得模型中未知参数的贝叶斯估计和贝叶斯数据删除影响诊断统计量.并利用诊断统计量的大小来识别数据的异常点.模拟研究和实例分析都表明提出的贝叶斯估计和诊断方法是可行有效的.     

RECENT PROGRESS ON SPHERICAL HARMONIC APPROXIMATION MADE BY BNU RESEARCH GROUP ——In memory of Professor Sun Yongsheng

As early as in 1990, Professor Sun Yongsheng, suggested his students at Beijing Normal University to consider research problems on the unit sphere. Under his guidance and encouragement his students started the research on spherical harmonic analysis and approximation. In this paper, we incompletely introduce the main achievements in this area obtained by our group and relative researchers during recent 5 years （2001-2005）. The main topics are： convergence of Cesaro summability, a.e. and strong summability of Fourier-Laplace series; smoothness and K-functionals; Kolmogorov and linear widths.     

Boundedness for commutators of multipliers on Hardy type spaces

In this paper, we study the commutators generalized by multipliers and a BMO function. Under some assumptions, we establish its boundedness properties from certain atomic Hardy space Hb^p（R^n） into the Lebesgue space L^p with p 〈 1.     

A unified approach to certain problems of best local approximation

In this paper we study best local quasi-rational approximation and best local approximation from finite dimensional subspaces of vectorial functions of several variables. Our approach extends and unifies several problems concerning best local multi-point approximation in different norms.     

Towards Excellence in Science Chinese Academy of Sciences

<正>May 26,2014,Beijing Science is a human enterprise in the pursuit of knowledge.The scientific revolution that occurred in the 17th Century initiated the advances of modern science.The scientific knowledge system created by     

THE 8~(th) INTERNATIONAL CONGRESS ON INDUSTRIAL AND APPLIED MATHEMATICS

<正>August 10-14,2015Beijing,ChinaThe International Congress on Industrial and Applied Mathematics(ICIAM)is the premier international congress in the field of applied mathematics held every four years under the auspices of the International Council for Industrial and Applied Mathematics.From August 10 to 14,2015,mathematicians,scientists     

Bounds for the zeros of polynomials

Let P(z)=∑↓j=0↑n ajx^j be a polynomial of degree n. In this paper we prove a more general result which interalia improves upon the bounds of a class of polynomials. We also prove a result which includes some extensions and generalizations of Enestrǒm-Kakeya theorem.     

Boundedness of commutators related to Marcinkiewicz integrals on hardy type spaces

In this paper, the authors study the boundedness of the operator [μΩ, b], the commutator generated by a function b ∈ Lipβ(Rn)(0 ＜β≤ 1) and the Marcinkiewicz integrals μΩ, on the classical Hardy spaces and the Herz-type Hardy spaces in the case Ω∈ Lipα(Sn-1)(0 ＜α≤ 1).     

Generalization of the interaction between Haar approximation and polynomial operators to higher order methods

In applications it is useful to compute the local average empirical statistics on u. A very simple relation exists when of a function f（u） of an input u from the local averages are given by a Haar approximation. The question is to know if it holds for higher order approximation methods. To do so, it is necessary to use approximate product operators defined over linear approximation spaces. These products are characterized by a Strang and Fix like condition. An explicit construction of these product operators is exhibited for piecewise polynomial functions, using Hermite interpolation. The averaging relation which holds for the Haar approximation is then recovered when the product is defined by a two point Hermite interpolation.     

Jacobi polynomials used to invert the laplace transform

Given the Laplace transform F(s) of a function f(t), we develop a new algorithm to find an approximation to f(t) by the use of the classical Jacobi polynomials. The main contribution of our work is the development of a new and very effective method to determine the coefficients in the finite series expansion that approximation f(t) in terms of Jacobi polynomials. Some numerical examples are illustrated.