线性回归模型Liu估计的新研究

考虑线性回归模型Y=Xβ+ε,E()ε=0,Cov()ε=2σI(1),当设计矩阵X的列存在共线性时,最小二乘估计^β=(X′X)-1X′Y的性质变坏,为此给出了有偏估计^(βK,d)=(X′X+K)-1(X′Y+d^β),其中K为对角矩阵,K=diag(k1,…kp),ki≥0,d>0为参数,讨论了这种有偏估计与广义岭估计、Liu估计的比较,并证明了其可容许性估计.     

非主成份与广义岭型估计

针对线性回归模型Y=Xp e,e～(0,σ2I)在设计矩阵X呈病态(存在复共线性关系)时,从主成分估计的思想出发,结合岭估计减少均方误差的方法,提出并推导了一类新的估计β(k)=(X'X Φx2kΦ'2)-1X'Y,称之为广义岭型估计.优点是只对主成分和非主成分添加两个不同的常数,均方误差大幅度降低的同时,相对于一般的广义岭估计,计算量减少,相对于主成分估计,便于对原变量做出解释.文中进一步讨论了该估计与主成分估计和岭估计的优劣.     

线性模型中误差方差的非齐次估计的可容许性

考虑模型Y=Xβ+e,其中X_(n×p)是设计矩阵,e的各分量e_1,e_2,…,e_n相互独立,E(e_i)=E(e_i~3)=0,E(e_i~2)=σ~2,E(e_i~4)=3σ~4,i=1,2,…,n。本文讨论误差方差σ~2的估计在估计类={Y′AY+l′AY+f;Y′AY+l′AY +f≥0对一切Y}中的可容许性问题。当X为满秩矩阵时,给出了σ~2的估计在中可容许的充分必要条件,当X_(n×p)=1时,给出了估计类的一个完全类以及估计可容许的充分条件。     

强收敛意义下岭回归的 C_L 和 GCV 准则的渐近最优性

设有回归模型Y_i=μ_i+e_i,i=1,2,…,n (1)假定 e_1,…,e_n 为 iid.的正态随机变量序列,具有共同的均值0和方差σ~2.每个 Y_i 可通过设计点列 x_(i1),x_(i2),…,x_i_p_n 观察到.为估计 Y=(Y_1,…,Y_n)′的未知均值 μ=(μ_1,…,μ_n)′,可构造一族岭估计(?)(h)=X(X′X+hI)~-1X′Y,h≥0,(2)其中 X=(x_ij)_(n×ρn) 为设计阵,I 为 p_n 阶单位阵.在这里,岭参数 h 的选择是一个十分     

方差分量模型中回归系数的线性Minimax估计

考虑方差分量(混合线性)模型y=Xβ+U1ξ1+U2ξ2+…+Ukξk,这里Xn×p,Ui,n×ti为已知设计矩阵,βp×1是固定效应,iξ是ti×1随机效应向量,满足E(iξ)=0,cov(iξ)=σ2iIti,iξ都不相关.往往Uk=In,ξk=ek,即最后一项为随机误差,热β∈RP和i2σ>0(i=1,2,…,k)为未知参数.我们考虑β的可估函数Sβ,选取二次损失函数L(d,Sβ)=(d-Sβ)′(d-Sβ)∑ki=1ciσi2+β′X′Vk-1Xβ,然后在线性估计类中给出Sβ的惟一的mini max估计.     

相依回归系统回归系数的待定系数估计

对于两个相依线性回归方程组成的系统(1．1)，本文提出了β1的待定系数估计β^*1（k，c)=（x′1x1 k1)^-1(x′1y1-cσ12/σ22x′1N2y2)，其中岭参数k≥0．c是待定系数．与β^*1(k，c)对应的非限定两步估计记为β^41(T，k，c)．当c=1时β^*1(k,1)=β1(k)和β^*1(T,k,1)=β1(T,k)等干[6]引入的一双有偏估计，结果表明总可以选取适当的c值和k值使β^*1(k，c)和β^*1(T,k，c)在均方误差阵准则下分别优于β1和β1(T)，并讨论了c值的最佳选择问题．     

带有不完全椭球约束的增长曲线模型中的可容许性

对于带有不完全椭球约束的增长曲线模型Y=X1ΘX2+ε,ε～(0,σ2VI),X′2(Θ-Θ0)′X′1(Θ-Θ0)X2σ2Iq本文在矩阵损失函数(d-KΘL)′(d-KΘL)下给出了KΘL在齐次线性估计类和非齐次线性估计类中可容许的充分条件.     

半序约束下多维正态总体均值和协方差阵的最大似然估计

对给定K个P维正态总体,未知均值和协方差阵分别为θi和Λi,i=1,2,…,k,本文考虑均值和协方差阵之间都在一个简单半序约束θ1≤θ2≤…≤θk,Λ1≥Λ2≥…≥Λk>0条件下的估计问题.讨论θi和Λi的最大似然估计的性质,并给出一个求解的迭代方法.     

任意阶Laplace算子的特征值估计

设D为n维Euclid空间Rn的一个有界区域,且0＜λ1≤λ2≤…≤λk≤…是l阶Laplace算子的Dirichlet问题{(-△)lu=λu, 在D中,u=(e)u/(e)n=…=(e)l-1u/(e)nl-1=0,在(e)D上的特征值.得到了该问题用其前k个特征值来估计第(k+1)个特征值λk+1的不等式k∑i=1(λk+1-λi)≤1/n(4l(n+2l-2)]1/2{k∑i=1(λk+1-λi)1/2λil-1/lk∑i=1(λk+1-λi)1/2λi1/l}1/2,此不等式不依赖于区域D.对l≥3,上述不等式比所有已知的结果都要好.陈庆民与杨洪苍考虑了l=2的情形.我们的结果是他们结果的自然推广.当l=1时,我们的不等式蕴含杨洪苍不等式的弱形式.文中还给出了陈和杨的一个断言的直接证明.     

可加模型低维分量估计平均偏差的指数界

设(X,Z,y),(X1,Z1,Y1),…,(Xn,Zn,Yn)为取值于Rp×Rq×R中的I.I.d.随机向量,E|Y|<∞,Y关于(X,Z)的回归函数m(x,z)△E(Y|(X,Z)＝(x,z))具有可加结构:m(x,z)＝m1(x)+m2(z).为估计可加分量,采用Linton&Nielsen(1995)提出的直接估计法,给出了可加分量的最近邻估计和核估计.在较弱的条件下,建立了可加分量最近邻估计和核估计的平均偏差的指数界.     

关于James-Stein估计的小样本性质

首先给出了Jam es-S te in估计优于岭估计的充分条件,随后在P itm an准则下给出了Jam es-S te in估计优于最小二乘估计的简短证明.     

Riemann Zeta函数的六阶和

利用概率论与组合数学的方法,研究了与Riemann-zeta函数ξ(k)的部分和ξ_n(k)有关的一些级数,计算出了一些重要的和式．特别的,Euler的著名结果5ξ(4)= 2ξ~2(2)能够从四阶和式直接推出．因此,通过计算全部的11个六阶和式,研究它们之间的非平凡关系,就有可能得到ξ(3)的数值．     

Post selection shrinkage estimation for high‐dimensional data analysis

In high‐dimensional data settings where p  ? n , many penalized regularization approaches were studied for simultaneous variable selection and estimation. However, with the existence of covariates with weak effect, many existing variable selection methods, including Lasso and its generations, cannot distinguish covariates with weak and no contribution. Thus, prediction based on a subset model of selected covariates only can be inefficient. In this paper, we propose a post selection shrinkage estimation strategy to improve the prediction performance of a selected subset model. Such a post selection shrinkage estimator (PSE) is data adaptive and constructed by shrinking a post selection weighted ridge estimator in the direction of a selected candidate subset. Under an asymptotic distributional quadratic risk criterion, its prediction performance is explored analytically. We show that the proposed post selection PSE performs better than the post selection weighted ridge estimator. More importantly, it improves the prediction performance of any candidate subset model selected from most existing Lasso‐type variable selection methods significantly. The relative performance of the post selection PSE is demonstrated by both simulation studies and real‐data analysis. Copyright © 2016 John Wiley & Sons, Ltd.     

回归系数的岭型主相关估计及其优良性

本文研究了线性模型中的一种有偏估计,利用均方误差和残差平方和,得到了岭型主相关估计的一些性质,是对[1]中相关结果的推广.     

岭回归中确定K值的一种方法

本文给出了岭估计中确定K值的一种新方法,这种方法改进了Hoerl和Kennard的相应方法。     

纵向数据下半参数回归模型估计的收敛速度

考虑纵向数据下半参数回归模型:yij=x′ijβ+g(tij)+eij,i=1,…,n,j=1,…,mi.基于最小二乘法和一般的非参数权函数方法给出了模型中参数β和回归函数g(·)的估计,并在适当条件下证明了参数分量β的估计量的强收敛速度和未知函数g(·)的估计量的一致强收敛速度.     

A Recursive Construction for 2-Designs

This paper contains two main results: given a symmetric (v, k, ) design, D, and a resolvable design which has the parameters of a residual design of D, there exists a symmetric (dv + 1, v, k) design, where d = (v - k)/(k - ), and d is a prime power; given a symmetric (v, k, ) design, D, and a resolvable design with the parameters of a derived design of D, there exists a 2 - (ek + v, 2k, k) design, where e = k/,and e is a prime power.     

Exact constants in inequalities between norms of derivatives of functions

In this article we shall concern ourselves with determining exact (least possible) constants in the inequalities of the form $$\parallel f^{(k)} \parallel _{L_q } \leqslant K\parallel f\parallel _{L_p } ^{\tfrac{{l - k - r - 1 + q - 1}}{{l - r - 1 + p - 1}}} \parallel f^{(l)} \parallel _{L_r } ^{\tfrac{{k - q - 1 + p - 1}}{{l - r - 1 + p^{n - 1} }}} $$ for functions defined on the entire (?∞, ∞), absolutely continuous on any interval together with their (l?1)-th derivatives, and having finite $$l = 2,k = 0,k = 1,q = r = \infty ,1 \leqslant p< \infty $$ is considered.     

亚纯函数结合其导数的值分布

研究了亚纯函数结合其导数的值分布问题,得到了一个有趣的不等式,此不等式概括了方-杨和I.Lahiri和S.Dewan的结果,应用此不等式还得到关于θ(a(z);φ)的一个估计,这里φ(z)=α(z)f~nM[f],M[f]=(f′)~(n_1)(f″)~(n_2)…(f~((k)))~(n_k),n_1,n_2,…,n_k,n为非负整数满足:n_1+n_2+…+n_k≥1,α(z),a(z)(≠00,∞)为f的小函数.     

Optimization based on quasi-Monte Carlo sampling to design state estimators for non-linear systems

State estimation for a class of non-linear, continuous-time dynamic systems affected by disturbances is investigated. The estimator is assigned a given structure that depends on an innovation function taking on the form of a ridge computational model, with some parameters to be optimized. The behaviour of the estimation error is analysed by using input-to-state stability. The design of the estimator is reduced to the determination of the parameters in such a way as to guarantee the regional exponential stability of the estimation error in a disturbance-free setting and to minimize a cost function that measures the effectiveness of the estimation when the system is affected by disturbances. Stability is achieved by constraining the derivative of a Lyapunov function to be negative definite on a grid of points, via the penalization of the constraints that are not satisfied. Low-discrepancy sampling techniques, typical of quasi-Monte Carlo methods, are exploited in order to reduce the computational burden in finding the optimal parameters of the innovation function. Simulation results are presented to investigate the performance of the estimator in comparison with the extended Kalman filter and in dependence of the complexity of the computational model and the sampling coarseness.