多参数指数族的逆Bhattacharyyà信息阵并应用于正态分布

本文进一步讨论多参数指数族中给定可估函数的 UMVUE 的方差计算问题.设定义于(X,B_X)上的 r.v.X 的分布为 P_θ,θ∈Θ.P_θ受某σ-有限测度μ(x)所控,称{P_θ,θ∈Θ)为自然指数族,是指     

多参数指数族的 UMVUE 及 Bhattacharyyà下界

对自然多参数指数族{f(x,θ)=exp[θ~τT(x)-ψ(θ)],θ∈Θ},这里Θ为 R_r 中非空开区域(r>1)及 T(x)=(T_1(x),…,T_r(x)),我们用 m=E_ΘT(x)代替θ作为参数.本文讨论了各种阶的 Fisher-Bhattacharyà 信息阵的正定性,给出了 G(m)可估的充要条件,证明了可估向量 G(m)=(g_1(m),…,g_p(m))的具有处处有限方差的 UMVUE 的协方差阵或等于某有限阶 B-下界或等于 B_n-下界的极限(n→∞).当 T(x)仅取有限个不同的向量值时,证明了可估函数 g(m)的方差必等于某给定阶的 B-下界.     

指数族的一致最小方差无偏估计与Bhattacharyyà下界

具有共同支撑的单参数分布族{pθ,θ∈},R,若它满足一定的正则条件,Cramér-Rao给出了可估函数g(θ)的所有无偏估计的方差的下界函数,即著名的C-R不等式。Fend和Wijsman指出:在这些正则条件下,g(θ)的无偏估计T(x)处处达到C-R下界的充要条件,是{pθ,θ∈}为指数族,对某σ-有限测度μ(x)的密度形为     

平方损失下的可容许估计

设(x)为样本空间,P_θ为其上的分布族,θ∈Θ为参数.欲估计g(θ),损失函数为L(g(θ),d).称R(g(θ),d(X))=EL(g(θ),d(X))为估计d(X)的风险函数.称d_0(X)是g(θ)的可容许估计,如果不存在其它估计d_1(X),使得R(g(θ),d_1(X))≤R(g(θ),d_0(X)),对一切θ∈Θ,且不等号至少对某θ_0∈Θ成立.设在参数空间Θ上建立了σ~-域θ,ξ为(Θ,θ)上σ~-有限测度.称g(θ)的估计δ_0(X)关于ξ是几乎可容     

条件马尔可夫过程的表示

设函数空间型马氏过程 X=(Ω,(?),(?)_t,X_t θ_t,P~x)是以(E,(?))为状态空间的暂留的Hunt 过程,ξ为(E,(?))上 Radon 测度,X 的位势核 U(x,A)=integral A u(x,y)ξ(dy),而 u(x,y)满足 chung、Rao[6]的基本假定。我们找到了一个由 u(x,y)确定的零势集∧(等价于ξ(∧)=0),证明了下述结论:定理 设μ为(?)上测度,μ(∧)=0,h=Uμ(?)∫u(·y)ξ(dy).记 E~h={0相似文献   

论达到多参数C-R下界的条件

问题和主要结果众所周知,在很宽的条件下,可估函数的无偏估计的方差(或协差阵)有 Cramér-Rao 下界.此问题的研究在统计参数估计理论中具有重要意义.首先设分布族{p_θ,θ∈(?)}被样本空间(?)上某σ有限测度μ所控,dp_θ/     

矩阵损失函数下参数估计的几个结果

在本文中,设随机向量 Y 的样本空间和分布族为((?)P_θ),θ∈(?),(?)为 p 维欧氏空间 R~p 中的 Borel 集.要估计θ的函数的向量 h(θ)=(h_1(θ),…,h_k(θ))'.文献[1]中第二章的定理1.4指出,若存在 h(θ)的无偏估计δ(Y),使得 E_θ(δ(Y)—h(θ))′(δ(Y)—h(θ))<∞,一切θ∈(?),则在损失函数(α—h(θ))′(α—h(θ))下,(?)(Y)是 h(θ)的一致最优无偏估计的充要条件是对 h(θ)的任何风险函数有限的无偏估     

二维单边截断型分布族中EB估计及其收敛速度

本文讨论了二维单边截断型分布族(I)中参数函数EB估计及其收敛速度。(I) f_0(x,y)dxdy=c(θ_1,θ_2)f_0(x,y)I_([α,θ_1;c,θ_2])(x,y)dxdy在适当的条件下,满足恰当条件参数函数Q(θ_1,θ_2)的EB估计的收敛速度可任意接近于 1。     

严格单调似然比分布族的完全类定理

引言和主要结果设X是一维随机变量,其分布函数为;。={F_θ:θ∈Θ}此处,v 为σ有限测度.记(?)={F_θ:θ∈Θ}.设参数空间Θ和行动空间(?)都是 R_1的子集,损失函数 L(θ,α),对每一个固定的θ,是α的下半连续函数,且存在θ的单调上升函数 q(θ),使a)对每一个固定的θ,L(θ,α)在α=q(θ)处达到极小,且当 α>q(θ)时非降,当αα有 L(θ_α,α′)>     

具有非局部边值约束的中子迁移问题单调衰减解

中子迁移方程是核动力学中描述中子的分布过程.本文讨论具有非局部边值约束的含任意空穴的非均匀介质中具连续能量的中子迁移系统其中D_1是R~3中含任意空穴的有界凸区域,且其边界(?)D_1分片光滑,D_2是R~3中有界可测集,表示位于x=(x_1,X_2,X-3)处,具有速度v=(v_1,v_2,v_3)在时刻t的中子分布密度.N_0(x,v)是初始分布,σ(x,v)表示含任意空穴的非均匀介质的总截面,k(x,v,v′)是能量迁移核,f(x,y)是边界约束函数,σ、k、N_0、f均为非负连续有界函数.另记     

THE INTRINSIC QUALITATIVE PROPERTIES OF THE CLASSICAL OPTIMAL STOPPING PROBLEM ARE INVARIANT TO THE FUNCTIONAL FORM OF THE DISCOUNT FUNCTION

Abstract The intrinsic qualitative properties of a generic optimal stopping model are shown to be invariant to the functional form of the discount function. If the discount function is assumed to be a member of particular infinite parametric family—a family that includes the exponential and classical hyperbolic discount functions as special cases—an additional refutable comparative statics result is produced that holds for the entire family. Consequently, if one limits econometric tests of the model to its qualitative properties, one cannot determine the form of the discount function used by the decision maker. It is also shown that the only discount function that yields a time‐consistent stopping rule is the exponential function with a constant rate of discount.     

On a fundamental optimality of the maximum likelihood estimator

It is well-known that the rate of exponential convergence for any consistent estimator is less than or equal to the Bahadur bound. In this paper we have proven, for the one-dimensional case, that the rate of exponential convergence for the maximum likelihood estimator (m.l.e.) attains the Bahadur bound if and only if the underlying distribution is a member of the exponential family of distributions.     

More comparisons of MLE with UMVUE for exponential families

Under some regularity conditions, the asymptotic expected deficiency (AED) of the maximum likelihood estimator (MLE) relative to the uniformly minimum variance unbiased estimator (UMVUE) for a given one-parameter estimable function of an exponential family is obtained. The exact expressions of the AED for normal, lognormal, inverse Gaussian, exponential (or gamma), Pareto, hyperbolic secant, Bernoulli, Poisson and geometric (or negative binomial) distributions are also derived.     

Nonuniform Dichotomies via Lyapunov Functions

For a nonautonomous linear equation x′ =  A(t)x we show how to characterize a nonuniform exponential dichotomy using strict Lyapunov functions. In particular, the stable and unstable subspaces are obtained from invariant families of cones determined by each Lyapunov function. We also obtain converse theorems, constructing explicitly a family of strict Lyapunov functions for each nonuniform exponential dichotomy. We emphasize that nonuniform exponential dichotomies include as a very particular case (uniform) exponential dichotomies.     

The generalized linear exponential distribution

This paper deals with a new generalization of the linear exponential distribution. This distribution is called the generalized linear exponential distribution (GLED). Some statistical properties such as moments, modes and quantiles are derived. The failure rate function and the mean residual lifetime are also discussed. The maximum likelihood estimators of the parameters are obtained using a simulation study. Real data are used to determine whether the GLED is better than other well-known distributions in modeling lifetime data or not.     

GLOBAL EXISTENCE, UNIFORM DECAY AND EXPONENTIAL GROWTH FOR A CLASS OF SEMI-LINEAR WAVE EQUATION WITH STRONG DAMPING

In this paper, we consider the nonlinearly damped semi-linear wave equation associated with initial and Dirichlet boundary conditions. We prove the existence of a local weak solution and introduce a family of potential wells and discuss the invariants and vacuum isolating behavior of solutions. Furthermore, we prove the global existence of solutions in both cases which are polynomial and exponential decay in the energy space respectively, and the asymptotic behavior of solutions for the cases of potential well family with 0相似文献   

Multivariate maximum entropy identification, transformation, and dependence

This paper shows that multivariate distributions can be characterized as maximum entropy (ME) models based on the well-known general representation of density function of the ME distribution subject to moment constraints. In this approach, the problem of ME characterization simplifies to the problem of representing the multivariate density in the ME form, hence there is no need for case-by-case proofs by calculus of variations or other methods. The main vehicle for this ME characterization approach is the information distinguishability relationship, which extends to the multivariate case. Results are also formulated that encapsulate implications of the multiplication rule of probability and the entropy transformation formula for ME characterization. The dependence structure of multivariate ME distribution in terms of the moments and the support of distribution is studied. The relationships of ME distributions with the exponential family and with bivariate distributions having exponential family conditionals are explored. Applications include new ME characterizations of many bivariate distributions, including some singular distributions.     

An Algorithm for the Computation of Hermite–Padé Approximations to the Exponential Function: Divided Differences and Hermite–Padé Forms

We present the first of two different algorithms for the explicit computation of Hermite–Padé forms (HPF) associated with the exponential function. Some roots of the algebraic equation associated with a given HPF are good approximants to the exponential in some subsets of the complex plane: they are called Hermite–Padé approximants (HPA) to this function. Our algorithm is recursive and based upon the expression of HPF as divided differences of the function texp(xt) at multiple integer nodes. Using this algorithm, we find again the results obtained by Borwein and Driver for quadratic HPF. As an example, we give an interesting family of quadratic HPA to the exponential.     

A class of infinitely divisible variance functions with an application to the polynomial case

Let be a natural exponential family on ??? with variance function (V, Ω). Here, Ω is the mean domain of and V is its variance expressed in terms of the mean μ ε Ω. In this note we prove the following result. Consider an open interval Ω = (0, b), 0 < b ∞, and a positive real analytic function V on Ω. If V² is absolutely monotone on [0, b) and V has the form μt(μ), where 1 and t is real analytic in a neighborhood of zero, then there exits an infinitely divisible natural exponential family with variance function (V, Ω). We illustrate this result with several examples of general nature.     

On the equivalence of brushlet and wavelet bases

We prove that the Meyer wavelet basis and a class of brushlet systems associated with exponential type partitions of the frequency axis form a family of equivalent (unconditional) bases for the Besov and Triebel-Lizorkin function spaces. This equivalence is then used to obtain new results on nonlinear approximation with brushlets in Triebel-Lizorkin spaces.