基于充分降维的半参数不可忽略无响应光滑分位回归

文章在响应变量不可忽略缺失假定下,考察了分位回归的估计问题.文章首先建立半参数指数倾斜响应模型,为克服不可忽略缺失数据的识别性困难,避免多元非参数核估计造成的维数灾难,文章基于充分降维假设,利用数据驱动方法构造缺失工具变量,得到倾斜参数的轮廓两步广义矩估计量和非参数部分的降维核估计量;基于上述估计量建立逆概率加权(IPW)、核辅助估计方程插补(EEI)和增强逆概率加权(AIPW)三种分位回归估计方程,并利用卷积平滑分位损失函数代替经典的分位损失函数克服检查函数不平滑造成的理论和计算困难,回归系数的估计量由经验似然方法得到.理论研究证明了三种估计量等价的渐近正态性和相应对数经验似然比函数的渐近χ²加权和性质.数值模拟比较了上述估计量的有限样本性能.最后对HIV-CD4实际数据进行分析.     

一种惩罚最大似然方法估计混合回归模型

本文考虑具有正态误差假设下混合回归模型的参数估计问题.由于似然函数的无界性,混合回归模型普通的最大似然估计不存在.本文提出一种惩罚最大似然方法来估计混合回归模型的参数,证明惩罚最大似然估计量(penalized maximum likelihood estimation, PMLE)具有强相合和渐近正态性.通过深入模拟研究,从估计精确性角度看,惩罚最大似然估计量有很好的表现.本文还给出一个音调感知的例子来说明理论结果的应用.     

离散抽样Gamma-OU过程的参数估计

平稳Gamma-OU过程是用于刻画金融资产波动的一类重要模型. 本文主要考虑基于离散观察的Gamma-OU过程的参数估计. 文中给出了强度参数λ的估计量及其收敛性，模拟显示这一估计是相当准确的. 在假设参数λ已被估计出来的条件下, 又研究了形状参数c和尺度参数α的最大似然估计, 其中关于这两个参数的似然函数是难于计算的. 通过Gaver-Stehfest算法, 我们构造了一个似然函数的具体估计序列, 它收敛于真实(但未知）的似然函数. 最大化这一序列可以得到收敛于真实最大似然估计的一列估计量, 并且这一估计序列具有与最大似然估计相同的收敛性. 模拟显示在大多数有关波动率的实际背景下, 我们的方法是非常准确的.     

缺失数据下含几何分布的对数线性模型的EM算法

本文研究缺失数据下对数线性模型参数的极大似然估计问题.通过Monte-Carlo EM算法去拟合所提出的模型.其中,在期望步中利用Metropolis-Hastings算法产生一个缺失数据的样本,在最大化步中利用Newton-Raphson迭代使似然函数最大化.最后,利用观测数据的Fisher信息得到参数极大似然估计的渐近方差和标准误差.     

缺失数据下的两步抽样法估计

在数据缺失机制形式未知时,通过两步抽样得到了分布函数的相合估计量,证明了该估计量的渐近正态性.文中假设第二次抽样时的数据缺失机制与第一次抽样时的数据缺失机制函数形式类似,允许两者有一个一维未知参数的差别.     

有缺失数据的条件独立正态母体中参数的最优同变估计

在缺失数据机制是可忽略的假设下,导出了有单调缺失数据的条件独立正态模型中协方差阵和精度阵的Cholesky分解的最大似然估计和无偏估计.通过引入一类特殊的变换群并在更广义的损失下,获得了其最优同变估计.这表明最大似然估计和无偏估计是非容许的.最后,通过数值模拟验证了相关结果的有效性.     

缺失数据均值推断的经验似然方法

研究了缺失数据的均值推断问题.在随机缺失及半参数模型的假设下,设计了基于影响函数理论的经验似然推断方法,证明了所构造的对数经验似然比检验统计量具有非参数Wilks性质.此外,该经验似然方法可以利用辅助协变量中提供的附加信息来提高检验的功效.在近邻备择假设下,计算了检验统计量的功效,并且通过一些模拟考察了该方法在有限样本下的表现.     

缺失数据均值推断的经验似然方法

研究了缺失数据的均值推断问题.在随机缺失及半参数模型的假设下,设计了基于影响函数理论的经验似然推断方法,证明了所构造的对数经验似然比检验统计量具有非参数Wilks性质.此外,该经验似然方法可以利用辅助协变量中提供的附加信息来提高检验的功效.在近邻备择假设下,计算了检验统计量的功效,并且通过一些模拟考察了该方法在有限样本下的表现.     

正倒向随机微分方程的参数估计

本文的研究目标是离散观测下正倒向随机微分方程中未知参数的估计及其性质.作为第一步,本文考虑一个线性模型.本文先导出两个状态过程的关系式,进而找到离散观测数据的似然函数.最后详细讨论最大似然估计量的相合性和渐近正态性.     

基于纵向数据与生存时间数据联合建模的变量选择

纵向数据和生存时间数据联合建模能减少由单独建模所引起的偏差,本文研究了基于纵向数据和生存时间联合建模的变量选择问题。对于生存时间数据,把生存时间做离散化处理,引入离散风险函数的Probit模型;对于纵向数据,利用线性混合效应模型建模。采用共享随机效应的方法对纵向数据和生存时间进行联合建模,通过利用多元高斯隐截断分布,构造出联合模型的精确似然。然后对似然函数加惩罚,重新构造目标函数,得到回归系数的稀疏估计量。理论证明以及数值模拟研究展示了稀疏估计量的良好性质。     

Distribution of types of symmetry of tensors of high degree

The asymptotic distribution of tensors of degree N in symmetry types is studied in this paper.Translated from Zapiski Nauchnykh Seminarov Leningradskogo Otdeleniya Matematicheskogo Instituta im. V. A. Steklova AN SSSR, Vol. 155, pp. 181–186, 1986.     

Estimates of stability of characterization of additive types of distributions

An estimate of stability of characterization of distribution types is obtained for the case of additive types. Under some conditions, the estimate has the order ε^1/3L(ε), where L(ε) is a slowly varying function. Proceedings of the Seminar on Stability Problems for Stochastic Models, Moscow, Russia, 1996, Part I.     

Study of the blow-up of solutions of systems of equations of hydrodynamic type

Mathematical Notes - We study the initial boundary-value problem for three-dimensional systems of equations of pseudoparabolic type. The system is similar to the Oskolkov system, but differs from...     

一类完全正则半群簇子簇格的性质

本文研究了完全正则半群簇的子簇格［Ｖ＋∩ＰＶ，Ｖ＋∩ＰＶ］的某些格运算性质，我们证明了簇Ｖ＋∩ＰＶ可分解为Ｖ与Ｖ＋∩ＰＶ的并；对任意完全正则半群簇Ｗ，有Ｗ∩（Ｖ∨Ｖ＋∩ＰＶ）＝（Ｗ∩Ｖ）∨（Ｗ∩Ｖ＋∩ＰＶ）．特别地，我们得到了等式Ｖ＋∩ＰＶ＝Ｖ成立的若干条件．     

Characterization of types of asymptotic discernibility of families of hypotheses

We give a characterization of the types of asymptotic discernibility of families of hypotheses in the case of hypothetical measures that are not, in general, mutually absolutely continuous. The case when the logarithm of the likelihood ratio admits an asymptotic expansion of the type of an expansion with local asymptotic normality is examined in detail. Examples are studied.Translated fromTeoriya Sluchainykh Protsessov, Vol. 15, pp. 64–71, 1987.     

Bases of the identities of products of certain varieties of groups

The following theorem is proved: The product of any variety of two-step solvable groups and a variety having a finite basis of identity relations has a finite basis of identity relations.Translated from Matematicheskie Zametki, Vol. 5, No. 1, pp. 137–144, January, 1969.     

Preservation of equivalence of derivations under reduction of depth of formulas

In this paper we consider derivations in the (&, )-fragment of the intuitionistic propositional calculus. As is known, replacement of any occurrence of a formula [F] in a sequent S by an occurrence of the formula [p], where p is a new propositional variable, with the simultaneous addition to the antecedent of the formula F p or p F depending on the sign of the occurrence of F in S, leaves the derivability unchanged. We give a proof of the fact that the natural extension of this transformation to derivations preserves the relation of equivalence of derivations, i.e., transformed derivations are equivalent if and only if the originals are equivalent. (Derivations are considered equivalent if certain of their normal forms coincide, or, what is the same, if their deductive terms coincide.) It is proved that by the iteration of this transformation, each derivation of an arbitrary sequent S can be transformed into a derivation of a sequent S, depending only on S, whose succedent is a variable, and in the antecedent there occur only formulas of the form a,a & b, a b,,(a b) c, a & b c, a (b & c), wherea, b, c are variables. Here if S is balanced, then S is also balanced. (A sequent is called balanced if each variable occurs in it no more than twice.) The familiar correspondence between certain concepts of the theory of categories and concepts of the theory of proofs allows one to assert that there has been constructed a univalent functor, mapping a free Cartesian closed category into itself.Translated from Zapiski Nauchnykh Seminarov Leningradskogo Otdeleniya Matematicheskogo Instituta im. V. A. Steklova AN SSSR, Vol. 88, pp. 197–207, 1979.