数学杂志  2016, Vol. 36 Issue (4): 859-866   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
高慧
郭明乐
祝东进
行为负相依随机变量阵列加权和的完全收敛性
高慧, 郭明乐, 祝东进     
安徽师范大学数学计算机科学学院, 安徽 芜湖 241003
摘要:本文研究了行为NOD随机变量阵列加权和的完全收敛性.运用NOD随机变量列的矩不等式以及截尾的方法, 得到了关于行为NOD随机变量阵列加权和的完全收敛性的充分条件.利用获得的充分条件, 推广了Baek(2008) 关于行为NA随机变量阵列加权和的完全收敛性的结论, 得到了比吴群英(2012) 更为一般的结果.
关键词NOD随机变量列    NA随机变量列    完全收敛性    矩不等式    
COMPLETE CONVERGENCE OF WEIGHTED SUMS FOR ARRAYS OF NOD RANDOM VARIABLES
Gao Hui, Guo Mingle, Zhu Dongjin     
School of Mathematics and Computer Science, Anhui Normal University, Wuhu 241003, China
Abstract: In this article, the complete convergence theorem for weighted sums for arrays of rowwise NOD random variables is investigated. By using moment inequality of negatively dependent random variables and truncation method, the sufficient conditions for complete convergence of weighted sums for arrays of NOD random variables are obtained. Using the sufficient conditions, we can promote the Bake's (2008) conclusion on complete convergence of weighted sums for arrays of NOD random variables, and we can get more general results than Wu (2012).
Key words: NOD random variables     NA random variables     complete convergence     moment inequality    
1 引言

称随机变量$X$$Y$是NOD(Negatively Orthant Dependent)的, 如果

$P(X\leq x, Y\leq y)\leq P(X\leq x)P(Y\leq y), \forall x, y\in R.$ (1.1)

事实上, (1.1) 式与下面的(1.2) 式是等价的,

$P(X>x, Y> y)\leq P(X> x)P(Y> y), \forall x, y\in R.$ (1.2)

但是, 当$n\geq3$时, (1.1) 与(1.2) 却并不等价, 因此, 需要下面的定义来定义NOD随机变量列, Joag-Dev和Proschan[1]在1983年提出了NOD和NA(Negatively Associated)随机变量列的概念.

定义1.1 称随机变量$X_1, X_2, \cdots, X_n$是NOD的, 如果对任意实数$x_1, x_2, \cdots$ $, x_n$

$\begin{array}{l} P\left( {\bigcap\limits_{i = 1}^n {({X_i} \le {x_i})} } \right) \le \prod\limits_{i = 1}^n P ({X_i} \le {x_i});\\ P\left( {\bigcap\limits_{i = 1}^n {({X_i} > {x_i})} } \right) \le \prod\limits_{i = 1}^n P ({X_i} > {x_i}), \end{array}$

称随机变量列$\{X_n, n\geq1\}$是NOD的, 如果它的每一个有限子集序列$X_1, X_2, \cdots, X_n$都是NOD的.

定义1.2 称随机变量$X_{1}, X_{2}, \cdots, X_{n}(n\geq2)$是NA的, 如果对于集合$\{1, 2\cdot\cdot\cdot, n\}$的任何两个不相交的非空子集$A$$B$, 都有

$\text{Cov}(f_{1}(X_{i}, i\in{A}), f_{2}(X_{j}, j\in{B}))\leq0, $

其中$f_{1}$$f_{2}$是任何两个使得协方差存在的且对每个变元均非降(或同为对每个变元均非升)的函数.称随机变量列$\{X_{n}, n\geq 1\}$是NA的, 如果对任何$n\geq2, X_{1}, \cdots, X_{n}$都是NA的.

易见, NOD随机变量列是比NA随机变量列弱的一种序列, 由于NOD随机变量列在工程技术领域都有较为广泛的应用, 所以将NA随机变量列的某些性质推广到NOD随机变量列上具有重要的理论和应用价值.

我国著名统计学家许宝禄与Robbins[2]提出了完全收敛性的这一概念, 它是随机变量列一种非常重要的收敛概念, 当假设$\{X_n, n\geq1\}$是概率空间$(\Omega, \mathfrak{\Im}, P)$上的随机变量列, 称$\{X_n\}$是完全收敛于常数C, 如果对于任意的$\epsilon >0, $

$\sum\limits_{n = 1}^\infty P (|{X_n} - C| > \varepsilon ) < \infty . $

许宝禄与Robbins[2]证明了独立同分布随机变量列在方差有限的情况下, 序列的样本均值完全收敛到总体均值, 随后Erdos[3]证明了它的逆命题也是成立的.由Borel-Cantelli引理可知, 完全收敛性可以推出几乎处处收敛.因此, 随机变量列的完全收敛性的研究就显的更为基本.目前, 在许多方向对其都进行了推广和完善.例如Baum与Katz[4]通过独立随机变量的一些收敛定理得到了Baum -Katz型的完全收敛性.

下面我们先介绍本文中几个必要的概念和符号.

定义1.3 称随机变量序列$\{X_n, n\geq1\}$尾概率有界于随机变量$X$(记为$\{X_n\}\prec X$), 如果存在正常数$C$, 使得对充分大的$x\geq 0$, 有

$\mathop {\sup }\limits_{n \ge 1} P(|{X_n}| \ge x) \le CP(|X| \ge x).$

本文中的常数一律以$C(>0)$表示, 在不同的地方$C$可表示不同值; $a_n\simeq b_n$表示存在正常数$C$, 使得对任意$n\geq 1, $$a_n\leq Cb_n$.

下面这个引理是为了保证, 对NOD随机变量列采取单调截尾法后, 能够使得截尾后的随机变量列仍是NOD的.

引理1.4[5] 设$X_1, X_2, \cdots, X_n$是NOD随机变量列, $f_1, f_2, \cdots, f_n$全部是单调增(或单调减)函数, 则$f_1(X_1), f_2(X_2), \cdots, f_n(X_n)$仍是NOD随机变量列.

下面这个引理是由Asadian[6]等建立的, 是关于NOD随机变量列的Rosenthal型矩不等式, 该不等式在证明中起着重要作用.

引理1.5 设$\{X_n, n\geq 1\}$是零均值的NOD随机变量列, 且$E|X_n|^t<\infty$, $ n\geq 1$.则存在仅依赖于$t$的正常数$C$(与$n$无关), 使得对任意$n\geq 1, $

$\begin{array}{l} E{\left| {\sum\limits_{i = 1}^n {{X_i}} } \right|^t} \le C\sum\limits_{i = 1}^n E |{X_i}{|^t},{\mkern 1mu} {\mkern 1mu} \forall 1 < t \le 2,\\ E{\left| {\sum\limits_{i = 1}^n {{X_i}} } \right|^t} \le C\left( {\sum\limits_{i = 1}^n E |{X_i}{|^t} + {{\left( {\sum\limits_{i = 1}^n E X_i^2} \right)}^{t/2}}} \right),{\mkern 1mu} {\mkern 1mu} \forall t > 2. \end{array}$

下面这个引理是由吴群英[7]建立的, 关于尾概率有界于X的随机变量列$\{X_n, n\geq 1\}$的矩的不等式.

引理1.6 设随机变量序列$\{X_n, n\geq 1\}$尾概率有界于随机变量$X$, 则对任意$t>0, \, u>0, \, n\geq 1$, 有

(ⅰ) $E|X_{n}|^u I(|X_{n}|> t)\leq CE|X|^u I(|X|> t)$;

(ⅱ) $E|X_{n}|^u I(|X_{n}|\leq t)\leq C\bigg(E|X|^u I(|X|\leq t)+t^u P(|X|> t)\bigg).$

引理1.7[9] 设$\{X_{ni}, i\geq 1, n\geq1\}$是行为ND随机变量阵列, 满足以下三个条件,

(ⅰ)$\sum\limits_{n = 1}^\infty {{c_n}} \sum\limits_{i = 1}^{{k_n}} P (|{X_{ni}}| > \epsilon ) < \infty ,{\mkern 1mu} {\mkern 1mu} \forall \epsilon > 0, $

(ⅱ)存在$q>2, \delta>0, $使得

$\sum\limits_{n = 1}^\infty {{c_n}} {(\sum\limits_{i = 1}^{{k_n}} E X_{ni}^2I(|{X_{ni}}| \le \delta ))^{\frac{q}{2}}}{\rm{ < }}\infty , $

(ⅲ)对于任意$n\rightarrow\infty$, 有$\sum\limits_{i = 1}^{{k_n}} E {X_{ni}}I(|{X_{ni}}| \le \delta ) \to 0, $

则对于任意$\epsilon >0$,有

$\sum\limits_{n = 1}^\infty {{c_n}} P\left( {\left| {\sum\limits_{i = 1}^{{k_n}} {{X_{ni}}} } \right| > \epsilon } \right) < \infty . $

$\sum\limits_{i = 1}^\infty {{X_{ni}}} $几乎处处收敛, 则上述定理对于$k_n=\infty$仍然成立, 即定理1.7就是完全收敛性的充分条件.

2 主要结果及证明

定理2.1 设$\{X_{ni}, i\geq 1, n\geq 1\}$是行为NOD随机变量阵列, 存在正常数$C$, 使得对充分大的$x\geq 0$, 有

$P(|{X_{ni}}| \ge x) \le CP(|X| \ge x),\;\forall {\mkern 1mu} {\mkern 1mu} n \ge 1,{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} i \ge 1,{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} x \ge 0.$ (2.1)

$\beta\geq -1$, 实数阵列$\{a_{ni}, i\geq 1, n\geq1\}$满足

$\mathop {\sup }\limits_{i \ge 1} |{a_{ni}}| = O({n^{ - r}}),{\mbox{对某个 }} r > 0$ (2.2)

$\sum\limits_{i = 1}^\infty {{{\left| {{a_{ni}}} \right|}^\theta }} = O({n^\alpha }),{\mbox{对某个}}\alpha < 2r,{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} 0 < \theta < \min (2,2 - \alpha /r).$ (2.3)

(ⅰ)当$\alpha+\beta+1>0$时, 若存在某个$\sigma>0, $使得${\displaystyle\frac {\alpha}{r}}+\theta<\sigma\leq2, $$s=\max(\theta+\displaystyle\frac{\alpha+\beta+1}r, \sigma)$, 且当$s>1$时, $EX_{ni}=0, i\geq 1, n\geq 1, $$E|X|^s<\infty, $

$\sum\limits_{n = 1}^\infty {{n^\beta }} P\left( {\left| {\sum\limits_{i = 1}^\infty {{a_{ni}}} {X_{ni}}} \right| > \epsilon } \right) < \infty , \forall \epsilon > 0.$ (2.4)

(ⅱ)当$\alpha+\beta+1=0$时, 若$E|X|^\theta\log(1+|X|)<\infty, $则(1.5) 成立.

 运用引理1.7, 在这里采用$c_n=n^\beta, k_n=\infty, $以及用$\{a_{ni}X_{ni}, i\geq1, n\geq1\}$代替$\{X_{ni}, i\geq1, n\geq1\}.$假设$a_{ni}^{+}=\max\{a_{ni}, 0\}\geq 0$$a_{ni}^{-}=\max\{-a_{ni}, 0\}\geq 0, $这样(2.4) 的证明就转化为证明下面两个式子, 对于任意的$\epsilon >0, $

$\sum\limits_{n = 1}^\infty {{n^\beta }} P\left( {\left| {\sum\limits_{i = 1}^\infty {a_{ni}^ + } {X_{ni}}} \right| > \epsilon /2} \right) < \infty , \sum\limits_{n = 1}^\infty {{n^\beta }} P\left( {\left| {\sum\limits_{i = 1}^\infty {a_{ni}^ - } {X_{ni}}} \right| > \epsilon /2} \right) < \infty .$

因此不失一般性, 对于任意的$i, n\geq 1, $我们假设$a_{ni}>0$

$\mathop {\sup }\limits_{i \ge 1} {a_{ni}} \le {n^{ - r}}.$ (2.5)

由引理1.7, 仅需证

$\begin{array}{l} {I_1} = :\sum\limits_{n = 1}^\infty {{n^\beta }} \sum\limits_{i = 1}^\infty P (|{a_{ni}}{X_{ni}}| > \epsilon ) < \infty ,{\mkern 1mu} {\mkern 1mu} \forall \epsilon > 0;\\ {I_2} = :\sum\limits_{n = 1}^\infty {{n^\beta }} {\left( {\sum\limits_{i = 1}^\infty E a_{ni}^2X_{ni}^2I(|{a_{ni}}{X_{ni}}| \le \delta )} \right)^{\frac{q}{2}}},{\mbox{对某个}} q > 2,\delta > 0;\\ {I_3} = :\sum\limits_{i = 1}^\infty E {a_{ni}}{X_{ni}}I(|{a_{ni}}{X_{ni}}| \le \delta ) \to 0. \end{array}$

首先, 我们来证明$I_1<\infty$, 根据引理1.6, (2.1), (2.2), (2.3), (2.5) 以及Makov不等式有

$\begin{array}{*{20}{l}} {{I_1}}&{ \le C\sum\limits_{n = 1}^\infty {{n^\beta }} \sum\limits_{i = 1}^\infty P (|{a_{ni}}X| > \epsilon ) \le C\sum\limits_{n = 1}^\infty {{n^\beta }} \sum\limits_{i = 1}^\infty E {{\left| {\frac{{{a_{ni}}X}}{\varepsilon }} \right|}^\theta }I(|{a_{ni}}X| > \epsilon )}\\ {}&{ \le C\sum\limits_{n = 1}^\infty {{n^\beta }} \sum\limits_{i = 1}^\infty E |{a_{ni}}X{|^\theta }I(|{a_{ni}}X| > \epsilon ) \le C\sum\limits_{n = 1}^\infty {{n^\beta }} \sum\limits_{i = 1}^\infty | {a_{ni}}{|^\theta }E|X{|^\theta }I(|X| > \epsilon {n^r})}\\ {}&{ \simeq \sum\limits_{n = 1}^\infty {{n^{\beta + \alpha }}} E|X{|^\theta }I(|X| > \epsilon {n^r}).} \end{array}$

下面分两种情况讨论:

(1) 当$\alpha+\beta+1>0$时,

$\begin{array}{*{20}{l}} {{I_1}}&{ \simeq \sum\limits_{n = 1}^\infty {{n^{\beta + \alpha }}} \sum\limits_{j = n}^\infty E |X{|^\theta }I({j^r} < \left| X \right| \le {{(j + 1)}^r})}\\ {}&{ = \sum\limits_{j = 1}^\infty E |X{|^\theta }I({j^r} < |X| \le {{(j + 1)}^r})\sum\limits_{n = 1}^j {{n^{\beta + \alpha }}} }\\ {}&{ \le \sum\limits_{j = 1}^\infty {{j^{\beta + \alpha + 1}}} E|X{|^\theta }I({j^r} < |X| \le {{(j + 1)}^r})}\\ {}&{ \simeq \sum\limits_{j = 1}^\infty E |X{|^{\theta + \frac{{\beta + \alpha + 1}}{r}}}I({j^r} < |X| \le {{(j + 1)}^r})}\\ {}&{ \simeq E{{\left| X \right|}^s} < \infty .} \end{array}$ (2.6)

(2) 当$\alpha+\beta+1=0$时,

$\begin{array}{*{20}{l}} {{I_1}}&{ \simeq \sum\limits_{n = 1}^\infty {{n^{ - 1}}} \sum\limits_{j = n}^\infty E |X{|^\theta }I({j^r} < |X| \le {{(j + 1)}^r})}\\ {}&{ = \sum\limits_{j = 1}^\infty E |X{|^\theta }I({j^r} < |X| \le {{(j + 1)}^r})\sum\limits_{n = 1}^j {{n^{ - 1}}} }\\ {}&{ \le \sum\limits_{j = 1}^\infty {\ln } jE|X{|^\theta }I({j^r} < |X| \le {{(j + 1)}^r})}\\ {}&{ \simeq E|X{|^\theta }\ln (1 + |X|) < \infty .} \end{array}$ (2.7)

结合(2.6), (2.7) 式, 可知$I_1<\infty$.

其次来证明$I_2<\infty, $为此根据$s$的范围分以下两种情形讨论.

情形1 $s>2$.

此时$s=\theta+\displaystyle\frac {\alpha+\beta+1}r, $因为高阶矩有限, 低阶矩一定有限, 所以当$E|X|^s<\infty$时, $E|X|^2<\infty.$又因为$\alpha+r(\theta-2)<0, $所以存在$q\geq2, $使得

$ \beta+(\alpha+r(\theta-2)){\displaystyle\frac{q}{2}}<-1. $

所以根据(1.4), 引理1.6以及示性函数的性质, 可以得到

$\begin{array}{*{20}{l}} {{I_2}}&{ \simeq \sum\limits_{n = 1}^\infty {{n^\beta }} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^2} EX_{ni}^2I(|{X_{ni}}| \le \delta )} \right)}^{\frac{q}{2}}}}\\ {}&{ \le \sum\limits_{n = 1}^\infty {{n^\beta }} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^2} } \right)}^{\frac{q}{2}}} = \sum\limits_{n = 1}^\infty {{n^\beta }} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^\theta } a_{ni}^{2 - \theta }} \right)}^{\frac{q}{2}}}}\\ {}&{ \simeq \sum\limits_{n = 1}^\infty {{n^\beta }} {{({n^{\alpha + r(\theta - 2)}})}^{\frac{q}{2}}} = \sum\limits_{n = 1}^\infty {{n^{\beta + (\alpha + r(\theta - 2))\frac{q}{2}}}} < \infty .} \end{array}$ (2.8)

情形2 $s\leq2$

$\beta>-1, $此时$s=\max(\theta+\displaystyle\frac {\alpha+\beta+1}r, \sigma), $${\displaystyle\frac {\alpha}{r}}+\theta<\sigma\leq s\leq2.$$\alpha+r(\theta-s)<0, $所以存在$q\geq2, $使得

$ \beta+(\alpha+r(\theta-s)){\displaystyle\frac{q}{2}}<-1, $

根据引理1.6, (2.3) 式以及$E|X|^s<\infty$, 有

$\begin{array}{*{20}{l}} {{I_2}}&{ \simeq \sum\limits_{n = 1}^\infty {{n^\beta }} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^s} E|{X_{ni}}{|^s}I(|{a_{ni}}{X_{ni}}| \le \delta )} \right)}^{\frac{q}{2}}}}\\ {}&{ \le \sum\limits_{n = 1}^\infty {{n^\beta }} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^s} } \right)}^{\frac{q}{2}}} = \sum\limits_{n = 1}^\infty {{n^\beta }} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^\theta } a_{ni}^{s - \theta }} \right)}^{\frac{q}{2}}}}\\ {}&{ \simeq \sum\limits_{n = 1}^\infty {{n^\beta }} {{({n^{\alpha + r(\theta - s)}})}^{\frac{q}{2}}} = \sum\limits_{n = 1}^\infty {{n^{\beta + (\alpha + r(\theta - s))\frac{q}{2}}}} < \infty .} \end{array}$ (2.9)

$\beta=-1, $此时$s=\max(\theta+\displaystyle\frac {\alpha}r, \sigma)=\sigma, $以及${\displaystyle\frac {\alpha}{r}}+\theta<\sigma\leq2.$$\alpha+r(\theta-\sigma)<0, $所以存在$q\geq2, $使得$-1+(\alpha+r(\theta-\sigma)){\displaystyle\frac{q}{2}}<-1$, 所以类似$\beta>-1$时的证法,有

$\begin{array}{*{20}{l}} {{I_2}}&{ \simeq \sum\limits_{n = 1}^\infty {{n^{ - 1}}} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^\sigma } E|{X_{ni}}{|^\sigma }I(|{a_{ni}}{X_{ni}}| \le \delta )} \right)}^{\frac{q}{2}}}}\\ {}&{ \le \sum\limits_{n = 1}^\infty {{n^{ - 1}}} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^\sigma } } \right)}^{\frac{q}{2}}} = \sum\limits_{n = 1}^\infty {{n^{ - 1}}} {{\left( {\sum\limits_{i = 1}^\infty {a_{ni}^\theta } a_{ni}^{\sigma - \theta }} \right)}^{\frac{q}{2}}}}\\ {}&{ \simeq \sum\limits_{n = 1}^\infty {{n^{ - 1}}} {{({n^{\alpha + r(\theta - \sigma )}})}^{\frac{q}{2}}} = \sum\limits_{n = 1}^\infty {{n^{ - 1 + (\alpha + r(\theta - \sigma ))\frac{q}{2}}}} < \infty .} \end{array}$ (2.10)

综上所述, 由(2.8)-(2.10) 式可知$I_2<\infty.$

最后证明$I_3\rightarrow 0, $仅需要证明

${I_3} = \sum\limits_{i = 1}^\infty {{a_{ni}}} E{X_{ni}}I(|{a_{ni}}{X_{ni}}| \le \delta ) \to 0,\, \, n \to \infty . $

根据证明的需要, 分下面两种情形.

情形1 $s\leq1 $.

根据引理1.6, (2.1)-(2.3), (2.5) 式以及$E|X|^s<\infty, $可以得到

$\begin{array}{*{20}{l}} {\sum\limits_{i = 1}^\infty {{a_{ni}}} |E{X_{ni}}I(|{a_{ni}}{X_{ni}}| \le \delta )|}&{ \simeq \sum\limits_{i = 1}^\infty {a_{ni}^s} E|X{|^s}I(|{a_{ni}}X| \le \delta )}\\ {}&{ \simeq \sum\limits_{i = 1}^\infty {a_{ni}^s} = \sum\limits_{i = 1}^\infty {a_{ni}^\theta } a_{ni}^{s - \theta }}\\ {}&{ \le {n^\alpha }{n^{(\theta - s)r}} = {n^{\alpha + (\theta - s)r}}.} \end{array}$ (2.11)

根据${\displaystyle\frac {\alpha}{r}}+\theta<\sigma\leq s\leq2, $$\alpha+(\theta-s)r<0, $所以

$I_3\simeq n^{\alpha+(\theta-s)r}\rightarrow 0, \, \, n\rightarrow\infty.$ (2.12)

情形2 $s>1. $

根据$EX_{ni}=0, E|X|^s<\infty, $可以得到

$\begin{array}{*{20}{l}} {\sum\limits_{i = 1}^\infty {{a_{ni}}} |E{X_{ni}}I(|{a_{ni}}{X_{ni}}| \le \delta )|}&{ = \sum\limits_{i = 1}^\infty {{a_{ni}}} |E{X_{ni}}I(|{a_{ni}}{X_{ni}}| > \delta )|}\\ {}&{ \simeq \sum\limits_{i = 1}^\infty {a_{ni}^s} E|X{|^s}I(|{a_{ni}}X| > \delta )}\\ {}&{ \simeq \sum\limits_{i = 1}^\infty {a_{ni}^s} \le {n^\alpha }{n^{(\theta - s)r}} = {n^{\alpha + (\theta - s)r}}.} \end{array}$ (2.13)

此时

$I_3\simeq n^{\alpha+(\theta-s)r}\rightarrow 0, \, \, n\rightarrow\infty.$ (2.14)

结合(2.12), (2.14) 式可以得到$I_3\rightarrow 0.$

因此, 根据(2.6) 至(2.14) 式可以满足定理1.7中的(ⅰ), (ⅱ), (ⅲ)三个条件, 所以我们可以得到对于任意$\epsilon >0$,有

$\sum\limits_{n = 1}^\infty {{n^\beta }} P\left( {\left| {\sum\limits_{i = 1}^\infty {{a_{ni}}} {X_{ni}}} \right| > \epsilon } \right) < \infty ,$

则表明(2.4) 式成立, 即成功的证明了定理(2.1).

由于吴群英[7]对于该问题的证明方法, 无法处理当$\alpha+\beta+1>0, \beta=-1$时的情况, 如果使用他的证明方法, 需要加强矩条件.而在本文中, 我们采用了不同于吴群英的方法, 可以处理$\beta=-1$时的情况.

邱德华[9]在研究行为NA随机变量列的完全收敛性时, 利用陈平炎等[10]研究的结果, 得到了行为NA随机变量列完全收敛性的充分条件, 受此启发, 这里我们利用李旭[11]建立的行为NOD随机变量列加权完全收敛性的充分条件, 验证充分条件的做法更为简便, Baek[12]和邱德华[9]的文章中研究了行为NA随机变量列加权和的完全收敛性, 且他们的证明方法也可以处理$\beta=-1$的情况, 本文将这个结果从行为NA随机变量列推广到行为NOD的随机变量列, 从而获得了NOD随机变量序列加权和的完全收敛性, 对吴群英[7]等的文章内容作了扩充, 且处理方法更为简便.

参考文献
[1] Joag-Dev K, Proschan F. Negative association of random variables with applications[J]. Ann. Stat., 1983, 11: 286–295. DOI:10.1214/aos/1176346079
[2] Hsu P L, Robbins H. Complete convergence and the law of large numbers[J]. Proc. National Acad.Sci. United States Amer., 1947, 33: 25–31. DOI:10.1073/pnas.33.2.25
[3] Erdos P. On a theorem of Hsu and Robbins[J]. Ann. Math. Stat., 1949, 20: 286–291. DOI:10.1214/aoms/1177730037
[4] Baum L E, Katz M. Convergence rates in the law of large numbers[J]. Trans. Amer. Math. Soc., 1965, 120: 108–123. DOI:10.1090/S0002-9947-1965-0198524-1
[5] Bozorgnia A, Patterson R F, Taylor R L. Limit theorems for ND r.v.'s[J]. Lith. Math. J., 1974, 14(4): 1639–1650.
[6] Asadian N, Fakoor V, Bozorgnia A. Rosenthal's type inequalities for negatively orthant dependent random variables[J]. J. Iranian Stat. Soc., 2006, 5: 66–75.
[7] Wu Q Y. Probability limit theory for mixed sequence[M]. Beijing: China Science Press, 2006.
[8] Wu Q Y. A complete convergence theorem for weighted sums of arrays of rowwise negatively dependent random variables[J]. J. Inequ. Appl., 2012: 50.
[9] 邱德华. 行为NA随机变量阵列的完全收敛性[J]. 数学杂志, 2013, 33: 138–146. DOI:10.3969/j.issn.0255-7797.2013.01.019
[10] Chen Pingyan, Hu T C, Liu X, Volodin A. On complete convergence for arrays of rowwise negatively associated random variables[J]. The. Prob. Appl., 2007, 52(2): 1–5.
[11] 李旭. NOD序列若干收敛性质的研究[D]. 安徽: 安徽大学硕士学位论文, 2012.
[12] Baek J I, Choi I B, Niu S l. On the complete convergence of weighted sums for arrays of negatively associated variables[J]. J. Korean Stat. Soc., 2008, 37: 73–80. DOI:10.1016/j.jkss.2007.08.001
[13] Guo M L, Zhu D J, Ren Y. Complete moment convergence of weighted sums for arrays of rowwise negatively associated random variables[J]. Chinese J. Appl. Prob. Stat. Soc., 2013, 29: 42–52.
[14] 谭成良, 吴群英, 何燕梅. 行为 ρ 混合阵列加权和的收敛性[J]. 数学杂志, 2011, 31(1): 178–184.