数学物理学报  2015, Vol. 35 Issue (4): 756-768   PDF (349KB)    
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
邱德华1
陈平炎2
END随机变量序列产生的移动平均过程的收敛性
邱德华1, 陈平炎2    
1 广东财经大学数学与统计学院, 广州 510320;
2 暨南大学数学系, 广州 510630
摘要: 设{Yn,-∞< n< +∞}是双向无穷的END随机变量序列(不必同分布), {an,-∞< n< +∞}是绝对可和的实常数序列, 该文利用END列的Rademacher-Menshov型矩不等式, 得到了移动平均过程Xn=$\sum\limits_{n = - \infty }^\infty $ai Yi+n,n≥部分和的最大值的完全收敛性和矩完全收敛性. 所得结果推广和改进了已知的相应的一些结果.
关键词: 完全收敛     矩完全收敛     END随机变量     移动平均过程    
large Convergence for Moving Average Processes Under END Set-up
Qiu Dehua1, Chen Pingyan2     
1 School of Mathematics and Statistics, Guangdong University of Finance and Economics, Guangzhou 510320;
2 Department of Mathematics, Jinan University, Guangzhou 510630
Abstract: Let {Yn,-∞< n< +∞} be a doubly infinite sequence of non-identically distributed extended negatively dependent (END) random variables, {an,-∞< n< +∞} an absolutely summable sequence of real numbers. Utilizing the Rademacher-Menshov's inequality of END random variables, the complete convergence and complete moment convergence of the maximal partial sums of moving average processes Xn=$\sum\limits_{n = - \infty }^\infty $,ai Yi+n,n≥ are obtained, the corresponding results in series of previous papers are enriched and extended.
Key words: Complete convergence     Complete moment convergence     END random variable     Moving average process    

1 引言

设$\{Y_n,-\infty< n< +\infty\}$是定义在同一概率空间 $\{\Omega,\Im,P \}$的双边无限随机变量序列, $\{a_n,-\infty< n< +\infty \}$是绝对可和的实常数序列,即 $\sum\limits_{n=-\infty}^\infty |a_n|< \infty$,令

\begin{equation}\label{eq:1.1} X_n=\sum _{i=-\infty}^\infty a_i Y_{i+n},\ n\ge 1, \end{equation} (1.1)
称$\{X_n,n\ge 1\}$ 为由$\{Y_n,-\infty< n< +\infty\}$产生的移动平均过程.

许多学者对移动平均过程$\{X_n,n\ge 1\}$ 的极限理论进行了研究. 当$\{Y_n,-\infty< n< +\infty\}$是独立同分布的随机变量序列时, 关于$\{X_n,n\ge 1\}$的极限理论已有许多结果, 如Ibragimov[1]得到了中心极限定理, Burton和Dehling[2] 在条件$E\exp(tY_1)< \infty$下得到了大偏差原理, Li等[3]则研究了完全收敛性, Chen和Wang[4]研究了移动平均过程中偏差的收敛速度和加细收敛速度, 陈平炎和管总平[5]研究了阵列移动平均和的对数律成立的充要条件, 等等.当$\{Y_n,-\infty< n< \infty\}$ 是相依的随机变量序列时, 关于$\{X_n,n\ge 1\}$的极限理论已有一些结果,如李云霞与李坚高[6] 研究了由$\rho$混合随机过程序列产生的移动过程的弱收敛性. 对同分布的$\varphi$混合随机变量序列产生的移动平均过程, Zhang[7]得到了完全收敛性, 推广和改进了Li等[3]的结果, Chen等[8]把关于同分布的$\varphi$混 合随机变量序列的完全收敛性的结果推广到移动平均过程, 得到了极大值的完全收敛性,改进了Zhang[7] 的结果,李炜与甘师信[9]用积分检验法刻画了 重尾随机变量序列滑动平均过程的极限性质并获得了Chover型重对数律,等等.

Chow[10]首先研究了独立同分布的随机变量序列部分和的矩完全收敛性, 它是完全收敛的深化. Li和Zhang[11]在较强的矩条件下研究了同分布的NA随机变量序 列产生的移动平均过程矩完全收敛性, Kim和Ko[12]则讨论了同分布的$\varphi$混合随机变量序列产 生的移动平均过程矩完全收敛性,结果与Li和Zhang[11]相同, Zhou[13]改进了Kim 和Ko[12]的结果, Chen等[14]则在最优的矩条件下得到了同分布的NA随机变量 序列产生的移动平均过程矩完全收敛性更好的结果, 陈平炎和李远梅[15]还得到了由鞅差序列产生的移动平均过程 的完全收敛性、矩完全收敛性等结果.受文献[16]的启发,本文利用END随机变量 序列的Rademacher-Menshov型矩不等式,研究了不同分布的END序列(定义见后) 产生的移动过程部分和的最大值的完全收敛性和矩完全收敛性, 推广和改进了Li等[3]、Chow[10]、 Li和Zhang[11]、 Chen等[14]等相应的结果.

定义1.1 称随机变量序列$\{X_n,n\ge 1\}$是END的,若存在常数$M>0$, 对$\forall~n\ge 1,x_1,x_2,\cdots,x_n\in {\Bbb R}$,都有下二式成立 $$ P(X_1>x_1,X_2>x_2,\cdots,X_n>x_n)\le M\prod_{j=1}^n P(X_j>x_j),$$ $$P(X_1\le x_1,X_2\le x_2,\cdots,X_n\le x_n)\le M\prod_{j=1}^n P(X_j\le x_j).$$

END这一概念是Liu[17]提出的, 以NOD (negatively orthant dependent)[18]为特殊情形($M=1$). Liu[17]通过例题指出END是既包含负相依结构又包含 正相依结构的非常广泛的概念. Joag-Dev与Proschan[18]指出 NA (negatively associated)列是NOD列,但反之不成立,因此,NA列也是END列. Liu[17]获得了具有重尾的END随机变量的精确大偏差, Liu[19]研究了END列中偏差的充要条件, Shen[20]获得了END列的Rosenthal型不等式并利用它研究了 END列的渐进逼近问题,Qiu等[21] 在不同条件下获得了END阵列加权和的完全收敛性定理, Wu和Guan[22]研究了END列弱大数律及$L_p$收敛性与完全收敛性, Wang等[23]研究了END列与END阵列的完全收敛性,等等.

定义1.2 称随机变量序列$\{Y_j,-\infty< j< +\infty\}$ 被随机变量$Y$所控制,若存在常数$D>0$,使 $$ P(|Y_j|>x)\le D P(|Y|>x),\forall\,x>0,\forall -\infty< j< +\infty. $$ 此时简记为$\{Y_j,-\infty< j< +\infty\}\prec X$.

为证本文结果,需要如下引理.

引理1.1 [17] 设$\{Y_n,n\ge 1\}$是END随机变量序列,$\{f_n,n\ge 1\}$同为单调递增(或同为单调递减)的函数列, 则 $\{f_n(Y_n),n\ge 1\}$ 仍是END随机变量序列.

下面引理2中的(i)可见文献[20]. 由(i)及类似文献[24,定理2.3.1]的证明可得(ii). (ii)即为所谓的Rademacher-Menshov型矩不等式.

引理1.2 设$v\ge 2,\{Y_n,n\ge 1\}$是均值为零的END随机变量序 列且$E|Y_n|^v< \infty,n\ge 1$.则存在只依赖于$v$的正常数$C_v$,使

(i) $E\left|\sum\limits_{i=j+1}^{j+n}Y_i\right |^v \le C_v\left\{\sum\limits_{i=j+1}^{j+n} E|Y_i|^v+ \left(\sum\limits_{i=j+1}^{j+n }E|Y_i|^2\right)^{v/2}\right\},$

(ii) $E\max\limits_{1\le k\le n}\left|\sum\limits_{i=j+1}^{j+k}Y_i\right |^v \le C_v\left(\log n\right)^v \left\{\sum\limits_{i=j+1}^{j+n } E|Y_i|^v+ \left(\sum\limits_{i=j+1}^{j+n }E|Y_i|^2\right)^{v/2}\right\}.$

引理1.3 设$\{Y_n,-\infty< n< +\infty\}\prec Y$,则对$\forall\,v>0,x>0,-\infty< n< +\infty,$有

(1) $E|Y_n|^v I(|Y_n|\le x)\le C\left\{E|Y|^v I(|Y|\le x)+x^v P(|Y|>x)\right\};$

(2) $E|Y_n|^vI(|Y_n|> x)\le C E|Y|^v I(|Y|> x).$

引理1.4 $Y$为一随机变量,$\alpha,\beta,p,D$是正常数,$\psi(x)=1$或$\psi(x)=\log (x),x\ge 1$,若$E|Y|^p\psi(1+|Y|) < \infty$,则

(1) $ \sum\limits_{n=1}^\infty n^{\alpha p-1}\psi(n)P(|Y|>n^\alpha)\le CE|Y|^p\psi(1+|Y|). $

(2)若$p< \beta,$则 $\sum\limits_{n=1}^\infty n^{\alpha ( p-\beta)-1}\psi(n)E|Y|^\beta I(|Y|\le Dn^\alpha)\le CE|Y|^p\psi(1+|Y|).$

(3) 若$p>\beta,$则$\sum\limits_{n=1}^\infty n^{\alpha ( p-\beta)-1}\psi(n)E|Y|^\beta I(|Y|> n^\alpha)\le CE|Y|^p\psi(1+|Y|).$

(4) 若$E\left\{|Y|^p\log(1+|Y|)\psi(1+|Y|)\right\}< \infty,$ 则 $$\sum\limits_{n=1}^\infty n^{-1}\psi(n)E|Y|^p I(|Y|> n^\alpha)\le CE\left\{|Y|^p\log(1+|Y|)\psi(1+|Y|)\right\}.$$

类似于文献[14,引理1]可证.

本文以下总用$C$代表与$n$ 无关的正常数,在不同的地方可表示不同的值,即使在同一式中也是如此.

2 主要结果和证明

定理2.1 设$p> 1,\alpha>1/2,\alpha p>1,\{Y_n,-\infty< n< +\infty\}$是 双边无限的END随机变量序列且$\{Y_n,-\infty< n< +\infty\}\prec Y$. 当$\alpha \le 1$时,进一步假设$EY_n=0,-\infty< n< +\infty$. $\{a_n,-\infty< n< +\infty \}$是绝对可和的实常数序列,$\{X_n,n\ge 1\}$是形如 (1.1)式定义的移动平均过程. 记$S_n=\sum\limits_{k=1}^n X_k$. 如果

\begin{equation}\label{eq:2.1} E|Y|^p< \infty. \end{equation} (2.1)
\begin{equation}\label{eq:2.2} \sum\limits_{n=1}^\infty n^{\alpha p-2}P \Big(\max_{1\le k\le n}|S_k|>\epsilon n^\alpha\Big)< \infty,\forall\,\epsilon>0. \end{equation} (2.2)

由(2.1)式和$\sum\limits_{i=-\infty}^\infty |a_i|< \infty$可知 $$ \sum\limits_{i=-\infty}^\infty E|a_i Y_{i+n}|\le \sup_{i\in Z}E|Y_{i+n}|\left(\sum\limits_{i=-\infty}^\infty |a_i|\right) < \infty,\forall\,n\ge 1, $$ 从而$X_n(n\ge 1)$是a.s.有意义. 取$q$使$1/(\alpha p)< q< 1$. 对$n\ge 1,-\infty< i< +\infty,$ 令 $$ Y_{i}^{(n,1)}=-n^{\alpha q}I(Y_i< -n^{\alpha q})+Y_iI(|Y_i|\le n^{\alpha q})+n^{\alpha q}I(Y_i>n^{\alpha q}), $$ $$ Y_{i}^{(n,2)}=(Y_i-n^{\alpha q})I(n^{\alpha q}< Y_i\le n^\alpha+ n^{\alpha q})+n^\alpha I(Y_i>n^\alpha+n^{\alpha q}), $$ $$ Y_{i}^{(n,3)}=(Y_i-n^{\alpha q}-n^\alpha )I(Y_i> n^\alpha+n^{\alpha q}), $$ $$ Y_{i}^{(n,4)}=(Y_i+n^{\alpha q})I(-n^\alpha-n^{\alpha q}\le Y_i< -n^{\alpha q})-n^\alpha I(Y_i< -n^\alpha-n^{\alpha q}), $$ $$ Y_{i}^{(n,5)}=(Y_i+n^{\alpha q}+n^\alpha )I(Y_i< -n^\alpha-n^{\alpha q}). $$ 因此,对任意正整数$m$,有 $$ S_n=\sum\limits_{k=1}^n X_k=\sum\limits_{k=1}^n\sum\limits_{i=-\infty}^\infty a_iY_{i+k}=\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+n}Y_j=\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+n}\sum\limits_{l=1}^5 Y_j^{(m,l)}, $$ 则

\begin{eqnarray}\label{eq:2.3} &&\sum\limits_{n=1}^\infty n^{\alpha p-2}P\Big(\max_{1\le k\le n}|S_k| >\epsilon n^\alpha\Big) \nonumber\\ & =&\sum\limits_{n=1}^\infty n^{\alpha p-2}P \left(\max_{1\le k\le n}\left|\sum\limits_{i=-\infty}^\infty a_i \sum\limits_{j=i+1}^{i+k}\sum\limits_{l=1}^5 Y_j^{(n,l)}\right|>\epsilon n^\alpha \right) \nonumber\\ & \le & \sum\limits_{n=1}^\infty n^{\alpha p-2}P\left(\max_{1\le k\le n}\left|\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}Y_j^{(n,1)}\right|>\epsilon n^\alpha /5\right) \nonumber\\ & & +\sum\limits_{l=2}^3 \sum\limits_{n=1}^\infty n^{\alpha p-2}P\left(\sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}Y_j^{(n,l)}>\epsilon n^\alpha/5 \right)\nonumber\\ & & + \sum\limits_{l=4}^5 \sum\limits_{n=1}^\infty n^{\alpha p-2}P\left(\sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}\left(-Y_j^{(n,l)}\right)>\epsilon n^\alpha/5 \right) \nonumber\\ & \stackrel{ def }{=} & I_1+I_2+I_3+I_4+I_5. \end{eqnarray} (2.3)
因此,要证(2.2)式,只需要证明$I_j< \infty,j=1,2,3,4,5.$先证明$I_1< \infty$,为此先证下式成立
\begin{equation}\label{eq:2.4} n^{-\alpha}\max_{1\le k\le n}\left|E\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}Y_j^{(n,1)}\right|\to 0,n\to \infty. \end{equation} (2.4)
由$q$取法知$\alpha pq>1.$ 因此,当$\alpha\le 1$时,由$EY_n=0,-\infty< n< +\infty$及$\sum\limits_{i=-\infty}^\infty |a_i|< \infty$和引理1.3及(2.1)式可得
\begin{eqnarray}\label{eq:2.5} && n^{-\alpha}\max_{1\le k\le n}\left|E\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}Y_j^{(n,1)}\right| \nonumber \\ &\le & n^{-\alpha}\sum\limits_{i=-\infty}^\infty |a_i|\max_{1\le k\le n} \left|\sum\limits_{j=i+1}^{i+k}EY_j^{(n,1)}\right| \nonumber \\ & \le & n^{-\alpha}\sum\limits_{i=-\infty}^\infty |a_i|\max_{1\le k\le n}\sum\limits_{j=i+1}^{i+k}\left\{E|Y_j|I(|Y_j|>n^{\alpha q})+n^{\alpha q}P(|Y_j|>n^{\alpha q})\right\}\nonumber\\ & \le & 2n^{-\alpha}\sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}E|Y_j| I(|Y_j|>n^{\alpha q})\le 2n^{-\alpha}\sum\limits_{i=-\infty}^\infty |a_i|\left\{CnE|Y|I(|Y|>n^{\alpha q})\right\}\nonumber\\ & \le & Cn^{1-\alpha}E |Y| I(|Y|>n^{\alpha q}) \le C n^{1-\alpha p q -\alpha(1-q)} E|Y|^p \to 0,n\to \infty. \end{eqnarray} (2.5)
当$\alpha>1$时,由引理1.3和(2.1)式有
\begin{eqnarray}\label{eq:2.6} && n^{-\alpha}\max_{1\le k\le n}\left|E\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}Y_j^{(n,1)}\right| \nonumber \\ & \le & n^{-\alpha} \sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}E\left|Y_j^{(n,1)}\right| \nonumber \\ & \le & n^{-\alpha}\sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}\left\{E|Y_j|I(|Y_j|\le n^{\alpha q}) +n^{\alpha q}P(|Y_j|>n^{\alpha q})\right\}\nonumber\\ & \le & n^{-\alpha}\sum\limits_{i=-\infty}^\infty |a_i|\left(Cn E|Y |\right) \le C n^{1-\alpha } E|Y|\to 0,n\to \infty. \end{eqnarray} (2.6)
因此(2.4)式成立.于是要证$I_1< \infty$,只需要证 $$ I_1^* := \sum\limits_{n=1}^\infty n^{\alpha p-2}P\left(\max_{1\le k\le n}\left|\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}\left(Y_j^{(n,1)}-EY_j^{(n,1)}\right)\right|>\epsilon n^\alpha /10\right) < \infty. $$ 由引理1.1知,对$\forall n\ge1,\{Y_j^{(n,1)}-EY_j^{(n,1)},-\infty< j< \infty\}$是END随机变量序列.对任意$v\ge 2$且$v>p$($v$待后确定),由Markov不等式,H"older不等式,$C_r$不等式,引理1.2、引理1.3及$\sum\limits_{i=-\infty}^\infty|a_i|< \infty$可得
\begin{eqnarray}\label{eq:2.7} I_1^* &\le & C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-2} E\left\{\max_{1\le k\le n}\left|\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}\left(Y_j^{(n,1)}-EY_j^{(n,1)}\right)\right|\right\}^v\nonumber\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-2}E\left\{\sum\limits_{i=-\infty}^\infty |a_i|\max_{1\le k\le n}\left|\sum\limits_{j=i+1}^{i+k}\left(Y_j^{(n,1)}-EY_j^{(n,1)}\right)\right|\right\}^v\nonumber\\ & = &C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-2}E\left\{\sum\limits_{i=-\infty}^\infty |a_i|^{1-1/v}\left(|a_i|^{1/v}\max_{1\le k\le n}\left|\sum\limits_{j=i+1}^{i+k}\left(Y_j^{(n,1)}-EY_j^{(n,1)}\right)\right|\right)\right\}^v \nonumber\\ &\le & C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-2}\left(\sum\limits_{i=-\infty}^\infty |a_i|\right)^{v-1}\left(\sum\limits_{i=-\infty}^\infty|a_i| E\max_{1\le k\le n}\left|\sum\limits_{j=i+1}^{i+k}\left(Y_j^{(n,1)}-EY_j^{(n,1)}\right)\right|^v \right) \nonumber\\ & \le &C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-2} \sum\limits_{i=-\infty}^\infty|a_i|(\log n)^v \left\{\sum\limits_{j=i+1}^{i+n}E|Y_j^{(n,1)}|^v+\left(\sum\limits_{j=i+1}^{i+n} E|Y_j^{(n,1)}|^2 \right)^{v/2}\right\}\nonumber\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-2}(\log n)^v \sum\limits_{i=-\infty}^\infty|a_i| \sum\limits_{j=i+1}^{i+n} \left\{E|Y_j|^vI(|Y_j|\le n^{\alpha q}) +n^{v\alpha q}P(|Y_j|>n^{\alpha q})\right\}\nonumber\\ & & +C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-2}(\log n)^v \nonumber\\ &&\times \sum\limits_{i=-\infty}^\infty|a_i|\left\{ \sum\limits_{j=i+1}^{i+n}\left[EY_j^2I(|Y_j|\le n^{\alpha q})+n^{2\alpha q}P(|Y_j|>n^{\alpha q})\right]\right \}^{v/2} \nonumber\\ &\le & C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-1}(\log n)^v \left\{E|Y|^vI(|Y|\le n^{\alpha q}) +n^{v\alpha q}P(|Y|>n^{\alpha q})\right\}\nonumber\\ & &+C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha+v/2-2}(\log n)^v \left\{EY^2I(|Y|\le n^{\alpha q})+n^{2\alpha q}P(|Y|>n^{\alpha q})\right\}^{v/2} \nonumber\\ &\stackrel{ def }{=} & I_{11}^{(v)}+I_{12}^{(v)}. \end{eqnarray} (2.7)
而由(2.1)式可得 $$ I_{11}^{(v)}\le C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha-1}(\log n)^v E\left\{|Y|^p n^{(v-p)\alpha q}\right\}\le C\sum\limits_{n=1}^\infty n^{\alpha (p-v)(1-q)-1}(\log n)^v < \infty. $$ 如果$p< 2$,令$v=2$,则$I_{12}^{(2)}=I_{11}^{(2)}< \infty.$ 如果$p\ge 2$,令$v>\max\{2,(\alpha p-1)/(\alpha-1/2)\}$,并注意到$E|Y|^2< \infty$,则$I_{12}^{(v)}\le C\sum\limits_{n=1}^\infty n^{\alpha p-v\alpha+v/2-2}(\log n)^v < \infty$. 由上所述可知$I_1< \infty.$ 再证$I_2< \infty$.由(2.1)式和引理1.3可得
\begin{eqnarray}\label{eq:2.8} 0 &\le & n^{-\alpha}E\sum\limits_{i=-\infty}^\infty | a_i|\sum\limits_{j=i+1}^{i+n}Y_j^{(n,2)}\le n^{-\alpha}\sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}EY_jI(Y_j>n^{\alpha q})\nonumber\\ & \le & Cn^{1-\alpha}E|Y| I(|Y|>n^{\alpha q})\le n^{1-\alpha pq -\alpha (1-q)} E|Y|^p\to 0,n\to\infty. \end{eqnarray} (2.8)
因此要证$I_2< \infty$,只需要证 $$ I_{2}^* = :\sum\limits_{n=1}^\infty n^{\alpha p-2}P\left( \sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n} \left( Y_j^{(n,2)}-EY_j^{(n,2)} \right)>\epsilon n^\alpha/10 \right)< \infty. $$ 由引理1.1知$\forall\,n\ge 1,\{Y_j^{(n,2)},-\infty< j< +\infty\}$是END随机变量, 对任意$v\ge 2$且$v>p$($v$待后确定),类似于(2.7)式可得
\begin{eqnarray}\label{eq:2.9} I_{2}^* &\le & C\sum\limits_{n=1}^\infty n^{\alpha p -v\alpha-2} E\left( \sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n} \left( Y_j^{(n,2)}-EY_j^{(n,2)} \right)\right)^v\nonumber\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha p -v\alpha-2}\sum\limits_{i=-\infty}^\infty|a_i| \left\{\sum\limits_{j=i+1}^{i+n}E|Y_j^{(n,2)}|^v+\left(\sum\limits_{j=i+1}^{i+n} E|Y_j^{(n,2)}|^2 \right)^{v/2}\right\}\nonumber\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha p -v\alpha-2}\sum\limits_{i=-\infty}^\infty|a_i| \sum\limits_{j=i+1}^{i+n} \left\{E|Y_j|^vI(|Y_j|\le 2n^{\alpha }) +n^{v\alpha }P(|Y_j|>n^{\alpha })\right\}\nonumber\\ & & +C\sum\limits_{n=1}^\infty n^{\alpha p -v\alpha-2} \sum\limits_{i=-\infty}^\infty|a_i|\left\{ \sum\limits_{j=i+1}^{i+n}\left[EY_j^2I(|Y_j|\le 2n^{\alpha })+n^{2\alpha }P(|Y_j|>n^{\alpha })\right]\right \}^{v/2} \nonumber\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha p -v\alpha-1} \left\{E|Y|^vI(|Y|\le 2 n^{\alpha }) +n^{v\alpha } P(|Y|>n^{\alpha })\right\}\nonumber\\ & & +C\sum\limits_{n=1}^\infty n^{\alpha p -v\alpha+v/2-2} \left\{EY^2I(|Y|\le 2n^{\alpha })+n^{2\alpha}P(|Y|>n^{\alpha })\right\}^{v/2}\nonumber\\ & \stackrel{ def }{=} & I_{21}^{(v)}+I_{22}^{(v)}. \end{eqnarray} (2.9)
由(2.1)式及引理1.4可得$I_{21}^{(v)}< \infty$. 如果$p< 2$,令$v=2$,则$I_{22}^{(2)}=I_{21}^{(2)}< \infty. $ 如果$p\ge 2$,令$v>\max\{2,(\alpha p-1)/(\alpha-1/2)\}$,并注意到$E|Y|^2< \infty$,则 $$ I_{22}^{(v)}\le C\sum\limits_{n=1}^\infty n^{\alpha p -v\alpha+v/2-2}(E|Y|^2)^{v/2} = C\sum\limits_{n=1}^\infty n^{\alpha p-v(\alpha-1/2)-2}< \infty. $$ 从而$I_2< \infty.$ 再证$I_3< \infty$.因$p>1$,故由Markov不等式,$\sum\limits_{i=-\infty}^\infty|a_i|< \infty$、引理1.3,引理1.4及(2.1)式可得 \begin{eqnarray*} I_{3}& \le & C\sum\limits_{n=1}^\infty n^{\alpha p-\alpha-2}E\left(\sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}Y_jI(Y_j> n^\alpha) \right)\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha( p-1)-1}E |Y |I(|Y|> n^\alpha)\le CE|Y|^p < \infty. \end{eqnarray*} 从而$I_3< \infty$.分别类似$I_2< \infty$与$I_3< \infty$的证明可证$I_4< \infty$与$I_5< \infty$,从而(2.2)式成立.

当$p=1$时,我们有

定理2.2 设$\alpha >1,\{Y_n,-\infty< n< +\infty\}$是双边无限的END随机变 量序列且$\{Y_n,-\infty< n< +\infty\}\prec Y,\{a_n,-\infty< n< +\infty \}$是绝对可和的实 常数序列,$\{X_n,n\ge 1\}$是形如(1.1)式定义的移动平均过程. 如 $$ E|Y|\log(1+|Y|)< \infty, $$ 则 $$ \sum\limits_{n=1}^\infty n^{\alpha -2}P \Big(\max_{1\le k\le n}|S_k|>\epsilon n^\alpha\Big)< \infty,\forall\,\epsilon>0. $$

与定理2.1证明方法一致,但此时$I_{3}\le CE|Y|\log(1+|Y|)< \infty$.

定理2.3 设$\gamma>0,p>1,\alpha>1/2,\alpha p>1,\{Y_n,-\infty< n< +\infty\}$是双边无限 的END随机变量序列且$\{Y_n,-\infty< n< +\infty\}\prec Y$,当$\alpha \le 1$时, 进一步假设$EY_n=0,-\infty< n< +\infty$. $\{a_n,-\infty< n< +\infty \}$是绝对可和的实常数序列, $\{X_n,n\ge 1\}$是形如(1.1)式定义的移动平均过程. 如果

\begin{equation}\label{eq:2.10} \left\{\begin{array}{ll} E|Y|^p< \infty,& \gamma< p,\\ E|Y|^p\log (1+|Y|)< \infty,& \gamma=p,\\ E|Y|^{\gamma}< \infty,& \gamma>p, \end{array}\right. \end{equation} (2.10)
\begin{equation}\label{eq:2.11} \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 } E \left \{\max_{1\le k \le n} |S_k|-\epsilon n^\alpha \right\}_+^\gamma< \infty,\forall\,\epsilon>0. \end{equation} (2.11)

由定理2.1的证明可知,对$\forall\,n\ge 1,X_n$有定义. 而对$\forall\,\epsilon>0$,有 \begin{eqnarray*} &&\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 } E \left \{\max_{1\le k \le n} |S_k|-\epsilon n^\alpha \right\}_+^\gamma \\ & = & \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 } \int_0^\infty P \Big(\max_{1\le k\le n}|S_k|-\epsilon n^\alpha>t^{1\gamma}\Big){\rm d}t \\ & = & \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 } \int_0^{n^{\gamma \alpha}} P \Big(\max_{1\le k\le n}|S_k|-\epsilon n^\alpha>t^{1/\gamma}\Big){\rm d}t \\ & & + \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 } \int_{n^{\gamma \alpha}}^\infty P \Big(\max_{1\le k\le n}|S_k|-\epsilon n^\alpha>t^{1/\gamma}\Big){\rm d}t \\ & \le & \sum\limits_{n=1}^\infty n^{\alpha p-2}P \Big(\max_{1\le k\le n}|S_k|>\epsilon n^\alpha\Big) +\sum\limits_{n=1}^\infty n^{\alpha(p-\gamma)-2}\int_{n^{\gamma \alpha}}^\infty P \Big(\max_{1\le k\le n}|S_k|>t^{1/\gamma}\Big){\rm d}t. \end{eqnarray*} 因此由定理2.1可知,要证(2.11)式,只需要证明 $$ \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty P\Big(\max_{1\le k\le n}|S_k|>t^{1/\gamma}\Big){\rm d}t< \infty . $$ 取$q,$使$1/(\alpha p)< q< 1$.对$\forall t>0,-\infty< j< +\infty$,令 $$ Y_j^{(t,1)}=-t^{q/\gamma}I(Y_j< -t^{q/\gamma})+Y_j I(|Y_j|\le t^{q/\gamma})+t^{q/\gamma}I(Y_j>t^{q/\gamma}), $$ $$ Y_j^{(t,2)}=(Y_j-t^{q/\gamma})I(t^{q/\gamma}< Y_j\le t^{1/\gamma}+t^{q/\gamma})+t^{1/\gamma}I(Y_j>t^{1/\gamma}+t^{q/\gamma}), $$ $$ Y_j^{(t,3)}=(Y_j-t^{q/\gamma}-t^{1/\gamma})I(Y_j>t^{1/\gamma}+t^{q/\gamma}), $$ $$ Y_j^{(t,4)}=(Y_j+t^{q/\gamma})I(-t^{1/\gamma}-t^{q/\gamma}\le Y_j< -t^{q/\gamma})-t^{1/\gamma}I(Y_j< -t^{1\gamma}-t^{q/\gamma}), $$ $$ Y_j^{(t,5)}=(Y_j+t^{q/\gamma}+t^{1/\gamma})I(Y_j< -t^{1/\gamma}-t^{q/\gamma}). $$ 于是 $Y_j=\sum\limits_{l=1}^5 Y_j^{(t,l)}$.类似于(2.3)式可得 \begin{eqnarray*} &&\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty P\Big(\max_{1\le k\le n}|S_k|>t^{1/\gamma}\Big){\rm d}t \\ & \le & \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty P\left(\max_{1\le k\le n}\left|\sum\limits_{i=-\infty}^\infty a_i \sum\limits_{j=i+1}^{i+k}Y_j^{(t,1)}\right|>t^{1/\gamma}/5\right){\rm d}t \\ & &+\sum\limits_{l=2}^3 \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty P\left(\sum\limits_{i=-\infty}^\infty |a_i| \sum\limits_{j=i+1}^{i+n}Y_j^{(t,l)}>t^{1/\gamma}/5\right){\rm d}t \\ & &+\sum\limits_{l=4}^5\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty P\left(-\sum\limits_{i=-\infty}^\infty |a_i| \sum\limits_{j=i+1}^{i+n}Y_j^{(t,l)}>t^{1/\gamma}/5\right){\rm d}t\\ & \stackrel{def}{=}& J_1+J_2+J_3+J_4+J_5. \end{eqnarray*} 于是要证(2.11)式,只要证$J_i< \infty,i=1,2,3,4,5.$ 先证$J_1< \infty$. 为此,先证

\begin{equation}\label{eq:2.12} \sup_{t\ge n^{\gamma \alpha}}t^{-1/\gamma}\max_{1\le k\le n}\left|E\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}Y_j^{(t,1)}\right|\to 0,\,n\to \infty. \end{equation} (2.12)
当$\alpha\le 1$时,由$EY_j=0,-\infty< j< +\infty,$类似于(2.5)式的证明有 \begin{eqnarray*} &&\sup_{t\ge n^{\gamma \alpha}}t^{-1/\gamma}\max_{1\le k\le n} \left|E\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}Y_j^{(t,1)}\right| \\ & \le & Cn\sup_{t\ge n^{\gamma \alpha}}t^{-1/\gamma}E|Y|I(|Y|> t^{q/\gamma}) \le Cn^{1-\alpha}E|Y|I(|Y|> n^{\alpha q})\to 0,\,n\to \infty. \end{eqnarray*} 当$\alpha>1$时,类似(2.6)式的证明可得,当$n\to \infty$时 $$ \sup_{t\ge n^{\gamma \alpha}}t^{-1/\gamma}\max_{1\le k\le n}\left|E\sum\limits_{i=-\infty}^\infty a_i\sum\limits_{j=i+1}^{i+k}Y_j^{(t,1)}\right| \le Cn \sup_{t\ge n^{\gamma \alpha}}t^{-1/\gamma}E|Y| \le Cn^{1-\alpha}E|Y|\to 0. $$ 因此(2.12)式成立. 于是要证明$I_1< \infty$,只需要证 $$ I_1^*:=\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty P\left(\max_{1\le k\le n}\left|\sum\limits_{i=-\infty}^\infty a_i \sum\limits_{j=i+1}^{i+k}\left(Y_j^{(t,1)}-EY_j^{(t,1)}\right)\right|>t^{1/\gamma}/10\right){\rm d}t< \infty. $$ 由引理1.1知,对$\forall\,t>0,\{Y_j^{(t,1)},-\infty< j< +\infty\}$是END随机变量序列, 对任意$v\ge 2 $且$v>\max\{p,\gamma\}$ ($v$待后确定),类似于(2.7)式的证明可得 \begin{eqnarray*} J_1^* &\le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma}E\left\{\sum\limits_{i=-\infty}^\infty |a_i|\max_{1\le k\le n}\left|\sum\limits_{j=i+1}^{i+k}\left(Y_j^{(t,1)}-EY_j^{(t,1)}\right)\right|\right\}^v {\rm d}t\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }(\log n)^v \int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma} \left\{E|Y|^vI(|Y|\le t^{ q/\gamma}) +t^{vq/\gamma}P(|Y|>n^{q/\gamma})\right\}{\rm d}t\\ & & + C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2+v/2 }(\log n)^v \\ &&\times \int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma}\left\{EY^2I(|Y|\le t^{ q/\gamma})+t^{2q/\gamma}P(|Y|>n^{q/\gamma})\right \}^{v/2}{\rm d}t\\ & \stackrel{def}{=} &J_{11}^{(v)}+J_{12}^{(v)}. \end{eqnarray*} 对$J_{11}^{(v)}$. 若$\max\{p,\gamma\}< 2$取$v=2$, 则由(2.10)式得$E|Y|^{\max\{p,\gamma\}}< \infty,$ 因此 \begin{eqnarray*} J_{12}^{(2)}=J_{11}^{(2)} & \le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }(\log n)^2 \int_{n^{\gamma \alpha}}^\infty t^{-[2-(2-\max\{p,\gamma\})q]/\gamma} E|Y|^{\max\{p,\gamma\}}{\rm d}t\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha [(p-2)-(\max\{p,\gamma\}-2)q]-1 } (\log n)^2 \\ &\le & C\sum\limits_{n=1}^\infty n^{\alpha (p-2)(1-q)-1 }(\log n)^2 < \infty. \end{eqnarray*} 若$\max\{p,\gamma\}\ge 2$,由(2.10)式得$E|Y|^2< \infty$. 取$v>\max\{p/(1-q),\gamma/(1-q),(\alpha p-1)/(\alpha-1/2)\}$,则有 $$ J_{11}^{(v)} \le C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }(\log n)^v \int_{n^{\gamma \alpha}}^\infty t^{-(v-vq)/\gamma}{\rm d}t \le \sum\limits_{n=1}^\infty n^{-v \alpha (1-q)+\alpha p-1 }(\log n)^v < \infty, $$ $$ J_{12}^{(v)} \le C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2+v/2 }(\log n)^v\int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma}{\rm d}t \le C\sum\limits_{n=1}^\infty n^{\alpha p-2-(\alpha-1/2)v }(\log n)^v< \infty. $$ 因此 $J_1^*< \infty$,从而$J_1< \infty.$下证$J_2< \infty,$类似于(2.8)式有 \begin{eqnarray*} 0 &\le & \sup_{t\ge n^{\gamma \alpha}}\left\{t^{-1/\gamma} \sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+n}EY_j^{(t,2)}\right\} \\ &\le& Cn\sup_{t\ge n^{\gamma \alpha}}t^{-1/\gamma}E|Y| I(|Y|>t^{q/\gamma}) \le Cn^{1-\alpha}E|Y| I(|Y|>n^{\alpha q}) \ \to 0,\,n\to \infty. \end{eqnarray*} 因此要证$J_{2}< \infty$,只需要证明 $$ J_{2}^* := \sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty P\left(\sum\limits_{i=-\infty}^\infty |a_i| \sum\limits_{j=i+1}^{i+n}\left(Y_j^{(t,2)}-EY_j^{(t,2)}\right)>t^{1/\gamma}/10\right){\rm d}t < \infty. $$ 由引理1.1知,对$\forall\,t>0,\{Y_j^{(t,2)},-\infty< j< +\infty\}$是END列. 对任意$v\ge 2 $且$v>\max\{p,\gamma\}$ ($v$待后确定),类似于(2.9)式的证明可得 \begin{eqnarray*} J_{2}^* & \le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma}E\left\{\sum\limits_{i=-\infty}^\infty |a_i|\sum\limits_{j=i+1}^{i+k}\left(Y_j^{(t,2)}-EY_j^{(t,2)}\right)\right\}^v{\rm d}t\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }\int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma} \left\{ E|Y|^vI(|Y|< 2t^{1/\gamma})+ t^{v/\gamma}P(|Y|>t^{1/\gamma})\right\}{\rm d}t\\ & & + C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2+v/2 }\int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma} \left\{ E|Y|^2I(|Y|< 2t^{1/\gamma})+t^{-2/\gamma} P(|Y|>t^{1/\gamma}) \right\}^{v/2}{\rm d}t\\ &\stackrel{def}{=} &J_{21}^{(v)}+J_{22}^{(v)}. \end{eqnarray*} 由中值定理及(2.10)式和引理1.4有 \begin{eqnarray*} J_{21}^{(v)} & = & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }\sum\limits_{j=n}^\infty \int_{j^{\gamma \alpha}}^{(j+1)^{\gamma \alpha}}\left\{ t^{-v/\gamma} E|Y|^vI(|Y|< 2t^{1/\gamma})+P(|Y|>t^{1/\gamma})\right\}{\rm d}t\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }\sum\limits_{j=n}^\infty j^{\gamma\alpha -1}\left\{j^{-v\alpha} E|Y|^vI(|Y|< 2(j+1)^{\alpha})+P(|Y|>j^\alpha )\right\}\\ & = & C\sum\limits_{j=1}^\infty j^{\gamma\alpha -1}\left\{j^{-v\alpha} E|Y|^vI(|Y|< 2(j+1)^{\alpha})+P(|Y|>j^\alpha )\right\}\sum\limits_{n=1}^j n^{\alpha ( p-\gamma)-1 }\\ & \le & \left\{\begin{array}{ll} C\sum\limits_{j=1}^\infty \left\{j^{\alpha(p-v)-1} E|Y|^vI(|Y|< 2(j+1)^{\alpha})+j^{\alpha p -1} P(|Y|>j^\alpha )\right\},& \gamma< p\\ [3mm] C\sum\limits_{j=1}^\infty \left\{ j^{\alpha (p-v)-1} \log j E|Y|^vI(|Y|< 2(j+1)^{\alpha})+ j^{\alpha p-1} \log j P(|Y|>j^\alpha )\right\},& \gamma=p\\ [3mm] C\sum\limits_{j=1}^\infty \left\{j^{\alpha (\gamma-v) -1} E|Y|^vI(|Y|< 2(j+1)^{\alpha})+j^{\gamma \alpha -1} P(|Y|>j^\alpha )\right\} ,& \gamma>p\\ \end{array}\right.\\ & \le & \left\{\begin{array}{ll} CE|Y|^p,& \gamma< p\\ CE|Y|^p\log(1+|Y|) ,~~ & \gamma=p\\ CE|Y|^\gamma ,& \gamma>p \end{array}\right.\\ & < & \infty. \end{eqnarray*} 当 $\max\{p,\gamma\}< 2$,令$v=2$,则$ J_{22}^{(2)} = J_{21}^{(2)}< \infty$. 当$\max\{p,\gamma\}\ge 2$,令 $v>\max\{\gamma,(\alpha p -1)/(\alpha-1/2)\}$并注意到 $E|Y|^2< \infty$,我们有 $$ J_{22}^{(v)} \le C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2+v/2 }\int_{n^{\gamma \alpha}}^\infty t^{-v/\gamma}{\rm d}t = C\sum\limits_{n=1}^\infty n^{\alpha p-2-(\alpha-1/2)v}< \infty.\\ $$ 因此,$J_{2}^*< \infty$,故$J_{2}< \infty.$再证 $J_3< \infty $. 由Markov不等式、引理1.3、中值定理及(2.10)式和引理1.4可得
\begin{eqnarray}\label{eq:2.13} J_{3} & \le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-2 }\int_{n^{\gamma \alpha}}^\infty t^{-1/\gamma}\left(\sum\limits_{i=-\infty}^\infty |a_i| \sum\limits_{j=i+1}^{i+n} EY_jI(Y_j>t^{1/\gamma})\right){\rm d}t \nonumber\\ &\le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }\int_{n^{\gamma \alpha}}^\infty t^{-1/\gamma} E|Y|I(|Y|>t^{1/\gamma}){\rm d}t\nonumber\\ & = & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }\sum\limits_{j=n}^\infty \int_{j^{\gamma \alpha}}^{(j+1)^{\gamma \alpha}} t^{-1/\gamma}E|Y|I(|Y|>t^{1/\gamma}){\rm d}t\nonumber\\ & \le & C\sum\limits_{n=1}^\infty n^{\alpha ( p-\gamma)-1 }\sum\limits_{j=n}^\infty j^{-\alpha-1+\gamma \alpha} E|Y| I(|Y|>j^{\alpha}) \nonumber\\ & = & C\sum\limits_{j=1}^\infty j^{-\alpha-1+\gamma \alpha} E|Y| I(|Y|>j^{\alpha}) \sum\limits_{n=1}^j n^{\alpha ( p-\gamma)-1 }\nonumber\\ & \le & \left\{\begin{array}{ll} C\sum\limits_{j=1}^\infty j^{\alpha(p -1)-1} E|Y| I(|Y|>j^{\alpha}),& \gamma< p\\ C\sum\limits_{j=1}^\infty j^{\alpha ( p-1)-1} \log j E|Y |I(|Y|>j^{\alpha}) ,~~ & \gamma=p\\ \sum\limits_{j=1}^\infty j^{\alpha (\gamma -1)-1} E|Y| I(|Y|>j^{\alpha}) ,& \gamma>p\\ \end{array}\right.\nonumber\\ & \le & \left\{\begin{array}{ll} CE|Y|^p,& \gamma< p\\ CE|Y|^p\log(1+|Y|) ,~~ & \gamma=p\\ CE|Y|^\gamma ,& \gamma>p \end{array}\right.\nonumber\\ & < & \infty. \end{eqnarray} (2.13)
分别类似$J_2< \infty$与$J_3< \infty$的证明可证 $J_4< \infty$与$J_5< \infty$,因此(2.11)式成立.

当$p=1$时,我们有

定理2.4 设$\gamma>0,\alpha >1,\{Y_n,-\infty< n< +\infty\}$ 是双边无限的END随机变量序列且$\{Y_n,-\infty< n< +\infty\}\prec Y, \{a_n,-\infty< n< +\infty \}$是绝对可和的实常数序列,$\{X_n,n\ge 1\}$ 是形如(1.1)式定义的移动平均过程.如果

\begin{eqnarray}\label{eq:2.14} \left\{\begin{array}{ll} E|Y|\log(1+|Y|)< \infty,& \gamma< 1,\\ E|Y|\log^2 (1+|Y|)< \infty,~~& \gamma=1,\\ E|Y|^{\gamma}< \infty,& \gamma>1 . \end{array}\right. \end{eqnarray} (2.14)
\begin{equation}\label{eq:2.15} \sum\limits_{n=1}^\infty n^{\alpha ( 1-\gamma)-2 } E \left \{\max_{1\le k \le n} |S_k|-\epsilon n^\alpha \right\}_+^\gamma< \infty,\forall\,\epsilon>0. \end{equation} (2.15)

由定理2.3的证明可知要证(2.15)式,只要证明定理2.3中的$J_{3}< \infty$. 事实上,由(2.13)式的证明和(2.14)式与引理1.4有 \begin{eqnarray*} J_3 & \le & \left\{\begin{array}{ll} C\sum\limits_{j=1}^\infty j^{-1} E|Y| I(|Y|>j^{\alpha}),& \gamma< 1\\ C\sum\limits_{j=1}^\infty j^{-1} \log j E|Y| I(|Y|>j^{\alpha}) ,~~& \gamma=1\\ \sum\limits_{j=1}^\infty j^{\alpha (\gamma -1)-1} E|Y| I(|Y|>j^{\alpha}) ,& \gamma>1\\ \end{array}\right.\nonumber\\ & \le & \left\{\begin{array}{ll} CE|Y|\log(1+|Y|),& \gamma< 1\\ CE|Y|^p\log^2 (1+|Y|) ,~~& \gamma=1\\ CE|Y|^\gamma ,& \gamma>1 \end{array}\right.\\ & < & \infty. \end{eqnarray*} 证毕.

参考文献
[1] 张建文. 具强迫项非线性梁方程解的渐近性. 应用数学, 2001, 14: 60-66
[2] 陈明勇, 杨晗. 一类非线性四阶波动方程解的爆破. 西南师范大学学报, 2004, 29(4): 545-548
[3] Chen Y M, Yang H. A characterization of existence of global solutions for some fourth-order wave equations. Chin Quart, 2008, 23: 109-114
[4] 任永华, 张建文. 非自治强阻尼梁方程的渐近行为. 太原理工大学学报, 2013, 44: 116-118
[5] 廖秋明, 赵红星. 一类具有耗散项的非线性四阶波动方程的整体弱解及其渐近性质. 工程数学学报, 2013, 30: 59-65
[6] Cazenave T, Haraux A. An Introduction to Semilinear Evolution Equations. Oxford: Clarrendon Press, 1998
[7] Payne L E, Sattinger D H. Saddle points and instability of nonlinear hyperbolic equations. Israel J Math, 1975, 22: 273-303