Message Board

Dear readers, authors and reviewers,you can add a message on this page. We will reply to you as soon as possible!

2019 Volume 44 Issue 3
Article Contents

Fan FENG, Zuo-xiang PENG. Higher-Order Expansions of Powered Order Statistics of Gaussian Sequences[J]. Journal of Southwest China Normal University(Natural Science Edition), 2019, 44(3): 18-22. doi: 10.13718/j.cnki.xsxb.2019.03.004
Citation: Fan FENG, Zuo-xiang PENG. Higher-Order Expansions of Powered Order Statistics of Gaussian Sequences[J]. Journal of Southwest China Normal University(Natural Science Edition), 2019, 44(3): 18-22. doi: 10.13718/j.cnki.xsxb.2019.03.004

Higher-Order Expansions of Powered Order Statistics of Gaussian Sequences

More Information
  • Corresponding author: Zuo-xiang PENG
  • Received Date: 17/09/2018
    Available Online: 20/03/2019
  • MSC: O211.4

  • In this paper, higher-order expansions on distributions and densities of powered order statistics formed by a sequence of independent identically Gaussian distributed random variables have been established under the optimal normalized constants.A byproduct is that rates of convergence of distributions and densities of powered statistics are the same order of $\frac{1}{{{\rm{log}}\;\mathit{n}}}$, respectively.
  • 加载中
  • [1] HALL P.On the Rate of Convergence of Normal Extremes[J].Journal of Applied Probability, 1979, 16(2):433-439. doi: 10.2307/3212912

    CrossRef Google Scholar

    [2] HALL P.Estimating Probabilities for Normal Extremes[J].Advances in Applied Probability, 1980, 12(2):491-500. doi: 10.2307/1426608

    CrossRef Google Scholar

    [3] ZHOU W, LING C X.Higher-Order Expansions of Powered Extremes of Normal Samples[J].Statistics and Probability Letters, 2016, 111:12-17. doi: 10.1016/j.spl.2016.01.003

    CrossRef Google Scholar

    [4] LIAO X, PENG Z X, NADARAJAH S, et al.Rates of Convergence of Extremes from Skew Normal Samples[J].Statistics and Probability Letters, 2012, 84(1):40-47.

    Google Scholar

    [5] LI C Q, LI T T.Density Expansions of Extremes from General Error Distribution with Applications[J].Journal of Inequalities and Applications, 2015, 2015(1):1-15.

    Google Scholar

    [6] 杨红艳, 彭作祥.有限混合偏正态极值的极限分布[J].西南大学学报(自然科学版), 2014, 36(3):59-62.

    Google Scholar

    [7] 章毓波, 彭作祥.t分布的极值分布渐近展开[J].西南大学学报(自然科学版), 2014, 36(3):67-70.

    Google Scholar

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Article Metrics

Article views(1039) PDF downloads(133) Cited by(0)

Access History

Other Articles By Authors

Higher-Order Expansions of Powered Order Statistics of Gaussian Sequences

    Corresponding author: Zuo-xiang PENG

Abstract: In this paper, higher-order expansions on distributions and densities of powered order statistics formed by a sequence of independent identically Gaussian distributed random variables have been established under the optimal normalized constants.A byproduct is that rates of convergence of distributions and densities of powered statistics are the same order of $\frac{1}{{{\rm{log}}\;\mathit{n}}}$, respectively.

  • X1X2,…为一列独立同分布于标准正态分布N(0,1)的随机变量. Φ(x)和ϕ(x)为标准正态分布函数和密度函数. Xn,1Xn,2≤…≤XnnX1X2,…,Xn的顺序统计量.文献[1]得到当anbn满足

    时,Φn(anx+bn)-Λ(x)收敛到0的一致收敛速度为$\frac{1}{{{\rm{log}}\;\mathit{n}}}$.进一步,文献[2]研究了高斯序列顺序统计量幂的渐近展开及收敛速度,得到对任意r≥1,t≥0,存在规范常数cn>0和dn$\mathbb{R}$,使得

    其中: Λr(x)=Λ(x)$\sum\nolimits_{j = 0}^{r-1} {\frac{{{{\rm{e}}^{-jx}}}}{{j!}}} $$x \in \mathbb{R}$规范常数cndn满足

    t=2时,亦可取如下的规范常数:

    bn由(1)式决定.文献[2]指出,当t=2且规范常数由(3)式给出时收敛速度最快.在此基础上,文献[3]研究了高斯序列最大值幂的分布函数与密度函数的高阶展开,得到与文献[2]一致的结论.

    文献[4-7]研究了其他给定分布序列的极值分布函数的渐近性质.

    本文旨在研究高斯序列顺序统计量幂|Xnr|t的分布函数与密度函数的高阶展开,并试图从中找出相应的收敛速度.

    定理1  对任意给定常数1≤rnt≥0及充分大的n,当t>0时,规范常数cndn取自(2)式,则

    t=2时,规范常数cndn取自(3)式,则

    推论1  由定理1有t>0时,P(|Xnn-r+1|tcnx+dn)收敛到$\mathit{\Lambda }\left( x \right)\sum\nolimits_{j = 0}^{r-1} {\frac{{{{\rm{e}}^{-jx}}}}{{j!}}} $的速度与$\frac{1}{{{\rm{log}}\;\mathit{n}}}$同阶;t=2时,P(|Xnn-r+1|2cnx+dn)收敛到$\mathit{\Lambda }\left( x \right)\sum\nolimits_{j = 0}^{r-1} {\frac{{{{\rm{e}}^{-jx}}}}{{j!}}} $的速度与$\frac{1}{{{{\left( {{\rm{log}}\;\mathit{n}} \right)}^2}}}$同阶.

    定理2  令fn(x)为P(|Xnn-r+1|tcnx+dn)的密度函数,当t>0时,规范常数cndn取自(2)式,对充分大的n,有

    t=2时,规范常数cndn取自(3)式,则对于Λ′(x)=e-xΛ(x),x$\mathbb{R}$

    推论2  由定理2知,当t>0时,fn(x)收敛到$\mathit{\Lambda '}\left( x \right)\frac{{{{\rm{e}}^{-\left( {r-1} \right)x}}}}{{\left( {r-1} \right)!}}$的速度与$\frac{1}{{{\rm{log}}\;\mathit{n}}}$同阶;当t=2时,fn(x)收敛到$\mathit{\Lambda '}\left( x \right)\frac{{{{\rm{e}}^{-\left( {r-1} \right)x}}}}{{\left( {r-1} \right)!}}$的速度与$\frac{1}{{{{\left( {{\rm{log}}\;\mathit{n}} \right)}^2}}}$同阶.

    为使得符号简化,当t>0,规范常数cndn取自(2)式时,令(cnx+dn)$\frac{1}{t}$=zn;当t=2,规范常数cndn取自(3)式时,令(cnx+dn)$\frac{1}{2}$=wn.

    引理1  对于自然数k及充分大的n

    (ⅰ) t≠2时,

    (ⅱ) t=2时,

    证明  由文献[3]中的(3.5)式有当t>0时,

    则对(n(1-Φ(zn)))k做泰勒展开即可证得结论. t=2时,同理可得

    引理证毕.

    引理2  对于自然数l及充分大的nΦ-l(zn)=1+O(n-1),Φ-l(wn)=1+O(n-1).

    证明  由(5)式可得

    同理可得Φ-l(wn)=1+O(n-1).

    定理1的证明  使用类似于文献[2]的方法有当t>0时,

    由(4)式知,

    由文献[3]中的(3.6)式有

    则由(7)式和(8)式知

    t=2时,由(5)式得

    故由(7)式知

    定理证毕.

    定理2的证明  令fn(x)=hn(x)+gn(x),其中

    首先考虑t>0时,

    由文献[3]之引理3.3知,t>0时,

    由(4)式,(9)式知此时

    t=2时,有

    故由(5)式及(11)式知,此时

Reference (7)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return