节点文献

几个新的重尾族上随机变量和的大偏差

Large Deviations for Sums of Heavy-tailed Random Variables of Several New Classes

【作者】 郭晓燕

【导师】 孔繁超;

【作者基本信息】 安徽大学 , 概率论与数理统计, 2005, 硕士

【摘要】 由于重尾分布族在应用概率领域中的广泛应用,人们对其的研究已经有多年的历史,重尾分布族的大偏差问题更是得到了众多学者的深入研究,Klüppelberg(1997)(文献[1])得到了如下的大偏差结果: 假设F∈ERV(-α,-β),对某(1<α≤β<∞),更新计数过程{N(t),t≥0}满足 N(t)/λ(t)(?)1(t→∞) (0-1) 且对任何固定的δ>0和某一个小的∈=∈(δ)>0使得: E(1+∈)N(t)IN(t)>(1+δ)λ(t)→(t→∞) (0-2)则对任意固定的γ>0,有i.e. P(S(t)-μ(t)>x)~λ(t)(?)(x),(t→∞)对于x≥γλ(t)一致成立。 文献[2]推广了上面的结论,指出当F∈C时,若(0-1),(0-2)成立,则同样的结果也成立,其中C族是一个比ERV族更大的重尾了类。 本义定义了两个新的重尾了族G和E,其中新类G包括了C族,Pareto分布族,Lognormal分布族等,E族包括G族及Weibull分布等,并得到了如下的结果: 1.假设F∈G,则有

【Abstract】 Since heavy-tailed distribution class has important function toward application probability fields , people have researched it for many years,many scholars deeply studied the large deviations for heavy-tailed random variables. Kluppelberg(1997) got large deviation result as follows:Suppose F ∈ ERV(-α,-β),(1 < α < β < ∞), if (0-1),(0-2) are satisfied , Then for any fixed γ> 0i.e. P(S(t) - μ(t) > x) ~ X(t)F(x), as t →∞ holds uniformly for x ≥γλ(t).It is shown in [2] that , suppose F ∈ C ,which is a bigger class than ERV, assume (0-1) and (0-2) are satisfied , the same result is also right .In the present paper we defined two new classes which named G and E . G includes class C ,Pareto distribution and Lognormal distribution etc., E includes G and Weibull distribution etc.,and got the following results:1.Suppose F ∈ G , then for any fixed γ> 0 ,P(S_n - ES_n > x) ~ nF(x), as n →∞holds uniformly for 2.Suppose F ∈ G , if (0-1) ,(0-2) are satisfied ,then for any fixed γ> 0i.e. P(S(t) - μ(t) > x) ~ X(t)F(x), as t→∞ holds uniformly for 3.Suppose F G E.there exists 0, 0 < (3 < l,then for any fixed 7 > 0 ,P{Sn - ESn > x)lim supnF{x)-1= 0i.e.P{Sn-ESn > x) n-s-ooholds uniformly for x@ >-fn .4.Suppose F E E , if (0-l),(0-2)are satisfied,then there exists /?, 0 < /? < l,for any fixed 7 > 0,we havelim supt->ooP(S(t) - is(t) > x)\(t)F(x)1= 0i.e.P(S(t) - fi(t) > X) ~ \{t)F{x) (t -4 OO)holds uniformly for x@ > jX(t).

  • 【网络出版投稿人】 安徽大学
  • 【网络出版年期】2006年 02期
  • 【分类号】O211.5
  • 【被引频次】6
  • 【下载频次】104
节点文献中: