Wu et al. (2009) studied the asymptotic approximation of inverse moments for nonnegative independent random variables. Shen et al. (2011) extended the result of Wu et al. (2009) to the case of ρ-mixing random variables. In the paper, we will further study the asymptotic approximation of inverse moments for nonnegative ρ-mixing random variables, which improves the corresponding results of Wu et al. (2009), Wang et al. (2010), and Shen et al. (2011) under the case of identical distribution.
1. Introduction
Firstly, we will recall the definition of ρ-mixing random variables.
Let {Xn,n≥1} be a sequence of random variables defined on a fixed probability space (Ω,ℱ,P). Let n and m be positive integers. Write ℱnm=σ(Xi,n≤i≤m) and ℱ𝒮=σ(Xi,i∈S⊂ℕ). Given σ-algebras ℬ,ℛ in ℱ, let
ρ(B,R)=supX∈L2(B),Y∈L2(R)|EXY-EXEY|Var(X)⋅Var(Y).
Define the ρ-mixing coefficients by ρ(n)=supk≥1ρ(F1k,Fk+n∞),n≥0.
Definition 1.1.
A sequence {Xn,n≥1} of random variables is said to be ρ-mixing if ρ(n)↓0 as n→∞.
ρ-mixing sequence was introduced by Kolmogorov and Rozanov [1]. It is easily seen that ρ-mixing sequence contains independent sequence as a special case.
The main purpose of the paper is to study the asymptotic approximation of inverse moments for nonnegative ρ-mixing random variables with identical distribution.
Let {Zn,n≥1} be a sequence of independent nonnegative random variables with finite second moments. Denote Xn=∑i=1nZiBn,Bn2=∑i=1nVarZi.
It is interesting to show that under suitable conditions the following equivalence relation holds, namely,
E(a+Xn)-r~(a+EXn)-r,n⟶∞,
where a>0 and r>0 are arbitrary real numbers.
Here and below, for two positive sequences {cn,n≥1} and {dn,n≥1}, we write cn~dn if cndn-1→1 as n→∞. C is a positive constant which can be different in various places.
The inverse moments can be applied in many practical applications. For example, they may be applied in Stein estimation and poststratification (see [2, 3]), evaluating risks of estimators and powers of tests (see [4, 5]). In addition, they also appear in the reliability (see [6]) and life testing (see [7]), insurance and financial mathematics (see [8]), complex systems (see [9]), and so on.
Under certain asymptotic-normality condition, relation (1.4) was established in Theorem 2.1 of Garcia and Palacios [10]. But, unfortunately, that theorem is not true under the suggested assumptions, as pointed out by Kaluszka and Okolewski [11]. The latter authors established (1.4) by modifying the assumptions as follows:
r<3 (r<4, in the i.i.d. case);
EXn→∞, EZn3<∞;
(Lc condition) ∑i=1nE|Zi-EZi|c/Bnc→0 (c=3).
Hu et al. [12] considered weaker conditions: EZn2+δ<∞, where Zn satisfies L2+δ condition and 0<δ≤1. Wu et al. [13] applied Bernstein’s inequality and the truncated method to greatly improve the conclusion in weaker condition on moment. Wang et al. [14] extended the result for independent random variables to the case of NOD random variables. Shi et al. [15] obtained (1.4) for Bn=1. Sung [16] studied the inverse moments for a class of nonnegative random variables.
Recently, Shen et al. [17] extended the result of Wu et al. [13] to the case of ρ-mixing random variables and obtained the following result.
Theorem A.
Let {Zn,n≥1} be a nonnegative ρ-mixing sequence with ∑n=1∞ρ(n)<∞. Suppose that
EZn2<∞, for all n≥1;
EXn→∞, where Xn is defined by (1.3);
for some η>0,
Rn(η)∶=Bn-2∑i=1nEZi2I(Zi>ηBn)⟶0,n⟶∞;
for some t∈(0,1) and any positive constants a,r,C,
limn→∞(a+EXn)r⋅exp{-C⋅(EXn)tn}=0.
Then for any a>0 and r>0, (1.4) holds.
In this paper, we will further study the asymptotic approximation of inverse moments for nonnegative ρ-mixing random variables with identical distribution. We will show that (1.4) holds under very mild conditions and the condition (iv) in Theorem A can be deleted. In place of the Bernstein type inequality used by Shen et al. [17], we make the use of Rosenthal type inequality of ρ-mixing random variables. Our main results are as follows.
Theorem 1.2.
Let {Zn,n≥1} be a sequence of nonnegative ρ-mixing random variables with identical distribution and let {Bn,n≥1} be a sequence of positive constants. Let a>0 and α>0 be real numbers. p>max{2,2α,α+1}. Assume that ∑n=1∞ρ2/p(2n)<∞. Suppose that
0<EZn<∞, for all n≥1;
μn≐EXn→∞ as n→∞, where Xn=Bn-1∑k=1nZk;
for all 0<ε<1, there exist b>0 and n0>0 such that
EZ1I(Z1>bBn)≤εEZ1,n≥n0.
Then (1.4) holds.
Corollary 1.3.
Let {Zn,n≥1} be a sequence of nonnegative ρ-mixing random variables with identical distribution and 0<EZ1<∞. Let {Bn,n≥1} be a sequence of positive constants satisfying Bn=O(nδ) for some 0<δ<1 and Bn→∞ as n→∞. Let a>0 and α>0 be real numbers. p>max{2,2α,α+1}. Assume that ∑n=1∞ρ2/p(2n)<∞. Then (1.4) holds.
By Theorem 1.2, we can get the following convergence rate of relative error in the relation (1.4).
Theorem 1.4.
Assume that conditions of Theorem 1.2 are satisfied and 0<EZn2<∞. p>max{2,4(α+1),2α+3}. If Bn≥Cn1/2 for all n large enough, where C is a positive constant, then
|(a+EXn)αE(a+Xn)-α-1|=O((a+EXn)-1).
Theorem 1.5.
Assume that conditions of Theorem 1.2 are satisfied and 0<EZn2<∞. p>max{2,4(α+1),2α+3}. Then
|(a+EXn)αE(a+Xn)-α-1|=O(n-1/2).
Taking Bn≡1 in Theorem 1.2, we have the following asymptotic approximation of inverse moments for the partial sums of nonnegative ρ-mixing random variables with identical distribution.
Theorem 1.6.
Let {Zn,n≥1} be a sequence of nonnegative ρ-mixing random variables with identical distribution. Let a>0 and α>0 be real numbers. p>max{2,2α,α+1}. Assume that ∑n=1∞ρ2/p(2n)<∞. Suppose that
0<EZn<∞, ∀n≥1;
νn≐EYn→∞ as n→∞, where Yn=∑k=1nZk;
for all 0<ε<1, there exist b>0 and n0>0 such that
EZ1I(Z1>b)≤εEZ1,n≥n0.
Then E(a+Yn)-α~(a+EYn)-α.
Remark 1.7.
Theorem 1.2 in this paper improves the corresponding results of Wu et al. [13], Wang et al. [14], and Shen et al. [17]. Firstly, Theorem 1.4 in this paper is based on the condition EZn<∞, for all n≥1, which is weaker than the condition EZn2<∞, for all n≥1 in the above cited references. Secondly, {Bn,n≥1} is an arbitrary sequence of positive constants in Theorem 1.2, while Bn2=∑i=1nVarZi in the above cited references. Thirdly, the condition (iv) in Theorem A is not needed in Theorem 1.2. Finally, (1.7) is weaker than (1.5) under the case of identical distribution. Actually, by the condition (1.5), we can see that
Bn-1∑i=1nEZiI(Zi>ηBn)≤η-1Bn-2∑i=1nEZi2I(Zi>ηBn)⟶0,n⟶∞,
which implies that for all 0<ε<1, there exists a positive integer n0 such that
Bn-1∑i=1nEZiI(Zi>ηBn)≤εμn=εBn-1∑i=1nEZi,n≥n0,
that is, (1.7) holds.
2. Proof of the Main Results
In order to prove the main results of the paper, we need the following important moment inequality for ρ-mixing random variables.
Lemma 2.1 (c.f. Shao [18, Corollary 1.1]).
Let q≥2 and {Xn,n≥1} be a sequence of ρ-mixing random variables. Assume that EXn=0, E|Xn|q<∞ and
∑n=1∞ρ2/q(2n)<∞.
Then there exists a positive constant K=K(q,ρ(·)) depending only on q and ρ(·) such that for any k≥0 and n≥1,
E(max1≤i≤n|Sk(i)|q)≤K[(nmaxk<i≤k+nEXi2)q/2+nmaxk<i≤k+nE|Xi|q],
where Sk(i)=∑j=k+1k+iXj, k≥0 and i≥1.
Remark 2.2.
We point out that if {Xn,n≥1} is a sequence of ρ-mixing random variables with identical distribution and the conditions of Lemma 2.1 hold, then we have
E(max1≤i≤n|Sk(i)|q)≤K[(nEX12)q/2+nE|X1|q],E(max1≤i≤n|∑j=1iXj|q)≤K[(nEX12)q/2+nE|X1|q]=K[(∑j=1nEXj2)q/2+∑j=1nE|Xj|q].
The inequality above is the Rosenthal type inequality of identical distributed ρ-mixing random variables.
Proof of Theorem 1.2.
It is easily seen that f(x)=(a+x)-α is a convex function of x on [0,∞), therefore, we have by Jensen’s inequality that
E(a+Xn)-α≥(a+EXn)-α,
which implies that
liminfn→∞(a+EXn)αE(a+Xn)-α≥1.
To prove (1.4), it is enough to prove that
limsupn→∞(a+EXn)αE(a+Xn)-α≤1.
In order to prove (2.6), we need only to show that for all δ∈(0,1),
limsupn→∞(a+EXn)αE(a+Xn)-α≤(1-δ)-α.
By (iii), we can see that for all δ∈(0,1),
EZ1I(Z1>bBn)≤δ2EZ1,n≥n0.
Let
Un=Bn-1∑k=1nZkI(Zk≤bBn),E(a+Xn)-α=E(a+Xn)-αI(Un≥μn-δμn)+E(a+Xn)-αI(Un<μn-δμn)≐Q1+Q2.
For Q1, since Xn≥Un, we have
Q1≤E(a+Xn)-αI(Xn≥μn-δμn)≤(a+μn-δμn)-α.
By (2.8), we have for n≥n0 that
μn-EUn=Bn-1∑k=1nEZkI(Zk>bBn)≤δμn2.
Therefore, by (2.12), Markov’s inequality, Remark 2.2 and Cr’s inequality, for any p>2 and all n sufficiently large,
Q2≤a-αP(Un<μn-δμn)=a-αP(EUn-Un>δμn-(μn-EUn))≤a-αP(EUn-Un>δμn2)≤a-αP(|Un-EUn|>δμn2)≤Cμn-pE|Un-EUn|p≤Cμn-p[Bn-2nEZ12I(Z1≤bBn)]p/2+Cμn-p[Bn-pnEZ1pI(Z1≤bBn)]≤Cμn-p[Bn-1nEZ1I(Z1≤bBn)]p/2+Cμn-pBn-1nEZ1I(Z1≤bBn)≤Cμn-p(μnp/2+μn)=C(μn-p/2+μn-(p-1)).
Taking p>max{2,2α,α+1}, we have by (2.10), (2.11), and (2.13) that
limsupn→∞(a+μn)αE(a+Xn)-α≤limsupn→∞(a+μn)α(a+μn-δμn)-α+limsupn→∞(a+μn)α[Cμn-p/2+Cμn-(p-1)]=(1-δ)-α,
which implies (2.7). This completes the proof of the theorem.
Proof of Corollary 1.3.
The condition Bn=O(nδ) for some 0<δ<1 implies that
μn≐EXn=Bn-1∑k=1nEZk=nBn-1EZ1,
thus, μn≥Cn1-δ→∞ as n→∞.
The fact 0<EZ1<∞ and Bn→∞ yield that EZ1I(Z1>bBn)→0 as n→∞, which implies that for all 0<ε<1, there exists n0>0 such that
EZ1I(Z1>bBn)≤εEZ1,n≥n0.
That is to say condition (iii) of Theorem 1.2 holds. Therefore, the desired result follows from Theorem 1.2 immediately.
Proof of Theorem 1.4.
Firstly, we will examine VarXn. By Remark 2.2, 0<EZ12<∞ and the condition Bn≥Cn1/2 for all n large enough, we can get that
VarXn=Bn-2Var(∑i=1nZi)≤Bn-2E(∑i=1nZi)2≤CnBn-2EZ12≤C1
for all n large enough.
Denote ϕ(x)=(a+x)-α for x≥0. By Taylor’s expansion, we can see that
ϕ(Xn)=ϕ(EXn)+ϕ′(ξn)(Xn-EXn),
where ξn is between Xn and EXn. It is easily seen that {ϕ′(x)}2 is decreasing in x≥0. Therefore, by (2.18), Cauchy-Schwartz inequality, (2.17) and (1.4), we have
[Eϕ(Xn)-ϕ(EXn)]2=E[ϕ′(ξn)(Xn-EXn)]2≤E[ϕ′(ξn)]2VarXn≤C1E[ϕ′(ξn)]2=C1E[ϕ′(ξn)]2I(Xn≤EXn)+C1E[ϕ′(ξn)]2I(Xn>EXn)≤C1E[ϕ′(Xn)]2+C1E[ϕ′(EXn)]2~2C1E[ϕ′(EXn)]2=2C1α2(a+EXn)-2(α+1).
This leads to (1.8). The proof is complete.
Proof of Theorem 1.5.
The proof is similar to that of Theorem 1.4. In place of VarXn≤C1, we make the use of VarXn≤CnBn-2EZ12≐C2nBn-2. The proof is complete.
Acknowledgments
The authors are most grateful to the Editor Tetsuji Tokihiro and an anonymous referee for the careful reading of the paper and valuable suggestions which helped to improve an earlier version of this paper. The paper is supported by the Academic innovation team of Anhui University (KJTD001B).
KolmogorovA. N.RozanovU. A.On the strong mixing conditions of a stationary Gaussian process19602222227WooffD. A.Bounds on reciprocal moments with applications and developments in Stein estimation and post-stratification1985472362371816102 (87b:62017)ZBL0603.62016PittengerA. O.Sharp mean-variance bounds for Jensen-type inequalities1990102919410.1016/0167-7152(90)90001-N1072493 (91i:60053)ZBL0705.60017MarciniakE.WesołowskiJ.Asymptotic Eulerian expansions for binomial and negative binomial reciprocals1999127113329333810.1090/S0002-9939-99-05105-91637400 (2000b:60033)ZBL0930.60004FujiokaT.Asymptotic approximations of the inverse moment of the noncentral chi-squared variable2001311991091863071 (2002h:62031)ZBL1031.62011GuptaR. C.AkmanO.Statistical inference based on the length-biased data for the inverse Gaussian distribution199831432533710.1080/023318898088026431711437 (2000e:62012)ZBL0930.62020MendenhallW.Lehman,E. H.Jr.An approximation to the negative moments of the positive binomial useful in life testing196022272420114279 (22#5103)ZBL0105.12305RamsayC. M.A note on random survivorship group benefits199323149156JurlewiczA.a.jurlewicz@im.pwr.wroc.plWeronK.karina@rainbow.if.pwr.wroc.plRelaxation of dynamically correlated clusters20023051–311212110.1016/S0022-3093(02)01087-6GarciaN. L.PalaciosJ. L.On inverse moments of nonnegative random variables200153323523910.1016/S0167-7152(01)00008-61841624 (2002b:60015)ZBL0991.60003KaluszkaM.OkolewskiA.On Fatou-type lemma for monotone moments of weakly convergent random variables2004661455010.1016/j.spl.2003.10.0092025652 (2004j:60029)ZBL1116.60308HuS. H.ChenGui JingWangX. J.ChenE. B.On inverse moments of nonnegative weakly convergent random variables20073023613672349267 (2009d:60031)ZBL1141.60309WuT.-J.ShiX. P.MiaoB. Q.Asymptotic approximation of inverse moments of nonnegative random variables200979111366137110.1016/j.spl.2009.02.0102537511 (2010h:60106)ZBL1168.60340WangX. J.HuS. H.YangW. Z.LingN. X.Exponential inequalities and inverse moment for NOD sequence2010805-645246110.1016/j.spl.2009.11.0232593586 (2011b:60066)ZBL1186.60015ShiX. P.WuY. H.LiuY.A note on asymptotic approximations of inverse moments of nonnegative random variables20108015-161260126410.1016/j.spl.2010.04.0042657491 (2011d:60142)ZBL1191.62020SungS. H.On inverse moments for a class of nonnegative random variables20102010132665482 (2011f:60032)ZBL1205.6004110.1155/2010/823767823767ShenA.baret@sohu.comShiY.zhqyueyue@163.comWangW.xiaojuan126@yahoo.com.cnHanB.ice05013861001@163.comBernstein-type inequality for weakly dependent sequence and its applications201111210.1007/s13163-011-0059-6ShaoQ. M.Maximal inequalities for partial sums of ρ-mixing sequences19952329489651334179 (96d:60027)10.1214/aop/1176988297