VMSTA Modern Stochastics: Theory and Applications 2351-6054 2351-6046 2351-6046 VTeXMokslininkų g. 2A, 08412 Vilnius, Lithuania VMSTA121 10.15559/18-VMSTA121 Research Article Convergence rates in the central limit theorem for weighted sums of Bernoulli random fields GiraudoDavidedavide.giraudo@rub.de Ruhr-Universität Bochum, Fakultät für Mathematik, IB 2/167, Universitätsstr. 150, 44780 Bochum, Germany 2019 2112201862251267 20122017 19102018 20102018 © 2019 The Author(s). Published by VTeX2019 Open access article under the CC BY license.

Moment inequalities for a class of functionals of i.i.d. random fields are proved. Then rates are derived in the central limit theorem for weighted sums of such randoms fields via an approximation by m-dependent random fields.

Random fields moment inequalities central limit theorem 60F05 60G60 This research was supported by the grand SFB 823.
Introduction and main results Goal of the paper

In its simplest form, the central limit theorem states that if (Xi)i1 is an independent identically distributed (i.i.d.) sequence of centered random variables having variance one, then the sequence (n1/2i=1nXi)n1 converges in distribution to a standard normal random variable. If X1 has a finite moment of order three, Berry  and Esseen  gave the following convergence rate: suptR|P{n1/2 i=1nXit}P{Nt}|CE[|X1|3]n1/2, where C is a numerical constant and N has the standard normal distribution. The question of extending the previous result to a larger class of sequences has received a lot of attention. When Xi can be represented as a function of an i.i.d. sequence, optimal convergence rates are given in .

In this paper, we will focus on random fields, that is, collections of random variables indexed by Zd and more precisely in Bernoulli random fields, which are defined as follows.

Let d1 be an integer. The random field (Xn)nZd is said to be Bernoulli if there exist an i.i.d. random field (εi)iZd and a measurable function f:RZdR such that Xn=f((εni)iZd) for each nZd .

We are interested in the asymptotic behavior of the sequence (Sn)n1 defined by Sn:=iZdbn,iXi, where bn:=(bn,i)iZd is an element of 2(Zd) . Under appropriated conditions on the dependence of the random field (Xi)iZd and the sequence of weights (bn)n1 that will be specified later, the sequence (Sn/bn2)n1 converges in law to a normal distribution . The goal of this paper is to provide bounds of the type Berry–Esseen in order to give convergence rates in the central limit theorem.

This type of question has been addressed for the so-called BL(θ) -dependent random fields , martingale differences random fields , positively and negatively dependent random fields  and mixing random fields .

In order to establish results of this kind, we need several ingredients. First, we need convergence rates for m-dependent random fields. Second, a Bernoulli random field can be decomposed as the sum of an m-dependent random field and a remainder. The control of the contribution of the remainder is done by a moment inequality in the spirit of Rosenthal’s inequality . One of the main applications of such an inequality is the estimate of the convergence rates in the central limit theorem for random fields that can be expressed as a functional of an i.i.d. random field. The method consists of approximating the considered random field by an m-dependent one and controlling the approximation with the help of the established moment inequality. In the one-dimensional case, probability and moment inequalities have been established in  for maxima of partial sums of Bernoulli sequences. The techniques used therein permit to derive results for weighted sums of such sequences.

The paper is organized as follows. In Subsection 1.2, we give the material which is necessary to understand the moment inequality stated in Theorem 1. We then give the results on convergence rates in Subsection 1.3 (for weighted sums, sums on subsets of Zd and in a regression model) and compare the obtained results in the case of linear random fields with some existing ones. Section 2 is devoted to the proofs.

Background

The following version of Rosenthal’s inequality is due to Johnson, Schechtman and Zinn : if (Yi)i=1n are independent centered random variables with a finite moment of order p2 , then i=1nYip14.5plogp(( i=1nYi22)1/2+( i=1nYipp)1/p), where Yq:=(E[|Y|q])1/q for q1 .

It was first establish without explicit constant in Theorem 3 of .

Various extensions of Rosenthal-type inequalities have been obtained under mixing conditions  or projective conditions . We are interested in extensions of (3) to the setting of dependent random fields.

Throughout the paper, we shall use the following notations.

For a positive integer d, the set {1,,d} is denoted by [d] .

The coordinatewise order is denoted by ≼, that is, for i=(iq)q=1dZd and j=(jq)q=1dZd , ij means that ikjk for any k[d] .

For q[d] , eq denotes the element of Zd whose qth coordinate is 1 and all the others are zero. Moreover, we write 0=(0,,0) and 1=(1,,1) .

For n=(nk)k=1dNd , we write the product k=1dnq as |n| .

The cardinality of a set I is denoted by |I| .

For a real number x, we denote by [x] the unique integer such that [x]x<[x]+1 .

We write Φ for the cumulative distribution function of the standard normal law.

If Λ is a subset of Zd and kZd , then Λk is defined as {lk,lΛ} .

For q1 , we denote by q(Zd) the space of sequences a:=(ai)iZd such that aq:=(iZd|ai|q)1/q<+ .

For i=(iq)q=1d , the quantity i is defined as max1qd|iq| .

Let (Yi)iZd be a random field. The sum iZdYi is understood as the L1 -limit of the sequence (Sk)k1 where Sk=iZd,ikYi .

Following  we define the physical dependence measure.

Let (Xi)iZd:=(f((εij))jZd)iZd be a Bernoulli random field, p1 and (εu)uZd be an i.i.d. random field which is independent of the i.i.d. random field (εu)uZd and has the same distribution as (εu)uZd . For iZd , we introduce the physical dependence measure δi,p:=XiXip, where Xi=f((εij)jZd) and εu=εu if u0 , ε0=ε0 .

In , various examples of Bernoulli random fields are given, for which the physical dependence measure is either computed or estimated. Proposition 1 of  also gives the following moment inequality: if Γ is a finite subset of Zd , (ai)iΓ is a family of real numbers and p2 , then for any Bernoulli random field (Xn)nZd , iΓaiXip(2piΓai2)1/2·jZdδj,p. This was used in  in order to establish functional central limit theorems. Truquet  also obtained an inequality in this spirit. If (Xi)iZd is an i.i.d. and centered random field, (3) would give iΓaiXipC(iΓai2)1/2X1p, while Rosenthal’s inequality (3) would give iΓaiXipC(iΓai2)1/2X12+C(iΓ|ai|p)1/pX1p, what is a better result in this context.

In the case of linear processes, equality δj,pKδj,2 holds for a constant K which does not depend on j . However, there are processes for which such an inequality does not hold.

We give an example of a random field for which there is no constant K such that δj,pKδj,2 holds for all jZd . Let p2 and let (εi)iZd be an i.i.d. random field and for each kZd , let fk:RR be a function such that the random variable Zk:=fk(ε0) is centered and has a finite moment of order p, and kZdZk22<+ . Define Xn:=limN+N1jN1fk(εnk) , where the limit is taken in L2 . Then XiXi=fi(ε0)fi(ε0) , hence δi,2 is of order Zi2 while δi,p is of order Zip .

Consequently, having the p -norm instead of the 2 -norm of the (ai)iΓ is more suitable.

Mains results

We now give a Rosenthal-like inequality for weighted sums of Bernoulli random fields in terms of the physical dependence measure.

Let (εi)iZd be an i.i.d. random field. For any measurable function f:RZdR such that Xj:=f((Xji)iZd) has a finite moment of order p2 and is centered, and any (ai)iZd2(Zd) , iZdaiXip14.5plogp(iZdai2)1/2 j=0+(4j+4)d/2X0,j2+14.5plogp(iZd|ai|p)1/p j=0+(4j+4)d(11/p)X0,jp, where for j1 , X0,j=E[X0σ{εu,uj}]E[X0σ{εu,uj1}] and X0,0=E[X0σ{ε0}] .

We can formulate a version of inequality (8) where the right-hand side is expressed in terms of the coefficients of the physical dependence measure. The obtained result is not directly comparable to (5) because of the presence of the p -norm of the coefficients.

Let {εi,iZd} be an i.i.d. set of random variables. Then for any measurable function f:RZdR such that Xj:=f((Xji)iZd) has a finite moment of order p2 and is centered, and any (ai)iZd2(Zd) , iZdaiXip214.5plogp(iZdai2)1/2 j=0+(4j+4)d/2(i=jδi,22)1/2+214.5plogpp1(iZd|ai|p)1/p j=0+(4j+4)d(11/p)(i=jδi,p2)1/2.

Let (Xj)jZd=f((εji)iZd) be a centered square integrable Bernoulli random field and for any positive integer n, let bn:=(bn,i)iZd be an element of 2(Zd) . We are interested in the asymptotic behavior of the sequence (Sn)n1 defined by Sn:=iZdbn,iXi. Let us denote for kZd the map τk:2(Zd)2(Zd) defined by τk((xi)iZd):=(xi+k)iZd.

In , Corollary 2.6 gives the following result: under a Hannan type condition on the random field (Xi)iZd and under the condition on the weights that for any q[d] , 1bn2τeq(bn)bn2=0, the series iZd|Cov(X0,Xi)| converges, and the sequence (Sn/bn2)n1 converges in distribution to a centered normal distribution with variance σ2 , where σ:=(iZdCov(X0,Xi))1/2. The argument relies on an approximation by an m-dependent random field.

The purpose of the next theorem is to give a general rates of convergence. In order to measure it, we define Δn:=suptR|P{Snbn2t}Φ(t/σ)|. The following quantity will also play an important role in the estimation of convergence rates: εn:=jZd|E[X0Xj]|iZd|bn,ibn,i+jbn21|.

Let p>2 }2$]]>, p:=min{p,3} and let (Xj)jZd=(f((εji)iZd))jZd be a centered Bernoulli random field with a finite moment of order p and for any positive integer n, let bn:=(bn,i)iZd be an element of 2(Zd) such that for any n1 , the set {kZd,bn,k0} is finite and nonempty, limn+bn2=+ and (12) holds for any q[d] . Assume that for some positive α and β, the following series are convergent: C2(α):= i=0+(i+1)d/2+αX0,i2andCp(β):= i=0+(i+1)d(11/p)+βX0,ip. Let Sn be defined by (11). Assume that iZd|Cov(X0,Xi)| is finite and that σ be given by (13) is positive. Let γ>0 }0$]]> and let n0:=inf{N1nN,σ2+εn29(log2)1C2(α)([bn2]γ)ασ/2}. Then for each nn0 , Δn˜150(29([bn2]+21)γ+21)(p1)dX0pp(bnpbn2)p(σ/2)p+(2|εn|σ2+80(log2)1bn2γασ2C2(α)2)(2πe)1/2+(14.5pσlogp4d/2bn2γαC2(α))pp+1+(bnpσbn214.5plogp4d(11/p)bn2γβCp(β))pp+1. In particular, there exists a constant κ such that for all nn0 , Δnκ(bn2γ(p1)dpbnpp+|εn|)+κ(bn2γαpp+1+bnppp+1bn2pp+1(γβ+1)).

If (12), limn+bn2=+ and the family (δi,2)iZd is summable, then the sequence (εn)n1 converges to 0 hence n0 is well defined. However, it is not clear to us whether the finiteness of C2(α) combined with (12) and limn+bn2=+ imply that jZd|E[X0Xj]| is finite. Nevertheless, we can show an analogous result in terms of coefficients δi,p with the following changes in the statement of Theorem 2:

the definition of C2(α) should be replaced with C2(α):=2 j=0+(j+1)d/2+α(i=jδi,22)1/2;

the definition of Cp(β) should be replaced with Cp(β):=2(p1) j=0+(j+1)d(11/p)+β(i=jδi,22)1/2.

In this case, the convergence of iZd|Cov(X0,Xi)| holds (cf. Proposition 2 in ).

Recall the notation (N.8). Let (Λn)n1 be a sequence of subsets of Zd . The choice bn,j=1 if jΛn and 0 otherwise yields the following corollary for set-indexed partial sums.

Let (Xi)iZd be a centered Bernoulli random field with a finite moment of order p2 , p:=min{p,3} and let (Λn)n1 be a sequence of subset of Zd such that |Λn|+ and for any kZd , limn+|Λn(Λnk)|/|Λn|=1 . Assume that the series defined in (16) are convergent for some positive α and β, that iZd|Cov(X0,Xi)| is finite and that σ defined by (13) is positive. Let γ>0 }0$]]> and n0 be defined by (17). There exists a constant κ such that for any nn0 , suptR|P{iΛnXi|Λn|1/2t}Φ(t/σ)|κ(|Λn|q+jZd|E[X0Xj]|||Λn(Λnj)||Λn|1|), where q:=max{γ(p1)dp2+1;γαp2(p+1);2ppγβ2(p+1)}. We consider now the following regression model: Yi=g(in)+Xi,iΛn:={1,,n}d, where g:[0,1]dR is an unknown smooth function and (Xi)iZd is a zero mean stationary Bernoulli random field. Let K be a probability kernel defined on Rd and let (hn)n1 be a sequence of positive numbers which converges to zero and which satisfies limn+nhn=+andlimn+nhnd+1=0. We estimate the function g by the kernel estimator gn defined by gn(x)=iΛnYiK(xi/nhn)iΛnK(xi/nhn),x[0,1]d. We make the following assumptions on the regression function g and the probability kernel K: The probability kernel K fulfills RdK(u)du=1 , is symmetric, non-negative, supported by [1,1]d . Moreover, there exist positive constants r, c and C such that for any x,y[1,1]d , |K(x)K(y)|rxy and cK(x)C . We measure the rate of convergence of ((nhn)d/2(gn(x)E[gn(x)]))n1 to a normal distribution by the use of the quantity Δn˜:=suptR|P{(nhn)d/2(gn(x)E[gn(x)])t}Φ(tσK2)|. Two other quantities will be involved, namely, An:=(nhn)d/2(iΛnK2(xi/nhn))1/2KL2(Rd)1(iΛnK(xi/nhn))1/2 and εn:=jZd|E[X0Xj]|(iΛn(Λnj)K(xi/nhn)K(x(ij)/nhn)kΛnK2(xk/nhn)1). Let p>2 }2$]]>, p:=min{p,3} and let (Xj)jZd=(f((εji)iZd))jZd be a centered Bernoulli random field with a finite moment of order p. Assume that for some positive α and β, the following series are convergent: C2(α):= i=0+(i+1)d/2+αX0,i2andCp(β):= i=0+(i+1)d(11/p)+βX0,ip.

Let gn(x) be defined by (26), (hn)n1 be a sequence which converges to 0 and satisfies (25).

Assume that iZd|Cov(X0,Xi)| is finite and that σ:=jZdCov(X0,Xj)>0 }0$]]>. Let n1N be such that for each nn1 , 12(nhn)dK(xi/nhn)32 and 12KL2(Rd)(nhn)dK2(xi/nhn)32KL2(Rd). Let n0 be the smallest integer for which for all nn0 , σ2+εn29(log2)1C2(α)([(iΛnK(1hn(xin))2)1/2]γ)ασ/2. Then there exists a constant κ such that for each nmax{n0,n1} , Δnκ|An1|pp+1+|εn|+κ(nhn)d2(γ(p1)dp+2)+(nhn)d2γαpp+1+(nhn)2dp(γβ+1)2(p+1). Lemma 1 in  shows that under (25), the sequence (An)n1 goes to 1 as n goes to infinity and that the integer n1 is well defined. We now consider the case of linear random fields in dimension 2, that is, Xj1,j2=i1,i2Zai1,i2εj1i1,j2i2, where (ai1,i2)i1,i2Z1(Z2) and (εu1,u2)u1,u2Z2 is an i.i.d. centered random field and ε0,0 has a finite variance. We will focus on the case where the weights are of the form bn,i1,i2=1 if 1i1,i2n and bn,i1,i2=0 otherwise. Mielkaitis and Paulauskas  established the following convergence rate. Denoting Δn:=supr0|P{|1n i1,i2=1nXi1,i2|r}P{|N|r}| and assuming that E[|ε0,0|2+δ] is finite and k1,k2Z(|k1|+1)2(|k2|+1)2ak1,k22<+, the following estimate holds for Δn : Δn=O(nr),r:=12min{δ,113+δ}. In the context of Corollary 2, the condition on the coefficients reads as follows: i=0+(i1+α+i22/p+β)((j1,j2):(j1,j2)=iaj1,j22)1/2<+, where p=2+δ . Let us compare (37) with (39). Let s:=max{1+α,22/p+β} . When s2 , (39) implies (37). However, this implication does not hold if s<3/2 . Indeed, let r(s+1,5/2) and define ak1,k2:=k1r if k1=k21 and ak1,k2:=0 otherwise. Then (39) holds whereas (37) does not. Let us discuss the convergence rates in the following example. Let ak1,k2:=2|k1||k2| and let p=2+δ , where δ(0,1] . In our context, ||Λn(Λnj)||Λn|1|n2(nj1)(nj2)n2j1+j2n, hence the convergence of j1,j2Z|Cov(X0,0,Xj1,j2)|(j1+j2) guarantees that εn in Corollary 2 is of order 1/n . Moreover, since (39) holds for all α and β, the choice of γ allows to reach rates of the form nδ+r0 for any fixed r0 . In particular, when δ=1 , one can reach for any fixed r0 rates of the form n1+r0 . In comparison, with the same assumptions, the result of  gives n3/8 . Proofs Proof of Theorem <xref rid="j_vmsta121_stat_004">1</xref> We define for j1 and iZd , Xi,j=E[Xiσ(εu,uij)]E[Xiσ(εu,uij1)]. In this way, by the martingale convergence theorem, XiE[Xiεi]=limN+ j=1NXi,j, hence iZdaiXip j=1+iZdaiXi,jp+iZdaiE[Xiεi]p. Let us fix j1 . We divide Zd into blocks. For vZd , we define Av:= q=1d([(2j+2)vq,(2j+2)(vq+1)1]Z), and if K is a subset of [d] , we define EK:={vZd,vqis even if and only ifqK}. Therefore, the following inequality takes place iZdaiXi,jpK[d]vEKiAvaiXi,jp. Observe that the random variable iAvaiXi,j is measurable with respect of the σ-algebra generated by εu , where u satisfies (2j+2)vq(j+1)uqj+1+(2j+2)(vq+1)1 for all q[d] . Since the family {εu,uZd} is independent, the family {iAvaiXi,j,vEK} is independent for each fixed K[d] . Using inequality (3), it thus follows that vEKiAvaiXi,jp14.5plogp(vEKiAvaiXi,j22)1/2+14.5plogp(vEKiAvaiXi,jpp)1/p. By stationarity, one can see that Xi,jq=X0,jq for q{2,p} , hence the triangle inequality yields vEKiAvaiXi,jp14.5plogpX0,j2(vEK(iAv|ai|)2)1/2+14.5plogpX0,jp(vEK(iAv|ai|)p)1/p. By Jensen’s inequality, for q{2,p} , (iAv|ai|)q|Av|q1iAv|ai|q(2j+2)d(q1)iAv|ai|q and using i=1Nxi1/qNq1q(i=1Nxi)1/q , it follows that K[d]vEKiAvaiXi,jp14.5plogpX0,j2(iZdai2)1/2(4j+4)d/2+14.5plogpX0,jp(iZd|ai|p)1/p(4j+4)d(11/p). Combining (43), (46) and (50), we derive that iZdaiXip14.5plogp j=1+X0,j2(iZdai2)1/2(4j+4)d/2+14.5plogp j=1+X0,jp(iZd|ai|p)1/p(4j+4)d(11/p)+iZdaiE[Xiεi]p. In order to control the last term, we use inequality (3) and bound E[Xiεi]q by X0,0q for q{1,2} . This ends the proof of Theorem 1. The following lemma gives a control of the Lq -norm of X0,j in terms of the physical measure dependence. For q{2,p} and jN , the following inequality holds X0,jq(2(q1)iZd,i=jδi,q2)1/2. Let j be fixed. Let us write the set of elements of Zd whose infinite norm is equal to j as {vs,1sNj} where NjN . We also assume that vsvs1{ek,1kd} for all s{2,,Nj} . Denote Fs:=σ(εu,uj,εvt,1ts), and F0:=σ(εu,uj) . Then X0,j=s=1NjE[X0Fs]E[X0Fs1] , from which it follows, by Theorem 2.1 in , that X0,jq2(q1) s=1NjE[X0Fs]E[X0Fs1]q2. Then arguments similar as in the proof of Theorem 1 (i) in  give the bound E[X0Fs]E[X0Fs1]qδvs,q+δvs1,q . This ends the proof of Lemma 1. □ Now, Corollary 1 follows from an application of Lemma 1 with q=2 and q=p respectively. □ Proof of Theorem <xref rid="j_vmsta121_stat_006">2</xref> Denote for a random variable Z the quantity δ(Z):=suptR|P{Zt}Φ(t)|. We say that a random field (Yi)iZd is m-dependent if the collections of random variables (Yi,iA) and (Yi,iB) are independent whenever inf{ab,aA,bB}>m. }m.\]]]> The proof of Theorem 2 will use the following tools. By Theorem 2.6 in , if I is a finite subset of Zd , (Yi)iI an m-dependent centered random field such that E[|Yi|p]<+ for each iI and some p(2,3] and Var(iIYi)=1 , then δ(iIYi)75(10m+1)(p1)diIE[|Yi|p]. By Lemma 1 in , for any two random variables Z and Z and p1 , δ(Z+Z)2δ(Z)+Zppp+1. Let (εu)uZd be an i.i.d. random field and let f:RZdR be a measurable function such that for each iZd , Xi=f((εiu)uZd) . Let γ>0 }0$]]> and n0 defined by (17).

Let m:=([bn2]+1)γ and let us define Xi(m):=E[Xiσ(εu,im1ui+m1)]. Since the random field (εu)uZd is independent, the following properties hold.

The random field (Xi(m))iZd is (2m+1) -dependent.

The random field (Xi(m))iZd is identically distributed and Xi(m)pX0p .

For any (ai)iZd2(Zd) and q2 , the following inequality holds: iZdai(XiXi(m))q14.5qlogq(iZdai2)1/2jm(4j+4)d/2X0,j2+14.5qlogq(iZd|ai|q)1/qjm(4j+4)d(11/q)X0,jq. In order to prove (59), we follow the proof of Theorem 1 and start from the decomposition XiXi(m)=limN+j=mNXi,j instead of (42).

Define Sn(m):=iZdbn,iXi(m) . An application of (T.2) to Z:=Sn(m)bn21σ1 and Z:=(SnSn(m))bn21σ1 yields Δn2δ(Sn(m)σbn2)+σpp+11bn2pp+1SnSn(m)ppp+1. Moreover, δ(Sn(m)σbn2)=suptR|P{Sn(m)σbn2t}Φ(t)| =supuR|P{Sn(m)Sn(m)2u}Φ(uSn(m)2σbn2)| δ(Sn(m)Sn(m)2)+supuR|Φ(uSn(m)2σbn2)Φ(u)|, hence, by (P.1) and (T.1) applied with Yi:=Xi(m)/Sn(m)2 , p instead of p and 2m+1 instead of m, we derive that Δn(I)+(II)+(III) where (I):=150(20m+21)(p1)diZd|bn,i|pXi(m)ppSn(m)2p, (II):=2supuR|Φ(uSn(m)2σbn2)Φ(u)|and (III):=σpp+11bn2pp+1SnSn(m)ppp+1. By (P.2) and the reversed triangular inequality, the term (I) can be bounded as (I)150(20m+21)(p1)dX0ppbnpp(Sn2SnSn(m)2)p and by (P.3) with q=2 , we obtain that (Sn2SnSn(m)2)p(Sn229(log2)1mαbn2C2(α))p. By (15), we have Sn22bn22=σ2+εn, and we eventually get (I)150(20m+21)(p1)dX0pp(bnpbn2)p·(σ2+εn29(log2)1mαC2(α))p. Since nn0 , we derive, in view of (17), (I)150(20m+21)(p1)dX0pp(bnpbn2)p(σ/2)p.

In order to bound (II) , we argue as in  (p. 456). Doing similar computations as in  (p. 272), we obtain that (II)(2πe)1/2(infk1ak)1|an21|, where an:=Sn(m)2σ1bn21 . Observe that for any n, by (P.3), anSn2SnSn(m)2σbn2σ2+εn29(log2)1C2(α)mασ and using again (P.3) combined with Theorem 1 for p=q=2 , |an21|=|Sn(m)22σ2bn221| |Sn22σ2bn221|+|Sn(m)22Sn22|σ2bn22 |εn|σ2+|Sn(m)2Sn2|(Sn(m)2+Sn2)σ2bn22 |εn|σ2+Sn(m)Sn2(Sn(m)2+Sn2)σ2bn22 |εn|σ2+40(log2)1mασ2C2(α)2. This leads to the estimate (II)(2πe)1/2σ2+εn29(log2)1C2(α)mα(|εn|σ+40(log2)1mασC2(α)2), and since nn0 , we derive, in view of (17), (II)(2|εn|σ2+80(log2)1bn2γασ2C2(α)2)(2πe)1/2.

The estimate of (III) relies on (P.3): (III)σpp+1(14.5plogpjm(4j+4)d/2X0,j2)pp+1+σpp+1bn2pp+1bnppp+1(14.5plogpjm(4j+4)d(11/p)X0,jp)pp+1 hence (III)(14.5pσlogp4d/2bn2γαC2(α))pp+1+(bnpσbn214.5plogp4d(11/p)bn2γβCp(β))pp+1. The combination of (64), (71), (80) and (82) gives (18).

Proof of Theorem <xref rid="j_vmsta121_stat_009">3</xref>

Since the random variables Xi are centered, we derive by definition of gn(x) that (nhn)d/2(gn(x)E[gn(x)])=(nhn)d/2iΛnXiK(xi/nhn)iΛnK(xi/nhn). We define bn,i=K(1hn(xin)),iΛn, and bn,i=0 otherwise. Denote bn=(bn,i)iZd and bn2:=(iZdbn,i2)1/2 . In this way, by (83) and (28), 1KL2(Rd)σ(nhn)d/2(gn(x)E[gn(x)])=1σiZdbn,iXibn21An. Applying (T.2) to Z=iZdbn,iXibn21andZ=iZdbn,iXibn21σ1(An1) and using Theorem 1, we derive that Δn˜cpΔn+cp(σ1C2(α)+Cp(β))pp+1|An1|pp+1, where Δn=suptR|P{Zt}Φ(tσ)|. We then use Theorem 2 to handle Δn (which is allowed, by (A)). Using boundedness of K, we control the p and p norms by a constant times the 2 -norm. This ends the proof of Theorem 3.

Acknowledgments

The author would like to thank the referees for many suggestions which improved the presentation of the paper.

References Bentkus, V., Sunklodas, J.K.: On normal approximations to strongly mixing random fields. Publ. Math. (Debr.) 70(3-4), 253270 (2007). MR2310650 Berry, A.C.: The accuracy of the Gaussian approximation to the sum of independent variates. Trans. Am. Math. Soc. 49, 122136 (1941). MR0003498. https://doi.org/10.2307/1990053 Biermé, H., Durieu, O.: Invariance principles for self-similar set-indexed random fields. Trans. Am. Math. Soc. 366(11), 59635989 (2014). MR3256190. https://doi.org/10.1090/S0002-9947-2014-06135-7 Bulinski, A.V.: On the convergence rates in the CLT for positively and negatively dependent random fields. In: Probability Theory and Mathematical Statistics (St. Petersburg, 1993), pp. 314. Gordon and Breach, Amsterdam (1996). MR1661688 Bulinski, A., Kryzhanovskaya, N.: Convergence rate in CLT for vector-valued random fields with self-normalization. Probab. Math. Stat. 26(2), 261281 (2006). MR2325308 Bulinskii, A.V., Doukhan, P.: Vitesse de convergence dans le théorème de limite centrale pour des champs mélangeants satisfaisant des hypothèses de moment faibles. C. R. Acad. Sci. Paris Sér. I Math. 311(12), 801805 (1990). MR1082637 Chen, L.H.Y., Shao, Q.-M.: Normal approximation under local dependence. Ann. Probab. 32(3A), 19852028 (2004). MR2073183. https://doi.org/10.1214/009117904000000450 El Machkouri, M., Ouchti, L.: Exact convergence rates in the central limit theorem for a class of martingales. Bernoulli 13(4), 981999 (2007). MR2364223. https://doi.org/10.3150/07-BEJ6116 El Machkouri, M.: Kernel density estimation for stationary random fields. ALEA Lat. Am. J. Probab. Math. Stat. 11(1), 259279 (2014). MR3225977 El Machkouri, M., Stoica, R.: Asymptotic normality of kernel estimates in a regression model for random fields. J. Nonparametr. Stat. 22(8), 955971 (2010). MR2738877. https://doi.org/10.1080/10485250903505893 El Machkouri, M., Volný, D., Wu, W.B.: A central limit theorem for stationary random fields. Stoch. Process. Appl. 123(1), 114 (2013). MR2988107. https://doi.org/10.1016/j.spa.2012.08.014 Esseen, C.-G.: On the Liapounoff limit of error in the theory of probability. Ark. Mat. Astron. Fys. 28A(9), 19 (1942). MR0011909 Jirak, M.: Berry-Esseen theorems under weak dependence. Ann. Probab. 44(3), 20242063 (2016). MR3502600. https://doi.org/10.1214/15-AOP1017 Johnson, W.B., Schechtman, G., Zinn, J.: Best constants in moment inequalities for linear combinations of independent and exchangeable random variables. Ann. Probab. 13(1), 234253 (1985). MR0770640 Klicnarová, J., Volný, D., Wang, Y.: Limit theorems for weighted Bernoulli random fields under Hannan’s condition. Stoch. Process. Appl. 126(6), 18191838 (2016). MR3483738. https://doi.org/10.1016/j.spa.2015.12.006 Liu, W., Xiao, H., Wu, W.B.: Probability and moment inequalities under dependence. Stat. Sin. 23(3), 12571272 (2013). MR3114713 Merlevède, F., Peligrad, M.: Rosenthal-type inequalities for the maximum of partial sums of stationary processes and examples. Ann. Probab. 41(2), 914960 (2013). MR3077530. https://doi.org/10.1214/11-AOP694 Mielkaitis, E., Paulauskas, V.: Rates of convergence in the CLT for linear random fields. Lith. Math. J. 51(2), 233250 (2011). MR2805741. https://doi.org/10.1007/s10986-011-9122-8 Nakhapetyan, B.S., Petrosyan, A.N.: On the rate of convergence in the central limit theorem for martingale-difference random fields. Izv. Nats. Akad. Nauk Armenii Mat. 39(2), 5968 (2004). MR2167826 Pavlenko, D.A.: An estimate for the rate of convergence in the central limit theorem for a class of random fields. Vestnik Moskov. Univ. Ser. I Mat. Mekh. 3, 8587 (1993). MR1355526 Peligrad, M., Utev, S., Wu, W.B.: A maximal Lp -inequality for stationary sequences and its applications. Proc. Am. Math. Soc. 135(2), 541550 (2007). MR2255301. https://doi.org/10.1090/S0002-9939-06-08488-7 Rio, E.: Théorie Asymptotique des Processus Aléatoires Faiblement Dépendants. Mathématiques & Applications (Berlin) [Mathematics & Applications], vol. 31, p. 169. Springer, Berlin (2000). MR2117923 Rio, E.: Moment inequalities for sums of dependent random variables under projective conditions. J. Theor. Probab. 22(1), 146163 (2009). MR2472010. https://doi.org/10.1007/s10959-008-0155-9 Rosenthal, H.P.: On the subspaces of Lp (p>2) }2)\$]]> spanned by sequences of independent random variables. Isr. J. Math. 8, 273303 (1970). MR0271721. https://doi.org/10.1007/BF02771562 Shao, Q.M.: Maximal inequalities for partial sums of ρ-mixing sequences. Ann. Probab. 23(2), 948965 (1995). MR1334179 Truquet, L.: A moment inequality of the Marcinkiewicz-Zygmund type for some weakly dependent random fields. Stat. Probab. Lett. 80(21–22), 16731679 (2010). MR2684016. https://doi.org/10.1016/j.spl.2010.07.011 Wu, W.B.: Nonlinear system theory: another look at dependence. Proc. Natl. Acad. Sci. USA 102(40), 1415014154 (2005). MR2172215. https://doi.org/10.1073/pnas.0506715102 Yang, W., Wang, X., Li, X., Hu, S.: Berry-Esséen bound of sample quantiles for ϕ-mixing random variables. J. Math. Anal. Appl. 388(1), 451462 (2012). MR2869760. https://doi.org/10.1016/j.jmaa.2011.10.058