Communications in Statistics - Theory and Methods

ISSN: 0361-0926 (Print) 1532-415X (Online) Journal homepage: http://www.tandfonline.com/loi/lsta20

Complete convergence for weighted sums of

extended negatively dependent random variables

Aiting Shen, Mingxiang Xue & Wenjuan Wang

To cite this article: Aiting Shen, Mingxiang Xue & Wenjuan Wang (2017) Complete convergence
for weighted sums of extended negatively dependent random variables, Communications in
Statistics - Theory and Methods, 46:3, 1433-1444, DOI: 10.1080/03610926.2015.1019147

To link to this article: http://dx.doi.org/10.1080/03610926.2015.1019147

Published online: 27 Oct 2016.

Submit your article to this journal

Article views: 14

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=lsta20

Download by: [University of Regina] Date: 03 December 2016, At: 18:38

COMMUNICATIONS IN STATISTICS—THEORY AND METHODS
, VOL. , NO. , –
http://dx.doi.org/./..

Complete convergence for weighted sums of extended

negatively dependent random variables
Aiting Shen, Mingxiang Xue, and Wenjuan Wang
School of Mathematical Sciences, Anhui University, Hefei, P.R. China

ABSTRACT ARTICLE HISTORY

In this article, the complete convergence for weighted sums of extended Received  September 
negatively dependent (END, in short) random variables without identi- Accepted  February 
cal distribution is investigated. In addition, the complete moment con-
KEYWORDS
vergence for weighted sums of END random variables is also obtained. Complete convergence;
As an application, the Baum–Katz type result for END random vari- complete moment
ables is established. The results obtained in the article extend the corre- convergence; END random
sponding ones for independent random variables and some dependent variables.
random variables.
MATHEMATICS SUBJECT
CLASSIFICATION
F

1. Introduction
It is well known that the complete convergence plays an important role in probability limit
theory and mathematical statistics, especially in establishing the convergence rate. The con-
cept of complete convergence was introduced by Hsu and Robbins (1947) as follows. A
sequence of random variables {Un , n ≥ 1} is said to converge completely to a constant C if
 ∞
n=1 P(|Un − C| > ε) < ∞ for all ε > 0. In view of the Borel–Cantelli lemma, this implies
that Un → C almost surely (a.s.). The converse is true if {Un , n ≥ 1} are independent. Hsu
and Robbins (1947) proved that the sequence of arithmetic means of independent and iden-
tically distributed (i.i.d.) random variables converges completely to the expected value if the
variance of the summands is finite. Erdös (1949) proved the converse. The result of Hsu–
Robbins–Erdös is a fundamental theorem in probability theory and has been generalized and
extended in several directions by many authors. One of the most important generalizations is
Baum and Katz (1965) for the strong law of large numbers as follows.
Theorem 1.1 (Baum and Katz, 1965). Let 1/2 < α ≤ 1 and α p > 1. Let {Xn , n ≥ 1} be a
sequence of independent and identically distributed random variables. Assume further that
EX1 = 0 if α ≤ 1. Then the following statements are equivalent:
E|X1 | p < ∞;
(i)  
∞ α p−2
(ii) n=1 n P(max1≤k≤n | ki=1 Xi | > εnα ) < ∞ for all ε > 0.

Since then, many authors devoted to study the Baum and Katz type strong law of large num-
bers for dependent random variables, such as Peligrad and Gut (1999), Wu and Jiang (2008)
for ρ̃-mixing sequence, Stoica (2007) and Wang and Hu (2014) for martingales, Wu (2011)

CONTACT Aiting Shen shenaiting@.com School of Mathematical Sciences, Anhui University, Hefei , P.R.
China.
©  Taylor & Francis Group, LLC
1434 A. SHEN ET AL.

for negatively orthant dependent sequence, Wang and Hu (2012) for ϕ-mixing sequence,
Wu (2012) for PQND sequence, Wang et al. (2012) for AANA sequence, and so on. The main
purpose of the article is to study the Baum and Katz type result for weighted sums of extended
negatively dependent (END) sequence.
The concept of END sequence was introduced by Liu (2009) as follows.
Definition 1.1. A finite collection of random variables X1 , X2 , . . . , Xn is said to be extended
negatively dependent (END, in short) if there exists a constant M > 0 such that both

n
P(X1 > x1 , X2 > x2 , . . . , Xn > xn ) ≤ M P(Xi > xi )
i=1

and

n
P(X1 ≤ x1 , X2 ≤ x2 , . . . , Xn ≤ xn ) ≤ M P(Xi ≤ xi )
i=1

hold for all real numbers x1 , x2 , . . . , xn . An infinite sequence {Xn , n ≥ 1} is said to be END if
every finite subcollection is END.

In the case M = 1, the notion of END random variables reduces to the well-known notion
of so-called negatively orthant dependent (NOD, in short) random variables which was intro-
duced by Joag-Dev and Proschan (1983). They also pointed out that negatively associated
(NA, in short) random variables are NOD and thus NA random variables are END. Hence,
the class of END includes independent sequence, NA sequence, and NOD sequence as special
cases. Studying the limiting behavior of END random variables is of great interest.
Some applications for END sequence have been found. For example, Chen et al. (2010)
established the strong law of large numbers for END random variables and showed applica-
tions to risk theory and renewal theory. Shen (2011) presented some probability inequalities
for END sequences and gave some applications. Wu and Guan (2012) presented some con-
vergence properties for the partial sums of END random variables. Wang and Wang (2013)
investigated a more general precise large deviation result for random sums of END real-valued
random variables in the presence of consistent variation. Qiu et al. (2013) and Wang et al.
(2013a, b, 2014) provided some results on complete convergence for sequences of END ran-
dom variables or arrays of row-wise END random variables, and so forth. We will further
study the complete convergence for weighted sums of END random variables.
The following definitions of slowly varying function and stochastic domination play
important roles throughout the article.
Definition 1.2. A real-valued function l(x), positive and measurable on (0, ∞), is said to be
slowly varying if
l(xλ)
lim =1 (1)
x→∞ l(x)

for each λ > 0.

Definition 1.3. A sequence {Xn , n ≥ 1} of random variables is said to be stochastically dom-
inated by a random variable X if there exists a positive constant C such that
P(|Xn | > x) ≤ CP(|X| > x)
for all x ≥ 0 and n ≥ 1.
 COMMUNICATIONS IN STATISTICS—THEORY AND METHODS 1435

Throughout the article, C denotes a positive constant which may be different in various
places. Denote X + = max(X, 0) and X − = max(−X, 0). Let I(A) be the indicator function
of the set A and x be the integer part of x.

2. Main results
Our main results are as follows.
Theorem 2.1. Let 0 < p < 2, α > 0, α p > 1, and {Xn , n ≥ 1} be a sequence of END random
variables, which is stochastically dominated by a random variable X. Let l(x) > 0 be a slowly
varying function. Assume further that EXn = 0 if p > 1, and {ani , 1 ≤ i ≤ n, n ≥ 1} is an array
of real numbers such that

n
a2ni = O(n). (2)
i=1

If
 
E |X| p l(|X|1/α ) < ∞, (3)

then for any ε > 0,

 n 

∞  
 
nα p−2 l(n)P  ani Xi  > εnα < ∞. (4)
n=1
 
i=1

Theorem 2.2. Suppose that the conditions of Theorem 2.1 hold and 1 < p < 2, then for any
ε > 0,
 n  +
∞  
α p−2−α   α
n l(n)E  ani Xi  − εn < ∞. (5)
n=1
 
i=1

Theorem 2.3. Let 1/2 < α ≤ 1 and {Xn , n ≥ 1} be a sequence of mean zero END random vari-
ables, which is stochastically dominated by a random variable X. Let l(x) > 0 be a slowly varying
and monotone non decreasing function. Assume further that {ani , 1 ≤ i ≤ n, n ≥ 1} is an array
of real numbers such that (2) holds. If
 
E |X|1/α l(|X|1/α ) < ∞, (6)

then for any ε > 0

 n 

∞  
l(n)  
P  ani Xi  > εnα < ∞. (7)
n=1
n  
i=1

Remark 2.1. Denote Sn = ni=1 ani Xi . We point out that (5) implies (4). This can be obtained
by the following inequality:

∞ 
∞ ∞
nα p−2−α l(n)E (|Sn | − εnα )+ = nα p−2−α l(n) P (|Sn | − εnα > t ) dt
n=1 n=1 0


∞ εnα
α p−2−α
≥ n l(n) P (|Sn | − εnα > t ) dt
n=1 0
1436 A. SHEN ET AL.


∞
≥ε nα p−2 l(n)P (|Sn | > 2εnα ) .
n=1

Hence, the complete moment convergence is more general than complete convergence.

Remark 2.2. There are many examples of slowly varying functions which are positive and
monotone non decreasing, such as l(x) = 1, l(x) = log x, and so on.

3. Preliminaries
To prove the main results of the article, we need the following lemmas.

Lemma 3.1 (cf. Liu, 2010). Let random variables X1 , X2 , . . . , Xn be END. If

f1 , f2 , . . . , fn are all non decreasing (or non increasing) functions, then random variables
f1 (X1 ), f2 (X2 ), . . . , fn (Xn ) are END.

The next one is the Rosenthal type inequality for partial sums of END random variables,
which was obtained by Shen (2011).

Lemma 3.2. Let q ≥ 2 and {Xn , n ≥ 1} be a sequence of END random variables with EXi = 0
and E|Xi |q < ∞ for each i ≥ 1. Then there exists a positive constant Cq depending only on q
such that
 n  ⎡  n ⎤
 q n 
q/2
 
E Xi  ≤ Cq ⎣ E|Xi |q + E|Xi |2 ⎦. (8)
 
i=1 i=1 i=1

Lemma 3.3 (cf. Bai and Su, 1985).

If l(x) > 0 is a slowly varying function, then
(i) limx→∞ l(tx)
l(x)
= 1 for each t > 0; limx→∞ l(x+u)
l(x)
= 1 for each u ≥ 0;
(ii) limk→∞ sup2k ≤x<2k+1 l(2k ) = 1;
l(x)

(iii) limx→∞ xδ l(x) = ∞, limx→∞ x−δ l(x) = 0 for each δ > 0;

(iv) C1 2 l(ε2 ) ≤ kj=1 2 jr l(ε2 j ) ≤ C2 2kr l(ε2k ) for every r > 0, ε > 0, positive integer k
kr k

and some C1 > 0, C2 > 0;

(v) C3 2 l(ε2 ) ≤ ∞
kr k
j=k 2 l(ε2 ) ≤ C4 2 l(ε2 ) for every r < 0, ε > 0, positive integer k
jr j kr k

and some C3 > 0, C4 > 0.

The last one is a basic property for stochastic domination. For the proof, one can refer to
Wu (2006).

Lemma 3.4. Let {Xn , n ≥ 1} be a sequence of random variables which is stochastically domi-
nated by a random variable X. Then for any α > 0 and b > 0,

E|Xn |α I (|Xn | ≤ b) ≤ C1 [E|X|α I (|X| ≤ b) + bα P (|X| > b)] (9)

and

E|Xn |α I (|Xn | > b) ≤ C2 E|X|α I (|X| > b) , (10)

where C1 and C2 are positive constants. Consequently, E|Xn |α ≤ CE|X|α .

 COMMUNICATIONS IN STATISTICS—THEORY AND METHODS 1437

4. Proofs of the main results

Proof of Theorem 2.1. Noting that ani = a+ −

ni − ani , without loss of generality, we can assume
that

n
a2ni ≤ Cn, (11)
i=1

and ani ≥ 0 for all 1 ≤ i ≤ n and n ≥ 1. It follows by (11) and Hölder’s inequality that
 n 1/2
 n 
|ani | ≤ n a2ni ≤ Cn. (12)
i=1 i=1

For fixed n ≥ 1, denote for 1 ≤ i ≤ n that

Xi(n) = −nα I(Xi < −nα ) + Xi I(|Xi | ≤ nα ) + nα I(Xi > nα ),


k
T (n) = n−α ani (Xi(n) − EXi(n) ).
i=1

It is easily checked that for any ε > 0,

 n   n 
   n  
  α α  (n) 
 ani Xi  > εn ⊂ (|Xi | > n ) ∪  ani Xi  > εnα ,
   
i=1 i=1 i=1

which implies that

 n 

∞  
 
nα p−2 l(n)P  ani Xi  > εnα
n=1
 
i=1
  n 

∞ 
n 
∞
 (n)   
 
≤ α p−2
n l(n) α
P(|Xi | > n ) + nα p−2
l(n)P T  > ε − n 
−α (n)
Eani Xi 
n=1 n=1
 
i=1 i=1
.
= I1 + I2 . (13)

In order to prove (4), it suffices to show I1 < ∞ and I2 < ∞. By (11) and Lemma 3.3, we can
get that

∞ 
n 
∞
I1 = nα p−2 l(n) P(|Xi | > nα ) ≤ C nα p−1 l(n)P(|X| > nα )
n=1 i=1 n=1

∞ ∞
 
=C nα p−1 l(n) P j < |X|1/α ≤ j + 1
n=1 j=n


∞
   α p−1
j

=C P j < |X|1/α ≤ j + 1 n l(n)

j=1 n=1
log2 j+1 2 i

∞
   
≤C P j < |X|1/α ≤ j + 1 nα p−1 l(n)
j=1 i=1 n=2i−1
log2 j+1

∞
    
≤C P j < |X|1/α ≤ j + 1 2iα p l 2i
j=1 i=1
1438 A. SHEN ET AL.


∞
   
≤C P j < |X|1/α ≤ j + 1 2(log2 j+1)α p l 2log2 j+1
j=1


∞
 
≤C P j < |X|1/α ≤ j + 1 jα p l( j)
j=1
 
≤ CE |X| p l(|X|1/α ) < ∞. (14)

In the following, we prove that I2 < ∞. First, we show that

 n 
 
 
n−α  Eani Xi(n)  → 0, as n → ∞. (15)
 
i=1

By (3), one can get that

E|X| p−γ < ∞, for any 0 < γ < p. (16)

We consider the following two cases.

Case 1: 0 < α ≤ 1.
Noting that α p > 1, we have that p > 1. Taking γ such that 0 < γ < α p−1 α
, we have by
EXn = 0, (12), (16) and Lemma 3.4 that
 n   n 
     n
−α  (n)  −α  α 
n  Eani Xi  ≤ n  Eani Xi I(|Xi | ≤ n ) + |ani |P(|Xi | > nα )
   
i=1 i=1 i=1
≤ Cn1−α E|X|I(|X| > nα ) + CnP(|X| > nα )
≤ Cn1−α E|X|I(|X| > nα )
≤ Cn1−α E|X| p−γ |X|1−p+γ I(|X| > nα )
≤ Cn1−α p+αγ E|X| p−γ → 0, as n → ∞.

Case 2: α > 1.
Taking γ such that 1 − α p + αγ < 0, we have by Lemma 3.4 again that
 n 
 
−α  (n) 
n  Eani Xi  ≤ Cn1−α E|X|I(|X| ≤ nα ) + CnP(|X| > nα )
 
i=1

n
= Cn1−α E [|X|I((k − 1)α < |X| ≤ kα )] + CnP(|X| > nα ). (17)
k=1

It follows by (16) that


∞ 
∞
α α
k1−α
E|X|I((k − 1) < |X| ≤ k ) ≤ kP((k − 1)α < |X| ≤ kα )
k=1 k=1

∞
= P(|X| > iα )
i=0

∞
E|X| p−γ
≤ 1+ < ∞,
i=1
iα p−αγ
 COMMUNICATIONS IN STATISTICS—THEORY AND METHODS 1439

which together with Kronecker’s lemma yields that


n
n1−α E [|X|I((k − 1)α < |X| ≤ kα )] → 0, as n → ∞. (18)
k=1

On the other hand, we have by (16) and Markov’s inequality that

nP(|X| > nα ) ≤ n1−α p+αγ E|X| p−γ → 0, as n → ∞. (19)

In this case, (15) follows by (17)–(19) immediately.

Hence, (15) has been proved by the statements above. It follows that for all n large enough,
 n 
 
 
n−α  Eani Xi(n)  < ε/2,
 
i=1

which implies that


∞   ε
I2 ≤ C nα p−2 l(n)P T (n)  > .
n=1
2

By Lemma 3.1, we can see that for fixed n ≥ 1, {ani (Xi(n) − EXi(n) ), 1 ≤ i ≤ n} are still END
random variables. Hence, we have by Markov’s inequality, Lemma 3.2 (taking q = 2), (11), and
Lemma 3.4 that

∞ 
n
I2 ≤ C nα p−2α−2 l(n) a2ni E|Xi(n) |2
n=1 i=1

∞
  
∞
α p−2α−1 α
≤C n l(n)E X I(|X| ≤ n ) + C
2
nα p−1 l(n)P(|X| > nα )
n=1 n=1
.
= I3 + I4 . (20)

Similar to the proof of (14), we can get that I4 < ∞. It suffices to show I3 < ∞. By Lemma
3.3 and (3) again, we have

∞
 
I3 = C nα p−2α−1 l(n)E X 2 I(|X| ≤ nα )
n=1
j+1

∞ 2−1
 
=C nα p−2α−1 l(n)E X 2 I(|X| ≤ nα )
j=0 n=2 j

∞
 
(α p−2α−1) j j
≤C 2 2 l(2 j )E X 2 I(|X| ≤ 2α( j+1) )
j=1


∞ 
j
 
≤C 2α(p−2) j l(2 j ) E X 2 I(2αk < |X| ≤ 2α(k+1) )
j=1 k=1


∞
 
≤C 2α pk l(2k )P 2αk < |X| ≤ 2α(k+1)
k=1
 
≤ CE |X| p l(|X|1/α ) < ∞, (21)

This completes the proof of the theorem. 

1440 A. SHEN ET AL.

n
Proof of Theorem 2.2. Denote Sn = i=1 ani Xi and assume that ani ≥ 0. For any ε > 0, we
have by Theorem 2.1 that

∞
nα p−2−α l(n)E (|Sn | − εnα )+
n=1

∞ ∞
= nα p−2−α l(n) P (|Sn | − εnα > t ) dt
n=1 0

∞ nα
= nα p−2−α l(n) P (|Sn | − εnα > t ) dt
n=1 0


∞ ∞
+ nα p−2−α l(n) P (|Sn | − εnα > t ) dt
n=1 nα


∞ 
∞ ∞
α p−2 α
≤ n l(n)P (|Sn | > εn ) + nα p−2−α l(n) P (|Sn | > t ) dt
n=1 n=1 nα


∞ ∞
≤C nα p−2−α l(n) P (|Sn | > t ) dt. (22)
n=1 nα

Hence, it suffices to show that

.  α p−2−α
∞ ∞
H= n l(n) P (|Sn | > t ) dt < ∞. (23)
n=1 nα

For t > 0, denote

Zti = −tI(Xi < −t ) + Xi I(|Xi | ≤ t ) + tI(Xi > t ), i = 1, 2, . . .
and
Uti = tI(Xi < −t ) + Xi I(|Xi | > t ) − tI(Xi > t ), i = 1, 2, . . .
Since Xi = Uti + Zti , it follows that
 n 

∞  
∞
 
H≤ nα p−2−α l(n) P  aniUti  > t/3 dt
n=1 n α  
i=1
 
 ∞ ∞  n 
α p−2−α  
+ n l(n) P  ani EZti  > t/3 dt
n=1 nα  
i=1
 
 ∞ ∞  n 
α p−2−α  
+ n l(n) P  ani (Zti − EZti ) > t/3 dt
nα  
n=1 i=1
.
= H1 + H2 + H3 . (24)
Note that |Uti | ≤ 2|Xi |I(|Xi | > t ). Similar to the proof of (14), we have by Markov’s inequality,
(12), Lemma 3.3, and Lemma 3.4 that
 n 
∞ ∞  
α p−2−α −1  
H1 ≤ C n l(n) t E  aniUti  dt
n=1 nα  
i=1

∞ ∞
≤C nα p−1−α l(n) t −1 E [|X| I(|X| > t )] dt
n=1 nα
 COMMUNICATIONS IN STATISTICS—THEORY AND METHODS 1441


∞ 
∞ (m+1)α
α p−1−α
=C n l(n) t −1 E [|X| I(|X| > t )] dt
n=1 m=n mα


∞ 
∞
≤C nα p−1−α l(n) m−1 E [|X| I(|X| > mα )]
n=1 m=n

∞ 
m
=C m−1 E [|X| I(|X| > mα )] nα p−1−α l(n)
m=1 n=1

∞
≤C m−1 E [|X| I(|X| > mα )] mα p−α l(m)
m=1

∞
=C nα p−1−α l(n)E [|X| I(|X| > nα )]
n=1
∞ 
∞
 
=C nα p−1−α l(n) E |X| I(m < |X|1/α ≤ m + 1)
n=1 m=n

∞
 
m
=C E |X| I(m < |X|1/α ≤ m + 1) nα p−1−α l(n)
m=1 n=1

∞
 
≤C E |X| I(m < |X|1/α ≤ m + 1) mα p−α l (m)
m=1
 
≤ CE |X| p l(|X|1/α ) < ∞. (25)

According to the proof of (25), we have by Markov’s inequality, (12), and Lemma 3.3 that
 n 

∞ ∞  
 
H2 ≤ C nα p−2−α l(n) t −1  ani EZti  dt
n=1 nα  
i=1
∞ ∞ 
n
≤C nα p−2−α l(n) t −1 E[|ani Xi | I(|Xi | > t )]dt
n=1 nα i=1
∞ ∞
≤C nα p−1−α l(n) t −1 E [|X|I(|X| > t )] dt < ∞. (26)
n=1 nα

For fixed t > 0 and n ≥ 1, it is easily seen that {ani (Zti − EZti ), i ≥ 1} are still END random
variables by Lemma 3.1. Hence, we have by Markov’s inequality, Lemma 3.2 (taking q = 2),
Lemma 3.4, (11), (25), Lemma 3.3, and (21) that
 n 2

∞ ∞  
 
H3 ≤ C nα p−2−α l(n) t −2 E  ani (Zti − EZti ) dt
n=1 nα  
i=1
∞ ∞ 
n
≤C nα p−2−α l(n) t −2 a2ni EZti2 dt
n=1 nα i=1

∞ ∞ 
n
 
≤C nα p−2−α l(n) t −2 a2ni EXi2 I (|Xi | ≤ t ) + t 2 P (|Xi | > t ) dt
n=1 nα i=1
∞ ∞  
≤C nα p−1−α l(n) t −2 E X 2 I (|X| ≤ t ) dt
n=1 nα
1442 A. SHEN ET AL.


∞ ∞
+C nα p−1−α l(n) P (|X| > t ) dt
n=1 nα


∞ ∞  
≤C nα p−1−α l(n) t −2 E X 2 I (|X| ≤ t ) dt
n=1 nα


∞ ∞
+C nα p−1−α l(n) t −1 E [|X|I (|X| > t )] dt
n=1 nα


∞ 
∞ (m+1)α  
α p−1−α
≤C n l(n) t −2 E X 2 I (|X| ≤ t ) dt
n=1 m=n mα


∞ 
∞
 
≤C nα p−1−α l(n) mα−1−2α E X 2 I (|X| ≤ (m + 1)α )
n=1 m=n
∞
 
m
=C mα−1−2α E X 2 I (|X| ≤ (m + 1)α ) nα p−1−α l(n)
m=1 n=1

∞
 
≤C mα−1−2α E X 2 I (|X| ≤ (m + 1)α ) mα p−α l(m)
m=1

∞
 
=C nα p−1−2α l(n)E X 2 I (|X| ≤ (n + 1)α )
n=1

∞
 
=C nα p−1−2α l(n)E X 2 I (nα < |X| ≤ (n + 1)α )
n=1
∞
 
+C nα p−1−2α l(n)E X 2 I (|X| ≤ nα )
n=1

∞
   
≤C n−1 E |X| p l(|X|1/α )I(nα < |X| ≤ (n + 1)α ) + CE |X| p l(|X|1/α )
n=1
  
≤ CE |X| p l |X|1/α < ∞. (27)

Hence, (23) follows by (24)–(27) immediately. This completes the proof of the
theorem. 

Proof of Theorem 2.3. We use the same notations as those in Theorem 2.1. The proof is sim-
ilar to that of Theorem 2.1. We only need to show that
 n 
 
−α  (n) 
n  Ea ni i  → 0, as n → ∞.
X
 
i=1

Since l(x) > 0 is a monotone non decreasing function, we have

  1
|X|1/α = |X|1/α I (|X| ≤ 1) + |X|1/α l |X|1/α  1/α  I (|X| > 1)
l |X|
1
≤ 1 + |X|1/α l(|X|1/α ) , (28)
l(1)
 COMMUNICATIONS IN STATISTICS—THEORY AND METHODS 1443

which together with (6) yields that E|X|1/α < ∞. Noting that 1/2 < α ≤ 1 and EXi = 0, we
have by Lemma 3.4 that
 n   n 
     n
−α  (n)  −α  α 
n  Eani Xi  ≤ n  Eani Xi I(|Xi | ≤ n ) + |ani |P(|Xi | > nα )
   
i=1 i=1 i=1
≤ Cn1−α E|X|I(|X| > nα ) + CnP(|X| > nα )
≤ Cn1−α E|X|I(|X| > nα )
= Cn1−α E|X|1/α |X|1−1/α I(|X| > nα )
≤ CE|X|1/α I(|X| > nα ) → 0, as n → ∞.
This completes the proof of the theorem. 

Acknowledgments
The authors are most grateful to the Editor-in-Chief N. Balakrishnan and three anonymous referees
for careful reading of the manuscript and valuable suggestions which helped in improving an earlier
version of this article.

Funding
This work was supported by the National Natural Science Foundation of China (11501004, 11526033,
11671012), the Natural Science Foundation of Anhui Province (1508085J06, 1608085QA02), the Key
Projects for Academic Talent of Anhui Province (gxbjZD2016005), the Provincial Natural Science
Research Project of Anhui Colleges (KJ2015A018), the Quality Engineering Project of Anhui Province
(2015jyxm045), and the Doctoral Research Start-up Funds Projects of Anhui University.

References
Bai, Z.D., Su, C. (1985). The complete convergence for partial sums of i.i.d. random variables. Sci. China
Ser. A 28(12):1261–1277.
Baum, L.E., Katz, M. (1965). Convergence rates in the law of large numbers. Trans. Am. Math. Soc.
120(1):108–123.
Chen, Y., Chen, A., Ng, K.W. (2010). The strong law of large numbers for extend negatively dependent
random variables. J. Appl. Probab. 47:908–922.
Erdös, P. (1949). On a theorem of Hsu and Robbins. Ann. Math. Stat. 20:286–291.
Hsu, P., Robbins, H. (1947). Complete convergence and the law of large numbers. Proc Nat. Acad. Sci.
33:25–31.
Liu, L. (2009). Precise large deviations for dependent random variables with heavy tails. Stat. Probab.
Lett. 79:1290–1298.
Joag-Dev, K., Proschan, F. (1983). Negative association of random variables with applications. Ann. Stat.
11(1):286–295.
Liu, L. (2010). Necessary and sufficient conditions for moderate deviations of dependent random vari-
ables with heavy tails. Sci. China Ser. A 53(6):1421–1434.
Peligrad, M., Gut, A. (1999). Almost-sure results for a class of dependent random variables. J. Theor.
Probab. 12(1):87–104.
Qiu, D.H., Chen, P.Y., Antonini, R.G., Volodin, A. (2013). On the complete convergence for arrays of
rowwise extended negatively dependent random variables. J. Korean Math. Soc. 50(2):379–392.
Shen, A.T. (2011). Probability inequalities for END sequence and their applications. J. Inequalities Appl.
2011 (Article ID 98):12.
Stoica, G. (2007). Baum-Katz-Nagaev type results for martingales. J. Math. Anal. Appl. 336(2):1489–
1492.
1444 A. SHEN ET AL.

Wang, S.J., Wang, X.J. (2013). Precise large deviations for random sums of END real-valued random
variables with consistent variation. J. Math. Anal. Appl. 402:660–667.
Wang, X.J., Hu, S.H. (2012). Some Baum-Katz type results for ϕ-mixing random variables with different
distributions. RACSAM 106:321–331.
Wang, X.J., Hu, S.H., Yang, W.Z., Wang, X.H. (2012). On complete convergence of weighted sums
for arrays of rowwise asymptotically almost negatively associated random variables. Abstract Appl.
Anal. 2012(Article ID 315138):15.
Wang, X.J., Hu, T.C., Volodin, A., Hu, S.H. (2013a). Complete convergence for weighted sums and
arrays of rowwise extended negatively dependent random variables. Commun. Stat. Theory Methods
42:2391–2401.
Wang, X.J., Wang, S.J., Hu, S.H., Ling, J.M., Wei, Y.F. (2013b). On complete convergence of weighted
sums for arrays of rowwise extended negatively dependent random variables. Stochastics Int. J.
Probab. Stochastic Processes 85(6):1060–1072.
Wang, X.J., Hu, S.H. (2014). Complete convergence and complete moment convergence for martingale
difference sequence. Acta Math. Sinica English Ser. 30(1):119–13.
Wang, X.J., Li, X.Q., Hu, S.H., Wang, X.H. (2014). On complete convergence for an extended negatively
dependent sequence. Commun. Stat. Theory Methods 43:2923–2937.
Wu, Q.Y. (2006). Probability Limit Theory for Mixing Sequences. Beijing: Science Press of China.
Wu, Q.Y. (2011). Complete convergence for weighted sums of sequences of negatively dependent ran-
dom variables. J. Probab. Stat. 2011(Article ID 202015):16.
Wu, Q.Y. (2012). Sufficient and necessary conditions of complete convergence for weighted sums of
PNQD random variables. J. Appl. Math. 2012(Article ID 104390):10.
Wu, Q.Y., Jiang, Y.Y. (2008). Some strong limit theorems for ρ̃-mixing sequences of random variables.
Stat. Probab. Lett. 78:1017–1023.
Wu, Y.F., Guan, M. (2012). Convergence properties of the partial sums for sequences of END random
variables. J. Korean Math. Soc. 49(6):1097–1110.