Strong Law of Large Numbers for a 2-Dimensional Array
of Pairwise Negatively Dependent Random Variables
Karn Surakamhaeng1, Nattakarn Chaidee1,2, Kritsana Neammanee1
1Department of Mathematics and Computer Science, Chulalongkorn University, Bangkok, Thailand 2Centre of Excellence in Mathematics, CHE, Bangkok, Thailand
Email: nattakarn.c@chula.ac.th
Received October 30,2012; revised November 30, 2012; accepted December 14,2012
ABSTRACT
In this paper, we obtain the strong law of large numbers for a 2-dimensional array of pairwise negatively dependent random variables which are not required to be identically distributed. We found the sufficient conditions of strong law of large numbers for the difference of random variables which independent and identically distributed conditions are
regarded. In this study, we consider the limit as m n which is stronger than the limit as when m, n
are natural numbers.
,
m n
Keywords: Strong Law of Large Numbers; Negatively Dependent; 2-Dimensional Array of Random Variables
1. Introduction and Main Results
Let
Xi i N be a sequence of random variables. We say
Xthat i i N satisfies the strong law of large numbers (SLLN) if there exist sequences of real numbers
nn N
a
and
n such thatn N
b n n a.s. 0
n
S a b
as .
where and the abbreviation a.s. stands for
n
1
n
n i
S Xi
almost surely.
To study the strong law of large numbers, there is a simple question come in mind. When does the sequence
i N satisfy the SLLN? Many conditions of the se-
quence i N have been found for this question. The
SLLN are investigated extensively in the literature espe- cially to the case of a sequence of independent random variables (see for examples in [1-3]). After concepts of dependence was introduced, it is interesting to study the SLLN with condition of dependence.
Xi
Xi A sequence i N of random variables is said to be
pairwise positively dependent (pairwise PD) if for any
and i j,
Xi
,
a bR
i , j
i
j
P X a X b P X a P X b
and it is said to be pairwise negatively dependent (pair- wise ND) if for any a b, R and i j,
i , j
i
j
.P X a X b P X a P X b
Theorem 1.1-1.5 are examples of SLLN for a sequence of pairwise PD and pairwise ND random variables.
Theorem 1.1. (Birkel, [4]) Let i N be a se- quence of pairwise PD random variables with finite variances. Assume
Xi1) sup
i
i
,i N
E X E X
2)
2
1
Cov ,
.
n
i i i
X S i
Then Sn E S
n a.s. 0n
as n .
Theorem 1.2. (Azarnoosh, [5]) Let i i N be a se- quence of pairwise ND random variables with finite variances. Assume
X1) sup
i ,i N
E X
2)
21 Var
.
n
i i
X i
Then Sn E S
n a.s. 0n
as n .
Theorem 1.3. (Nili Sani, Azarnoosh and Bozorgnia, [6]) Let
an n N be a positive and increasing sequencesuch that an as n .
Let
Xi i N be a sequence of pairwise ND randomvariables with finite variances such that
1)
1
sup n i i ,
n N i n
E X E X
a
2)
21
Var
. i i i
X
a
Then n
n a.s. 0n
S E S
a
as n .
In this work, we study the SLLN for a 2-dimensional array of pairwise ND random variables. We say that
Xi j i j N, , satisfies the SLLN if there exist double se- quences of real numbers
am n, m n N, and
bm n, m n N,such that , , a.s.
,
0 m n m n
m n
S a
b
as where
,
m n
, ,
1 1
.
m n m n i j
i j
S X
In 1998, Kim, Beak and Seo investigated SLLN for a 2-dimensional array of pairwise PD random variables and it was generalized to a case of weighted sum of 2-dimensional array of pairwise PD random variables by Kim, Baek and Han in one year later. The followings are their results.
A double sequence
,,
i j i j N
X
is said to be pairwise
positively dependent (pairwise PD) if for any a b, R
and
i j, k l, ,
i j, , k l,
i j,
k l,
P X a X b P X a P X b
Theorem 1.4. (Kim, Beak and Seo, [7]) Let
Xi j i j N, , be a 2-dimensional array of pairwise PD random variables with finite variances. Assume1)
,
,
,
sup i j i j , i j N
E X E X
2)
, ,
2
, , 1 , ,
Cov ,
.
i j k l i j i j k l k l i j
X X
i j
Then Sm n, E S
m n, a.s. 0 m n
as m n .
Theorem 1.5. (Kim, Baek and Han, [8]) Let
ai j, i j N, be a 2-dimensional array of positive numbers andsuch that ,
1 1
n m m n i j
i j
b
a, ,,
0 m n m n
a
b and bm n, as
, .
m n
Let be a 2-dimensional array of pairwise
PD random variables with finite variances such that
Xi j, i j N,1)
,
,
,
sup i j i j , i j N
E X E X
2) , , 2
, ,
, , 1 , , ,
Cov ,
.
i j k l i j k l i j i j k l k l i j i j
a a X X
b
Then as m n where
, X
double indexed sequence of real ,
m n
,1 1
.
n m i j i j i j
W a
Observe that, for a number
am n, m n N, ,
, , a.s.
,
0
m n m n m n
W E W
b
m n the convergence as
im
nce
plies the convergence as m n, . However, a dou-
ble seque
a ,,
m n m n N where
,
m n
1m n
a
m n
Our goal is t he SLLN -
A double sequence
m n
shows us that the conv o obtai ray of random variab
erse is not true in general.
n t for 2-dimensional ar
les in case of pairwise ND.
Xi j, i j N, is said to be pairwise negtively dependent (pairwise ND) if for any a b, R and
i j, k l, ,
i j, , k l,
i j,
k l,
.P X a X b P X a P X
The
b
followings are SL of pairwise ND ran sults.
LNs for a 2-dimensional array dom variables which are all our re-
Theorem 1.6. Let
am m N and
bn n N be increas-ing sequences of positive numbers such that ,
w m n
a b e
hich am as m and bn as n .
Let
,,
i j i j N
X be a 2-dimensional array of
ND ran nite va . If th st
pairwise
with fi riances ere exi
dom variables real numbers p q, such that
,1 1 2
Var
,
i j p q i j
X
a b
2
i j
nce such that
then for any double seque
,,
m n m n
c N
2 2
,
p q m n m n
c a b for every ,m nN,
, , a.s.
,
m n
c 0
m n m n
S E S
as n .
The next theorem is the SLLN for the difference of random variables which independent and identically dis- tributed conditions are regarded.
m
Theorem 1.7. Let
Xi j i j N, , and
Yi j, i j N, be 2- dimensional arrays of random variables on a probability spthen
ace (Ω, F, P). If
, ,
1
,
i j i j i
P X Y
1
j
a.s., ,
1 1
1
0
m n
i j i j i j
X Y
m n
Corollary 1.8 and Corollary 1.9 follow directly from
Theorem 1.6 by choosing nd
b where with p = q =
ing sequences of positive nu an
a 2-dim l array
such
,
k m n m n
c a b
3
a m and bn3n
a ,
m n
4, r m n .
c a
espectively m
Corollary 1.8. Let
am and
b N be increas-mbers such that a bm, ne
which am as d as n .
m N n n
m bn
Let
,,
i j i j N
X be ensiona of pairwise
ND random variables with finite variances. If there exist
,
p qN that
, Var Xi j
1 1 2 2
,
p q i j
a b
i j
then for any k p q,
, , a.s. 0
m n m n k m n
S E S
a b
as m n .
Corollary 1.9. Let
Xi j i j N, , be a 2-dimensional ar-ray of pairwise ND random va s with finite vari-
ances. If
riable
, 2
1 1
,
i j i j
then
Var Xi j
, , a.s.
2 0
81
m n m n
S E S
m n
as m n .
2. Auxiliary Results
In this section, we present some materials which will be used in obtaining the SLLN’s in the next section.
icz, [9]) Let
Proposition 2.1. (Mór
i j, i j N, be aand
double sequence of positive numbers such that for all
, ,
i jN
1, , 0, , 1 , 0
i j i j i j i j
1, 1 1, , 1 , 0,
i j i j i j i j
,
i j
Let
aas
ence of r numbers.
Assume that
max ,i j .
be a double sequ
, ,
i j i j N eal
1) ,
1 1 ,
, i j
2)
i i j
a
j
,
i k
a
1 ,
k i k
for every iN and ,
1 ,
k j k k j
a
forevery jN. Then ,
1 1
m n i j i j
a
,
1 0
m n
as
max m n, .
The
for a sequence of d
osition
following proposition is a Borel-Cantelli lemma ouble indexed events
t
Prop 2.2. Le
Ei j, i j N, be a sequence probability spacedouble
of events on a
, ,F P
.Then
i j,
i j, i.o
P E P E
1 1
i j
. 0
where
Ei j, i.o.
Ei j,
1 , ,
.
k i j i j k
Proof. Let
,1 1
.
i j i j
L P E
LR be such that
First note that
,
,
1 1
k k k k
i j i j i j i j
P E P E P E
,1 1 , ,
i j i j i j k
where
k
denote the greatest integer smaller than or
equal k and hence
1 1 , ,
1 1
lim
k
i j i j i j k k k
k i j
P
, ,
,
lim lim
. k k
i j i j
k
i j
L P E P E
E L
Therefore
,, ,
lim i j k
i j i j k
P E L
and
,
, ,
, , , ,
, ,
1 1 , , 1
i.o. lim lim
lim 0.
i j i j i j
k k i j i j k i j i j k
i j i j k
i j i j i j k
P E P E P E
P E E
P
This completes the proof. □
3. Proof of Main Results
Proof of Theorem 1.6
Let m n, N and define f m
lnam and
ln n .g n b
Clearly, f and g are increasing whose facts
ln m
1f m a f m and g n
lnbn g n
1 1
f m
e and
which imply that f m
e
g g n m
a
n 1
n
e b e .
Let 0 be given. By using the fact that
, ,
ov Xi j,Xk l 0
C for
i j, k l, ([10], p. 313), weFrom this fact and Chebyshev’s ineq have
Var m n , Var , .
m n
i j i j
X X
1 1 1 1
i j i j
, ,1 1 ,
1 1 , 1 1 , 1 1
, 2
1 1 ,
, 1 1 , 1 1 Var 1 Var 1 Var .
m n m n m n m n
m n m n m n m ni j
i j
i j m i n j m n
i j p q i j m i n j m n
i j pf m qg n j m i n j
S E S P c c c C X c C X a b C X e
1 (3.1) i
For each let
, , 2 2 Var 1 Var m n m n i j S
C C X
, ,
i jN
1 1 : : f s i i f t j jA s N e a
B t N e b
and iminAi and jminBj . Since iAi and j
jB (3.1
, we have ), we have
ii and j j. From this facts and
, , 1 , 1 1 ,1 1 2
,
1 1 2 2
, 1 1 2 1 1 Var 1 1 Var 1 Var
m n m n
i j pf m qg n m i n j
i j pm qn i j m f i n g j
i j pm qn i j m f i n g j
i j pf i i j
S E S
P c X e C X e C X e e C X e
1 , Varm n m n
i j C
21 . (3.2)
qg j e
Since iAi and jBj, we have
1 1 f i i e a e
and
1 1 g j j e e assum 2) ve b .
From this facts and (3.2) together with our ption
, we ha
,
m n
c
, ,
1 1 , 1 1 2 2
Var
.
m n i j
p q
m n m n i j
S E S X
P C
ition 2.2 with
i j a b By Propos
, , , , ,m n m n m n
m n
S E S E c
m n, i.o.
0P E
we have and this hold for every 0.
, p. 77)
By us ith Theorem 4.2.2 ([11]
we can prove that
ing the same idea w ,
, , a.s.
m n m n
S E S ,
0
m n
c as m n . □
Proof of Theorem 1.7
Let 0
By Proposition 2.2,1 , ,
.
k i j i j k
X Y
i j, i j,we have
c 1P 0 P
0 1 P
Xi j, Yi j,
i.o.
1.0,
c
For every we will show that
,1 1 i j
i j X i j
,
1 , i j Y (3.3)
fo ry
r eve iN,
, ,
1
1
i j i j j
X Y
i j
(3.4)and for every jN,
, ,
1
1 .
i j i j i
X Y
i j
(3.5)From (3.3), (3.4) and (3.5), we can apply Proposition 2.1 with i j, i j that
a.s., ,
1 1
1
0
m n
i j i j i j
X Y
m n
as max
m n, . We here note that am n, a as
x m n, imp
ma lies am n, a as m n . Hence
a.s., ,
1 1 i j i j
i j
m n
as m n
1
0
m n
X Y
.
To prove (3.3), (3.4) and (3.5), let Then
there exists 0
. c
kN such that for i j, N,
, , .
i j i j
X Y
i j k (3.6)
Thus for each c
0, Xi j, i j N,
and
Yi j,
i j N, are different only finitely many terms.plies th t (3.3) holds.
This im a
For fixed iN, we can find a large j0N such
that (3.6) holds for all j j0 which m
are only finitely many different terms of
eans that there
, ,
i j i j N
X
Y
. So for fixedand ,
,
i j i j N iN,
, ,
1
i j i j
X Y
i j .
1
j
, , a.s.
,
0
m n m n m n
S E S
c
Similarly, for fixed jN,
, ,
1 .
i j i j
X Y
1
i i j
Now ) are now pr
proo
(3.4) and (3.5 oved and this ends the
f. □
Remark 3.1. In case of m fixed and by con-
sidering the limit as we the
corresponding results for a case of 1-dime nal pair-
wise N
4. Example
Example 4.1 A box contains pq balls of p different col- or
,
n
also obtain
,
m n
nsio D random variables.
s and q different sizes in each color. Pick 2 balls ran- domly.
Let Xi j, ,i1, 2, , p and j1, 2, q random
variable indicating th of a ball of the ith color
and the jth size such that
1,if theithcolo
,
ce
be a presen
e
,
0,other
r and the size of ball is picked, wise.
th i j
j X
For i j, N, let Xi j, be a random variable defined by
, ,
1 ,
0, otherwise. i j
i j
,if and 1
X i q
X
p j
Proof. By a direct calculation, we have Xi j, ’
,
k l R
s are pairwise ND random variables, i.e. for i j, , that
and
i j, k l, a b, R,
i j, , k l,
i j,
k l,
.P X a X b P X a P X b
Note that
,1 2 2
i j
pq E X
pq pq
and
, 2
2 4 . i j
pq
Var X
pq
Hence,
, ,
2 , 2
1 1 1 1
2 2
1 1
Var Var
lim
2 4 1
.
m n
i j i j
m n
i j i j
i j
X X
i j i j
pq pq i j
By applying Theorem 1.6, for any double sequence
that n for ever
nN, we have as m n .
5. Acknowledgements
have hel
ork. hor gives an ap
thanks to the Institute for the Prom hnology for financial su
REFERENCES
V. Totik, “On th□
otion of Teaching
e Strong Law
Th comm our w
Scienc
e authors would like to thank referees for valuable
ents and suggestions which ped improving
The first aut preciation and
e and Tec pport.
[1] S. Csörgő, K. Tandori and
of Large Numbers for Pairwise Independent Random Variables,” Acta Mathematica Hungarica, Vol. 42, No. 3- 4, 1983, pp. 319-330. doi:10.1007/BF01956779
[2] N. Etemadi, “An Elementary Proof of the Strong Law of Large Numbers,” Zeitschrift für Wahrscheinlichkeitstheo- rie und Verwandte Gebiete, Vol. 55, No. 1, 1981, pp. 119-122.
R. G. Laha and ility Theory,” John Wiley & Sons
[3] V. K. Rohatgi, “Probab , Hoboken, 1979.
[4] T. Birkel, “A Note on the Strong Law of Large Numbers for Positively Dependent Random Variables,” Statistics & Probability Letters, Vol. 7, No. 1, 1989, pp. 17-20.
doi:10.1016/0167-7152(88)90080-6
[5] H. A. Azarnoosh, “On the Law of Large Numbers for Ne- gatively Dependent Random Variables,” Pakistan Journal of Statistics, Vol. 19, No. 1, 2003, pp. 15-23.
[6] H. R. Nili Sani, H. A. Azarnoosh and A. Bozorgnia, “The Strong Law of Large Numbers for Pairwise Negatively Dependent Random Variables,” Iranian Journal of Sci-
pend- the Korean Mathe-
imensional ence & Technology, Vol. 28, No. A2, 2004, pp. 211-217. [7] T. S. Kim, H. Y. Beak and H. Y. Seo, “On Strong Laws
of Large Numbers for 2-Dimensional Positively De ent Random Variables,” Bulletin of
matical Society, Vol. 35, No. 4, 1998, pp. 709-718. [8] T. S. Kim, H. Y. Beak and K. H. Han, “On the Almost
Sure Convergence of Weighted Sums of 2-D
Arrays of Positive Dependent Random Variables,” Com- munications of the Korean Mathematical Society, Vol. 14, No. 4, 1999, pp. 797-804.
[9] F. Móricz, “The Kronecker Lemmas for Multiple Series and Some Applications,” Acta Mathematica Academiae Scientiarum Hungaricae, Vol. 37, No. 1-3, 1981, pp. 39- 50. doi:10.1007/BF01904871
[10] N. Ebrahimi and M. Ghosh, “Multivariate Negativ
c ,,
m n m n N such
2
, 81
m n
c m y m
e De- pendence,” Communications in Statistics—Theory and Methods, Vol. A10, No. 4, 1981, pp. 307-337.