http://dx.doi.org/10.4236/am.2014.510153
How to cite this paper: Wang, Y.F. and Wu, Q.Y. (2014) A Note on the Almost Sure Central Limit Theorem in the Joint Ver-sion for the Maxima and Partial Sums of Certain Stationary Gaussian Sequences. Applied Mathematics, 5, 1598-1608. http://dx.doi.org/10.4236/am.2014.510153
A Note on the Almost Sure Central Limit
Theorem in the Joint Version for the Maxima
and Partial Sums of Certain Stationary
Gaussian Sequences
*
Yuanfang Wang, Qunying Wu
College of Science, Guilin University of Technology, Guilin, China Email: wangyuanfang89@126.com, wqy666@glut.edu.cn
Received 29 April 2014; revised 29 May 2014; accepted 5 June 2014 Copyright © 2014 by authors and Scientific Research Publishing Inc.
This work is licensed under the Creative Commons Attribution International License (CC BY).
http://creativecommons.org/licenses/by/4.0/
Abstract
Considering a sequence of standardized stationary Gaussian random variables, a universal result in the almost sure central limit theorem for maxima and partial sum is established. Our result ge-neralizes and improves that on the almost sure central limit theory previously obtained by Marcin Dudzinski [1]. Our result reaches the optimal form.
Keywords
Almost Sure Central Limit Theorem, Stationary Gaussian Sequence, Slowly Varying Functions at Infinity
1. Introduction
In this paper, we let
(
X X, n n)
∈ be a standardized stationary Gaussian sequence, also let1
: max ,
n i
i n
M X
≤ ≤
= ,
1
: max
m n i m i n
M X
+ ≤ ≤
= ,
1
:
n n i
i
S X
=
=
∑
,σ
n:= Var( )
Sn ,1
n n k
k
D d
=
=
∑
,{ }
dkbe some sequence of weights, I . and
( )
Φ( )
. denote the indicator function and the standard normal distribu- tion function, respectively.*Supported by the National Natural Science Foundation of China (11361019) and Innovation Project of Guangxi Graduate Education
The ASCLT has been first introduced independently by Brosamler [2] and Schatte [3] for partial sum, since then the concept has already started to receive applications in many fields. For example, Fahrner and Sadtmuller [4] and Cheng et al. [5] extended this almost sure central limit theorem for partial sums to the case of maxima of i.i.d. random variables. Under some conditions, they proved as follows:
(
)
(
)
( )
1
1 1
lim ln
n
k k k n
k
a S b x G x
n k
→∞
∑
= Ι − ≤ = a.s.,for all x∈R, where ak >0 and bk∈R satisfy
(
)
dk k k
a S −b →G, where G(:) is some non-degenerate distribution function. Afterwards, Marcin Dudzinski [5] showed the ASCLT in its two-dimensional version, i.e.
(
)
( )
1
1
lim , e Φ
n
k k k k k n
k
n k
S
d I a M b x y y
D
τ
σ −
→∞ =
− ≤ ≤ =
∑
a.s. for ∀x y, ∈. (1.1)In this paper, we extend the weight dk 1 k
= to the weight
(
( )
)
exp ln
k
k d
k β
= , 0 1
2
β
≤ < .Our purpose is to prove that if
(
X X, n n N)
∈ is a standardized stationary Gaussian sequence, the covariance function r t( )
fulfills( )
L t( )
r t tα
= , α>0 and t=1, 2,, (1.2)
where L t
( )
is a positive slowly varying function at infinity. Moreover if the numerical sequence{ }
un satis- fies the following relation( )
(
)
lim 1 n
n→∞n − Φ u =τ for some 0≤ < ∞τ
, (1.3)
then we have
(
)
( )
1
1
lim , e
n
k k k k k
n k
n k
S
d I a M b x y y
D
τ σ
−
→∞ =
− ≤ ≤ = Φ
∑
a.s. for any x y, ∈R, (1.4)where
1
n n k
k
D d
=
=
∑
and(
( )
)
exp ln
k
k d
k β
= , 0 1
2
β
≤ < .In the following, denote by an ∼bn if n 1
n a
b → as n→ ∞, by anbn if there exists a constant c>0 such that an≤cbn for sufficiently large n. The c stands for a constant, which may vary from one line to another.
2. Main Results
Theorem 2.1. Assume that
(
X X, n n)
∈ be a standardized stationary Gaussian sequence, the covariance func- tion r t( )
satisfies (1.2) for some α >0. If the numerical sequence{ }
un satisfies (1.3), then (1.4) holds with( )
(
)
exp ln
k
k d
k β
= , 0 1
2
β
≤ < .Let
{ }
n nD= D ∈ be a positive non-decreasing sequence with lim n
n→∞D = ∞. We say that
{ }
xk k∈ is D-summable to a finite limit x if
1 1 lim
n
k k n
k n
d x x
D
→∞
∑
= = , (2.1)Remark 2.1. By a classical theorem of Harly (see Chandrasekharan and Minakshisundaram [6]), if two se- quences D and D* satisfy Dn∗=O D
( )
n , then D-summability implies the D*-summability, i.e., if a se- quence{ }
nn N
x ∈ is D-summable to x, then it is also *
D -summable to x. These results show that if (2.1)
holds with the weight
{ }
k kd ∈, then for 0 dk dk
∗
≤ ≤ , Dn∗ → ∞, (2.1) also holds with the weight
{ }
k k d∗∈. So,
if ASCLT holds with
{ }
k kd ∈, then for 0 dk dk
∗
≤ ≤ , Dn∗→ ∞, ASCLT also holds with
{ }
k k d∗∈. So, by this,
if we use larger weights, we should expect to get stronger results.Theorem 2.1 remains valid if we extend the
weights from dk 1 k
= to
(
( )
)
exp ln
k
k d
k β
= , 0 1
2
β
< < . When
β
=0, the weights dk 1 k = .Lemma 2.2. Under the same assumptions as in Theorem 2.1, if the numerical sequence
{ }
un fulfills (3), then there exists someγ
>0 and for any x y, ∈ and m<n,,
, n , n
n n m n n
n n
S S m
E I M u y I M u y
n γ
σ
σ
≤ ≤ − ≤ ≤
, (2.2)
,
, m , , n ,
m m m n n
m n
S S m
Cov I M u y I M u y
n γ
σ
σ
≤ ≤ ≤ ≤
(2.3)
where m m, n n.
m n
x x
u b u b
a a
= + = +
3. Proofs
3.1. Proof of Theorem 2.1
Under the assumptions of Theorem 1 on X X1, 2,, r t
( )
and{ }
un , by Theorem 4.3.3 in Leadbetter et al. [7], we have lim(
n n)
en P M u
τ −
→∞ ≤ = for τ which is defined in (1.3). Let y be a real number, for each n≥1,
n
Y denotes a standard normal variable, which is independent of
(
X1,,Xn)
and has the same distribution as nn S
σ . From the proof of Lemma 2.2, we get that
(
) (
)
1
, n 0
n n n n n
n S
P M u y P M u P Y y
nγ
σ
≤ ≤ − ≤ ≤ →
, as
n→ ∞. Thus we have
(
) (
)
( )
lim , n lim e
n n n n n
n n
n
S
P M u y P M u P Y y τ y
σ −
→∞ →∞
≤ ≤ = ≤ ≤ = Φ
,
by Toeplitz Lemma, we obtain
( )
1
1
lim , e
n
k k k k n
k
n k
S
d P M u y y
D
τ σ
− →∞ =
≤ ≤ = Φ
∑
(3.1)where
1
n n k
k
D d
=
=
∑
and(
( )
)
exp ln
k
k d
k β
= , 0 1
2
β
< < .To prove Theorem 2.1 for 0 1 2
β
< < . We should prove the following:
1 1
, , 0
n
k k
k k k k k
k
n k k
S S
d I M u y P M u y
D =
σ
σ
≤ ≤ − ≤ ≤ →
∑
a.s. (3.2)[8]) for some ε>0
(
)
(1 )2
1
Var , ln
n
k
k k k n n
k k
S
d I M u y D D ε
σ
− +
=
≤ ≤
∑
(3.3)Set , k , k
k k k k k
k k
S S
I M u y P M u y
ξ
σ σ
= ≤ ≤ − ≤ ≤
, by Lemma 2.2, we have
(
)
,
,
, , ,
, , , ,
, , ,
,
k l
k l k k l l
k l
k l l
k k l l k l l
k l l
k l
k k k l l
k l
l
l l
l
S S
E Cov I M u y I M u y
S S S
Cov I M u y I M u y I M u y
S S
Cov I M u y I M u y
S
E I M u y
ξ ξ
σ σ
σ σ σ
σ σ
σ
= ≤ ≤ ≤ ≤
≤ ≤ ≤ ≤ ≤ − ≤ ≤
+ ≤ ≤ ≤ ≤
≤ ≤
,
, ,
, , ,
for 1 .
l k l l
l
k l
k k k l l
k l
S
I M u y
S S
Cov I M u y I M u y
k
k l l
γ
σ
σ σ
− ≤ ≤
+ ≤ ≤ ≤ ≤
≤ <
We know that
(
)
( )2 ( )2
2
1 1 1
1 2
1 , ln 1 , ln
Var , 2
:
n n
n n
k
k k k k k k l k l
k k k k l n
k l k l n n
k k
k l n D k l n D
l l
S
d I M u y E d d d E
k
d d d d T T
l
γ γ
γ
ξ ξ ξ
σ
− −
= = ≤ ≤ ≤
≤ ≤ ≤ ≤ ≤ ≤ ≤ >
≤ ≤ ≤ ≤
+ = +
∑
∑
∑
∑
∑
For Tn1, we have the following estimate:
( ) 2 ( )2
(
)
(
)
(
)
2 2
1 2 2 1
1 , ln 1 , ln
1
.
ln ln ln
γ γ
γ
ε
− −
+
≤ ≤ ≤ ≤ ≤ ≤ ≤ ≤
≤
∑
∑
n n
n n
n k l k l
k k n n n
k l n D k l n D
l l
D D
k
T d d d d
l D D D
By the elementary calculation, it is easy to see that Dn 1
( )
lnn 1 βexp ln(
( )
n)
β β−
∼ , lnDn∼
( )
lnn β,ln lnDn ∼ln lnn.
For 0 1
2
β
< < , we have : 1 2 0 2
β ε
β −
= > and 1 1
2β = +ε. Then
( )
( )
(
)
( )
(
)
(
)
(
) (
)
(
)
(
)
2 2
2
2 1 1
1 1
1 , ln ln
2 2 2
1 1 2 1 1
2 2 2
exp ln ln ln
ln ln
ln ln
ln ln
ln
ln ln ln
γ λ
β
β β
β β
β ε
β β β
− −
− −
≤ ≤ ≤ =
≤ ≤ ≤ > ≤ <
− +
= =
∑
∑
∑
∑
n n
n
n n n
n k l k k n
k l n k
k
k l n D k l D n n
l
n n n n
n
n n n
n D D D
T d d d d D
l
D D
D D D D
D
D D D
.
3.2. Proof of Lemma 2.2
We first consider that (1.2) holds with some 0< <α 1. Let 1≤ ≤i n, we have
(
)
(
)
( )
1
1 1 1
1
1
1 1
0 , , 1
2 2
for some 0 1
n i n
n
i i j
j j j i
n n n
n
t
n n
S
Cov X Cov X X r i j r j i
L t tα
σ σ σ
α
σ σ
−
= = = +
−
=
< = = + − + −
+ < <
∑
∑
∑
∑
. (4.1)
By the application of the Karamata’s theorem (see Mielniczuk [9]), we obtain
(
)(
)
( )
1
2 1 1
2 2 2
1 2
n L n n
α σ
α α
−
∼
− −
. (4.2)
From (4.1) and (4.2), for some 0< <α 1, we have
( )
( )
( )
( )
(
)
( )
(
)
11 2
1
1
1 1 1
1
2 2 2 2
2
1 ,
n n
i
t n
L n
L t L n n
S Cov X
t
L n n L n n n
α
α α α α
σ
− −
− = −
=
∑
.Since L n
( )
is a slowly varying function at infinity, for any ε>0, L n( )
nε. So2 2
1 , n
i n
S n
Cov X
n n
ε
α α ε
σ −
=
for any ε>0 and some 0< <α 1. Hence,
1
1
0 sup , n
i i n n
S Cov X
nβ σ
≤ ≤
<
→
0
, as n→ ∞, for some1
0 ,
2
β
< < (4.3)
so there exist λ, n0 such that
1
0 sup , n 1
i i n n
S
Cov X λ
σ
≤ ≤
< < <
, for any n>n0. (4.4)
By (1.2), lim
( )
lim( )
0t t
L t r t
tα
→∞ = →∞ = , thus there exist δ , n1 such that
( )
1
0 sup 1
t n
r t δ >
< = < . (4.5)
Let y be an arbitrary real number and m<n. Suppose that Yn is a random variable, which has the same
distribution as n n S
σ , but Yn is independent of
(
X1,,Xn)
, then we have(
) (
)
(
)
(
)
(
)
(
)
,
,
, ,
,
1 2 3
, ,
, ,
,
,
:
n n
n n m n n
n n
n n
m n n n n
n n
n
n n n n n
n n
m n n m n n n n
m n n n n
S S
E I M u y I M u y
S S
P M u y P M u y
S
P M u y P M u P Y y
S
P M u y P M u P Y y
P M u P M u
A A A
σ σ
σ σ
σ
σ
≤ ≤ − ≤ ≤
= ≤ ≤ − ≤ ≤
≤ ≤ ≤ − ≤ ≤
+ ≤ ≤ − ≤ ≤
+ ≤ − ≤
= + +
By Theorem 4.2.1 in Leadbetter et al. (1983) and (4.3), (4.4) and (4.5), we get that
(
)
(
)
2 2
1 2
1
1
, exp exp
2 1 2 1
n
n n n
i
i n
S u u
A A Cov X n
nβ
σ
λ
λ
=
+ − + − +
∑
. (4.6)
As we know
{ }
un fulfills (1.3), which implies that2 2π
exp 2
n n
u c u
n
− ∼
,
(
)
1 2 2 ln
n
u ∼ n ,
(
)
( )
(1 )
2 2 1
1 1 ln exp
2 1
n n
u
n λ
λ λ
+
+
−
+
,
combine this and (4.6), we have
( )
2 1(1 )( )
2 1(1 )1 2 1 1
1
1 1
ln ln
1 n n
A A n
n
n n
λ λ
β β
λ λ
+ +
+ −
+ +
+ = .
We set 1 1 0
1
β λ + − >
+ , thus
1 2
1 m
A A
n n
γ
γ
+
for some 0 1 1
1
γ β λ < < + −
+ . (4.7) Next, we estimate A3.
(
)
(
)
(
)
( )
(
)
( )
( )
( )
3 ,
,
1 1
1 1
1 2 3
:
m n n n n
n n
m n n n n n n
i m i
n n
n n
i m i
A P M u P M u
P M u u P M u u
u u
B B B
= + =
= + =
= ≤ − ≤
≤ ≤ − Φ + ≤ − Φ
+ Φ − Φ
= + +
∏
∏
∏
∏
. (4.8)
Let δ is defined as in (4.5), set
2
α δ
α <
− such that
2
2 0
1+δ + − >α . By Theorem 4.2.1 in Leadbetter et
al. and (1.3), we have
( ) ( )
( ) ( )
( )
1 1
2 1 1
1 2 2 2
1 2
1 1 1 1
ln ln
exp 1
n n
n
t t
n L t n L n
u
B B n r t
t
n n
δ δ
α α
δ δ
δ
+ +
− + −
= + = +
+ −
+
∑
∑
,
since L n
( )
and lnn are slowly varying functions at infinity, we have( )
( )
1 1
lnn +δ L n nε for any ε>0, then
1 2
1
for some 0 m
B B
n n
γ
γ γ
+ >
. (4.9)
Meanwhile, following from the elementary inequality xn m xn m n − − ≤
, 0≤ ≤x 1. We get
3 for
m m
B m n
n n
γ
≤ <
. (4.10)
By (4.8), (4.9) and (4.10), if (1.2) holds with 0< <α 1, then (2.2) holds.
Provided (1.2) also holds with some α≥1, since r
( )
. is positive, we get( )
1 2 Var
n Sn n
( )
1 1
1 2
2
0 ,
n n
i
t n
L t S
Cov X
t n
α σ
−
=
<
∑
. As( )
1
n
t L t
tα =
∑
is a slowly varying function infinity, we get( )
1
n
t L t
n t
ε α =
∑
forany ε >0. We obtain that (4.3) and (4.4). By using Theorem 4.2.1 in Leadbetter et al. and by (4.6)-(4.10), let
1
A - A3 be replaced by C1 - C3 in (4.3), we also get C1 C2 m n
γ
+
, let D1 - D3 be replaced by B1-
3
B in (4.8), we also obtain that
( ) ( )
1
2 1
1 2 2
1
1 1
ln exp
1
n n
t
n
u m
D D n r t n
n n
γ δ ε
δ δ
+
−
= +
+ −
+
∑
for some
γ
>0,like (4.11), D3 m m
n n
γ
≤
, we get C3 m
n γ
, at last , we obtain when (1.2) holds with α≥1, then (2.2)
also holds.
Next, we prove (2.3). First we consider the situation of some 0< <α 1. Let i≥ +m 1. Since L t
( )
is a posi- tive slowly varying function, then we have(
( )
)
1 2
L m mε for any ε >0, L t
( )
tα is monotonic decreasing,
so
( )
( )
1
1
i m
t i m t
L t L t
tα tα
−
= −
∑
∑
=, by (1.2) and (4.2), we obtain
( )
( )
( )
( )
( )
( )
( )
( )
1 1
1 1 2 2
1 1
2
1 1 1 1
1
2 2 2
2 2
2
1 1
0 ,
1
1 1
for some 0 2
i i
m i
t i m t i m m m
m t
L t L t
S Cov X
t t
L m m
L t L m m L m
t
L m m L m m m
m
m m
α α α
α
α α α α
ε
α η
σ σ
η
− −
−
= − = −
−
− = −
<
=
< <
∑
∑
∑
,
then there exist numbers µ,m0, such that
1
0 sup , m 1
i i m m
S
Cov X µ
σ
≥ +
< < <
for all m>m0. By the Normal Comparison Lemma, we obtain
(
)
(
(
)
)
,
, ,
2 2 2
1 1 1
, , ,
, , , , ,
exp , exp
2 1
m n
m m m n n
m n
m n m n
m m m n n m m m n m
m n m n
m n m
m n n m
i
i j m i n
S S
Cov I M u y I M u y
S S S S
P M u y M u y P M u y P M u y
u u S u
r j i Cov X
r j i
σ σ
σ σ σ σ
σ
= = + =
≤ ≤ ≤ ≤
= ≤ ≤ ≤ ≤ − ≤ ≤ ≤ ≤
+
− − + −
+ −
∑ ∑
∑
2
1
1 2 3 4
2 1 ,
, exp ,
2 1 ,
:
n i
n n
m n m n
i
i m m m m n
i m
S Cov X
S u S S
Cov X Cov
S Cov X
E E E E
σ
σ σ σ
σ
= +
+
+ − +
+
= + + +
∑
By (4.5), set 1, 2 α δ α < <
− then
2
2 0
1+δ + − >α , we have
(
)
(
(
)
)
(
)
( )
(
)
( )
(
)
( )
( )
2 2 2 2
1
1 1 1 1
2 2 2 2
1 1 1 1 1 1 1 2 2 1 exp exp 2 1 2 1 exp exp
2 1 2 1
1
m n m n i
m n m n
i j m i t m i
n i n
m n m n
t m i t
u u u u
E r j i r t
r j i
L t L t
u u u u
m m t t m m L n n n n α α δ α δ δ δ δ − = = + = = + − − = + − = − − + − + + + + = − − − + + − + + − − + +
∑ ∑
∑ ∑
∑
∑
1 m for some 0 1
n γ δ γ + < < .
By (4.4), set 1
1 β λ
β < <
− , then
1
1 0
1
β
λ − + − <
+ , we have
(
)
(
)
( )(
)
( ) 2 2 1 1 12 1 1
2 1
1
1 1
1
1 2 1
, exp
2 1 ,
exp , ln
2 1
1
ln for 0
2
m
n m
i
i n n
i n m m n i i n S u
E Cov X
S Cov X
u S m
Cov X m
n
m m m
m m n n n λ λ β β γ β β λ λ β σ σ λ σ γ β = − + + = − + − + + = − + − +
< < <
∑
∑
. and(
)
(
)
( )
( )
( )
1 1 1 1 1
1 1 1 1 1
1 1
,
1 1
n n m m n
m i
i m m mi m j m i j m
m n i m n n
i k m i i k k
m m m
S
Cov X r i j r j i
L k m
r k r k
kα
σ σ σ
σ σ σ
= + = + = = = + − = = + − = = = = − = − =
∑
∑ ∑
∑ ∑
∑ ∑
∑∑
∑
(
)
(
)
( )
( ) ( )
(
)
( )
( ) 2 3 1 2 1 1 1 22 2 1 1 1 1 2 2 exp , 2 1 exp 2 1 1 ln for 0 2 n n m i i m m
n n
k m
u S
E Cov X
L k
u m
k
m
L n L m n
n n m m n n α α µ α µ α γ µ σ µ σ α γ = + = + + − + − + − +
< <
∑
∑
.2 2 2 2
4
2 2
1 2 3
,
:
m n m m m n n m m
m n m n n
m n n m m
m n n
S S S S S S S
E Cov E E
S S S S
E E
F F F
σ σ σ σ σ
σ σ σ
+ + + − − + − − − − = + +
SinceL t
( )
tα is monotonic decreasing, so for 1≤ ≤i m, r j
(
− ≤i) (
r j−m)
, by (4.2), Karamata’s theorem and the property of the slowly varying function, we obtain(
)
(
)
( )
( ) ( )
( )
2 2
1
1 2 1 1 2 1
2
1 2 1 1
1 2 2
1
1 1 1 1
2 2
2 2
2 4
1 1
,
1
( ) ( )
1 for 0
4 4
m m n m m n
i j
i j m i j m
m n m n
m m n n
i j m k
m n m n
F Cov X X r j i
m
r j m r k
m m L n
L n n
n L m
L m L n m n
m n m
n m n
α α
α α
α α γ
σ σ σ σ
σ σ σ σ
α γ
+ +
== = + = = +
+
= = + =
− − −
= −
≤ −
< < <
∑
∑
∑ ∑
∑ ∑
∑
.
By
σ
n = Var( )
Sn , then we get2
1
m
m S E
σ
=
, by the stationary of the sequence
(
X X, n n)
∈, we get2 2
d n m n m
S+ −S =S , thus E S
(
n+2m−Sn)
2 =E S( )
2m 2 and by the fact that EX Xi j>0 for all i j, and (4.2), we deduce that2 2
2 2 2 2
2
1 2
1 1
2 2
2 ( )
for some 1
( )
m n m n m m n m m
m n n m n
m
n
S S S S S S S
F E E E E
m L m m m
n L n n n
α α γ
σ σ σ σ σ
σ γ
σ
+ +
− −
− −
≤ − ≤
<
,
where the second inequality follows from Jensen inequality. For F3, we have
2 2
3 0
m n n m m
m n n
S S S S
F E E
σ σ + σ
−
= − − =
, so we prove (2.3) for some 0< <α 1. Next, we prove (2.3) for some α>1.
, 1 2 3 4
, m , , n :
m m m n n
m n
S S
Cov I M u y I M u y G G G G
σ
σ
≤ ≤ ≤ ≤ = + + +
,
where G1 - G4 are defined as E1 - E4 in (4.11), but for (1.2) holds for α>1. Similarly as the proof of
1
E - E4, it is easy to check that G1 G2 G3 m n
γ
+ +
for some
γ
>0.Define G4=:H1+H2+H3 as in (4.12), by Karamata’s theorem, we get
(
)
( )
( )
( )
2 2 1
1 1 1 1 1
1 2 1 1 1
2 2 2 2
1
2 2
1 1 1
1
2 2
1 1
1 1
for 0 2
m m n m m n
i j m i t m
m n
t m
H r j i r t
m n m n
m m m
r t L m
n m n
m n
γ
α γ
+ + −
= = + = = +
+
− = +
≤ −
< <
∑ ∑
∑ ∑
∑
.
By the definition of σn and Karamata’s theorem, it is easy to obtain that
(
) ( )
( )
( )
( )
( )
1 1 1 1
2
1 1 1 1 1
2 2 α 2 2 α 2 α
σ − − − − ∞
= = = = =
≤ n = +
∑
n − = +∑
n −∑
n × +∑
n +∑
+t t t t t
L t L t L t
n n n t r t n n t r t n n n n n cn
then we have
( )
1 2n O n
σ = . Similarly as the proof of F2,
1 2 2
2
m
n
m m
H
n n
γ σ
σ
for 0 1
2
γ
< < ,by the stationary of the sequence
(
X X, n n)
∈, it is easy to see that 2 23 0
m n n m m
m n n
S S S S
H E E
σ σ + σ
−
= − − =
.
So we have proved (2.3) for the case of α>1. Finally we should prove (2.3) for α=1.
, 1 2 3 4
, m , , n :
m m m n n
m n
S S
Cov I M u y I M u y I I I I
σ
σ
≤ ≤ ≤ ≤ = + + +
,
where we replaced E1 - E4 by I1 - I4 as in (4.11), but for (1.2) holds for some α =1.
Similarly, it is easy to check that I1 I2 I3 m n
γ
+ +
for some
γ
>0. Define I4=:J1+J2+J3. Since( )
r t is monotonic decreasing, define
( )
( )
1
1
ˆ 1 2n
t
L n r t
−
=
= +
∑
, by the application of the proposition (Potter’s TH)of the slowly varying function, for any 0 1 1 2
δ
< < , we have
( )
( )
(
)
( )
(
)
(
( )
)
(
)
(
)
( )
(
)
(
)
(
)
( )
(
)
1 2
1 1
2 1 1
1 1 1 1 1
1 1 2 2
2 2
1 1
1 1
2 2
2 2
1 1
2 2
1 1
2 2
ˆ
ˆ ˆ
ˆ ˆ
ˆ ˆ
2 1
for some 0
m n m n
t m t
m n m n
mL n m
m m
J r t r t
L m L n m n
L n m L n m
m m n m n m
n n m n
L m L n
m n m m n m
n m n m n
δ δ
δ δ γ
σ σ σ σ
γ
+ + + +
= + =
+
+ + + +
=
+
< <
∑
∑
2 .
( )
( )
1
1
1 1
2
2 2
2
2 1
2 ˆ ˆ
ˆ ˆ
m
n
L m
m m m m
J
n n n n
L n
δ γ
σ σ
−
for some 0 1
2
γ
< < .Similarly as F3 =0 when 0< <α 1, we have J3 =0, So we have I4 m n γ
. Finally, we get that
1 2 3 4
m
I I I I
n γ
+ + +
for some
γ
>0, when α=1.References
[1] Dudzinski, M. (2008) The Almost Sure Central Limit Theorems in The Joint Version for The Maxima and Sums of Certain Stationary Gaussian Sequence. Statistics & Probability Letters, 78, 347-357.
http://dx.doi.org/10.1016/j.spl.2007.07.007
[2] Brosamler, G.A. (1988) An Almost Everywhere Central Limit Theorem. Mathematical Proceedings of the Cambridge Philosophical Society, 104, 561-574. http://dx.doi.org/10.1017/S0305004100065750
[3] Schatte, P. (1988) On Strong Versions of the Central Limit Theorem. Mathematische Nachrichten, 137, 249-256.
http://dx.doi.org/10.1002/mana.19881370117
[4] Fahrner, I. and Stadtmuller, U. (1998) On Almost Sure Central Max-Limit Theorems. Statistics & Probability Letters,
[5] Cheng, H., Peng, L. and Qi, Y.C. (1998) Almost Sure Convergence in Extreme Value Theory. Mathematische Nach-richten, 190, 43-50. http://dx.doi.org/10.1002/mana.19981900104
[6] Chandrasekharan, K. and Minakshisundaram, S. (1952) Typical Means. Oxford University Press, Oxford.
[7] Leadbetter, M.R., Lindgren, G. and Rootzen, H. (1983) Extremes and Related Properties of Random Sequences and Processes. Springer, New York. http://dx.doi.org/10.1007/978-1-4612-5449-2
[8] Csaki, E. and Gonchigdanzan, K. (2002) Almost Sure Limit Theorems for The Maximum of Stationary Gaussian Se-quences. Statistics & Probability Letters, 58, 195-203. http://dx.doi.org/10.1016/S0167-7152(02)00128-1