This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
 Author's personal copy

Mathematical and Computer Modelling 53 (2011) 189–195

Contents lists available at ScienceDirect

Mathematical and Computer Modelling

journal homepage: www.elsevier.com/locate/mcm

Statistical A-summability of positive linear operators

Kamil Demirci, Sevda Karakuş ∗
Sinop University, Faculty of Sciences and Arts, Department of Mathematics, TR-57000, Sinop, Turkey

article info abstract

Article history: In this paper, using the concept of statistical A-summability which is stronger than the
Received 21 April 2010 A-statistical convergence we provide a Korovkin-type approximation theorem. We also
Received in revised form 4 August 2010 compute the rates of statistical A-summability of a sequence of positive linear operators.
Accepted 4 August 2010
© 2010 Elsevier Ltd. All rights reserved.
Keywords:
Statistical convergence
Statistical A-summability
Positive linear operator
Korovkin-type approximation theorem

1. Introduction

The idea of statistical convergence was introduced by Fast [1], which is closely related to the concept of natural density
or asymptotic density of subsets of the set of natural numbers N. Let K be a subset of N. The natural density of K is the
nonnegative real number given by δ (K ) := limn→∞ 1n |{k ≤ n : k ∈ K }| provided the limit exists, where |B| denotes the
cardinality of the set B (see [2] for details). Then, a sequence x = {xk } is said to be statistically convergent to a number L if
for every ε > 0,
δ ({k ∈ N : |xk − L| ≥ ε}) = 0.
This is denoted by st − limk→∞ xk = L (see [1,3]). It is easy to see that every convergent sequence is statistically convergent
but not conversely.
If x = {xk } is a number sequence and A = {ajk } is an infinite matrix, then Ax is the sequence whose jth term is given by
∞
−
Aj (x) := ajk xk
k=1

provided the series converges for each j ∈ N. Thus we say that x is A-summable to L if
lim Aj (x) = L.
j→∞

We say that A is regular if limj→∞ Aj (x) = L whenever limk→∞ xk = L. The well-known necessary and sufficient
conditions [4] (Silverman–Toeplitz) for A to be regular are
(R1) ‖A‖ = supj k=1 ajk  < ∞,
∑∞  
(R2) limj→∞ a jk = 0 for each k ∈ N,
(R3) limj→∞ k=1 ajk = 1.
∑ ∞

∗ Corresponding author.
E-mail addresses: [email protected] (K. Demirci), [email protected] (S. Karakuş).

0895-7177/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mcm.2010.08.003
 Author's personal copy

190 K. Demirci, S. Karakuş / Mathematical and Computer Modelling 53 (2011) 189–195

Freedman and Sember [5] introduced the following extension of statistical convergence. Let A = {ajk } be a nonnegative
regular matrix. The A-density of K is defined by
∞
−
δA (K ) := lim ajk χK (k)
j→∞
k=1

provided the limit exists, where χK is the characteristic function of K . Then the sequence x = {xk } is said to be A-statistically
convergent to the number L if, for every ε > 0,
δA ({k ∈ N : |xk − L| ≥ ε}) = 0
or equivalently
−
lim ajk = 0.
j→∞
k:|xk −L|≥ε

We denote this limit by stA − limk→∞ xk = L (see [5–7]). The case in which A = C1 , the Cesàro matrix of order one, reduces
to statistical convergence, and also if A = I, the identity matrix, then it coincides with the ordinary convergence.
Recently, the idea of statistical (C , 1)-summability was introduced in [8], statistical (H , 1)-summability in [9] by Moricz,
and statistical (N , p)-summability by Moricz and Orhan [10]. Then these statistical summability methods were generalized
by defining the statistical A-summability in [11].
Now we recall this convergence method.

Definition 1.1. Let A = {ajk } be a nonnegative regular matrix and x = {xk } be a sequence. We say that x is statistically
A-summable to L if for every ε > 0,
δ j ∈ N : Aj (x) − L ≥ ε = 0,
   

i.e.,
1 
 j ≤ n : Aj (x) − L ≥ ε  = 0.
  
lim
n→∞ n
Thus x = {xk } is statistically A-summable to L if and only if Ax is statistically convergent to L. In this case we write
(A)st − limk→∞ xk = L or, st − limj→∞ Aj (x) = L.
Using Definition 1.1, we see that, if a sequence is bounded and A-statistically convergent to L then it is A-summable to L
and hence statistically A-summable to L. However, its converse is not always true. Such examples were given in [11].
We note that the use of statistical convergence in approximation theory has been shown to be effective than the classical
methods (see, for instance [12–16]). On the other hand, most of the classical approximation operators tend to converge
to the value of the function being approximated. However, at points of discontinuity, they often converge to the average
of the left and right limits of the function. In fact, there are some sharp exceptions such as the interpolation operator of
Hermite–Fejér (see [17]). These operators do not converge at points of simple discontinuity. For such a misbehavior, the
matrix summability methods of Cesàro type are strong enough to correct the lack of convergence (see [18]).
In the present paper, we mainly use the statistical A-summability method in the approximation theory settings. We also
compute the rates of statistical A-summability of a sequence of positive linear operators.

2. A Korovkin-type approximation theorem

We denote by CM [a, b] the space of all continuous functions on [a, b] and bounded on the entire line, that is
|f (x)| ≤ Kf , −∞ < x < ∞,
where Kf is a constant depending on f . Also by B [a, b] we denote the space of all bounded functions on [a, b]. This space is
equipped with the supremum norm
‖f ‖B[a,b] = sup |f (x)| , (f ∈ B [a, b]) .
x∈[a,b]

Let L be a linear operator from CM [a, b] into B [a, b]. Then, as usual, we say that L is a positive linear operator provided
that f ≥ 0 implies L (f ) ≥ 0. Also, we denote the value of L (f ) at a point x ∈ [a, b] by L (f (u) ; x) or, briefly, L (f ; x).
Throughout the paper, we also use the following test functions
e0 (x) = 1, e1 (x) = x, e 2 ( x) = x2 .
First we recall the classical and statistical cases of the Korovkin-type results introduced in [19,20], respectively.

Theorem 2.1. Let {Lk } be a sequence of positive linear operators acting from CM [a, b] into B [a, b]. Then, for all f ∈ CM [a, b],
lim ‖Lk (f ) − f ‖B[a,b] = 0
k→∞
 Author's personal copy

K. Demirci, S. Karakuş / Mathematical and Computer Modelling 53 (2011) 189–195 191

if and only if
lim ‖Lk (ei ) − ei ‖B[a,b] = 0, (i = 0, 1, 2) .
k→∞

Theorem 2.2. Let {Lk } be a sequence of positive linear operators acting from CM [a, b] into B [a, b]. Then, for all f ∈ CM [a, b],
st − lim ‖Lk (f ) − f ‖B[a,b] = 0
k→∞

if and only if
st − lim ‖Lk (ei ) − ei ‖B[a,b] = 0, (i = 0, 1, 2) .
k→∞

Now we have the following main result.

Theorem 2.3. Let A = {ajk } be a nonnegative regular matrix and let {Lk } be a sequence of positive linear operators acting from
CM [a, b] into B [a, b]. Then, for all f ∈ CM [a, b],
 
− ∞ 
st − lim  ajk Lk (f ) − f  =0 (2.1)
 
j→∞  
k =1 B[a,b]

if and only if
 
− ∞ 
st − lim  ajk Lk (ei ) − (ei ) = 0, (i = 0, 1, 2) . (2.2)
 
j→∞  
k =1 B[a,b]

Proof. Since each ei (x) ∈ CM [a, b], (i = 0, 1, 2), the implication (2.1) H⇒ (2.2) is obvious. Now, to prove the implication
(2.2) H⇒ (2.1) assume that (2.2) holds. Let f ∈ CM [a, b] and x ∈ [a, b] be fixed. Since f is bounded on the entire line, we can
write
|f (x)| ≤ H , −∞ < x < ∞. (2.3)
Also, since f is continuous on [a, b], we write that for every ε > 0, there exists a number δ > 0 such that |f (y) − f (x)| < ε
holds for all y ∈ [a, b] satisfying |y − x| < δ . Hence, we get

2H
|f (y) − f (x)| ≤ ε + (y − x)2 (2.4)
δ2
for all y, x. Using the linearity and the positivity of the operators {Lk } and (2.4), we get, for any j ∈ N that,
   
− ∞  − ∞ − ∞ 
ajk Lk (f ; x) − f (x) ≤ ajk Lk (|f (y) − f (x)| ; x) + |f (x)|  ajk Lk (e0 ; x) − e0 (x)
   

 k=1  k=1
 k =1

 
∞   ∞
2H
− − 
≤ ajk Lk ε + 2 (y − x)2 ; x + |f (x)|  ajk Lk (e0 ; x) − e0 (x)
 
k=1
δ  k=1 
 
− ∞  2H − ∞
≤ ε + (ε + H )  ajk Lk (e0 ; x) − e0 (x) + 2 ajk Lk (y − x)2 ; x
   
 k=1  δ k=1
 

2Hc 2 −
 ∞ 
≤ ε+ ε+H + 2 ajk Lk (e0 ; x) − e0 (x)

δ

 k=1 
   
4Hc −
∞  2H − ∞ 
+ 2  ajk Lk (e1 ; x) − e1 (x) + 2  ajk Lk (e2 ; x) − e2 (x)
  
δ  k=1
 δ k =1
 
where c := maxx∈[a,b] |x|. Then taking the supremum over x ∈ [a, b], we have
   
− ∞  − ∞ 
ajk Lk (f ) − f  ≤ ε+B  ajk Lk (e0 ) − e0 
   

 k=1   k=1 
B[a,b] B[a,b]
    
− ∞  − ∞ 
+  ajk Lk (e1 ) − e1  + ajk Lk (e2 ) − e2  (2.5)
   
 k=1   k=1 
B[a,b] B[a,b]
 Author's personal copy

192 K. Demirci, S. Karakuş / Mathematical and Computer Modelling 53 (2011) 189–195

 2

where B := max ε + H + 2Hc
δ2
, 4Hc
δ2
, 2H
δ2
. Now, for a given r > 0, choose ε > 0 such that ε < r. Then, it follows from (2.5)
that
 ∞    ∞  
   
( )  j ≤ n :  ajk Lk (e0 ) − e0 
∑  ∑  r −ε 
 j≤n: a L f − f ≥ r ≥
  
jk k
 
3B
    
k=1 B[a,b] k=1 B[a,b]
   
≤
 n  n
∞   ∞  
   
( )  j ≤ n :  ajk Lk (e2 ) − e2 
∑  r −ε 
 ∑  r −ε 
 j≤n: a L e − e ≥ ≥
 
jk k 1 1
 
3B 3B
    
k=1 B[a,b] k=1 B[a,b]
   
+ + .
n n
Then using the hypothesis (2.2), we get
 ∞  
 
( )
∑ 
 j≤n: a L f − f ≥ r
 
jk k
 
 
k=1 B[a,b]
 
lim =0
n→∞ n
for every r > 0. The proof is complete. 

Remark 2.1. Let A be the Cesàro matrix, i.e.,


1
, 1 ≤ k ≤ j,
ajk = j
0, otherwise,

and let

1, if k is odd,

αk = (2.6)
−1, if k is even.

Observe now that, A = {ajk } is a nonnegative regular matrix and for the sequence α := {αk }

st − lim Aj (α) = 0.
j→∞

However, the sequence {αk } does not converge in the usual sense. Then, consider the following Bernstein–Kantorovich
operators:
k   ∫ n/(k+1)
− k
Uk (f ; x) = (k + 1) x (1 − x)
n k−n
f (t ) dt
n =0
n (n+1)/(k+1)

where x ∈ [0, 1], f ∈ CM [0, 1] and k ∈ N. Using these polynomials, we introduce the following positive linear operators on
CM [0, 1]:

Tk (f ; x) = (1 + αk ) Uk (f ; x) , x ∈ [0, 1] , f ∈ CM [0, 1] (2.7)

where α = {αk } is the same as (2.6). Then observe that

Tk (e0 ; x) = (1 + αk ) e0 (x)
 
k 1
Tk (e1 ; x) = (1 + αk ) e1 (x) +
k+1 2 (k + 1)
k (k − 1)
[ ]
2kx 1
Tk (e2 ; x) = (1 + αk ) e2 (x) + + .
(k + 1)2 (k + 1)2 3 (k + 1)2
Since st − limj→∞ Aj (α) = 0, we conclude that
 
− ∞ 
st − lim  ajk Tk (ei ) − (ei ) = 0, on [0, 1] for each i = 0, 1, 2.
 
j→∞  
k=1 B[0,1]

So {Tk } satisfy all hypotheses of Theorem 2.3 and we immediately see that
 
− ∞ 
st − lim  ajk Tk (f ) − (f ) = 0, on [0, 1] for all f ∈ CM [0, 1] .
 
j→∞  
k=1 B[0,1]
 Author's personal copy

K. Demirci, S. Karakuş / Mathematical and Computer Modelling 53 (2011) 189–195 193

However, since {αk } is not convergent and A-statistical convergent to 0, we can say that ‖Tk (ei ) − (ei )‖B[0,1] is not convergent
and A-statistical convergent to zero for each i = 0, 1, 2. So the classical Korovkin theorem (Theorem 2.1) and the statistical
Korovkin theorem (Theorem 2.2) do not work for our operators defined by (2.7).

3. Rate of convergence

In this section, using statistical A-summability we study the rate of convergence of positive linear operators with the help
of the modulus of continuity.

Definition 3.1. Let A = {ajk } be a nonnegative regular matrix. A sequence x = {xk } is statistical A-summable to a number L
at a rate of β ∈ (0, 1) if for every ε > 0,
 j ≤ n : Aj (x) − L ≥ ε 
   
lim = 0.
n→∞ n1−β
In this case, it is denoted by

xk − L = o n−β ((A)st ) .
 

Using this definition, we obtain the following auxiliary result.

 matrix. Let x = {xk } and y = {yk } be bounded sequences. Assume that

Lemma 3.1.Let A = {ajk } be a nonnegative regular
xk − L1 = o n−β1 ((A)st ) and yk − L2 = o n−β2 ((A)st ). Let β := min {β1 , β2 }. Then we have

(i) (xk − L1 ) ∓ (yk − L2 ) = o n−β ((A)st )

 

(ii) λ (xk − L1 ) = o n ((A)st ), for any real number λ.

 −β1


Proof. (i) Assume that xk − L1 = o n−β1 ((A)st ) and yk − L2 = o n−β2 ((A)st ). Then, since β := min {β1 , β2 }, for ε > 0
   
observe that
 j ≤ n :  Aj (x) − L1 ∓ Aj (y) − L2  ≥ ε   j ≤ n : Aj (x) − L1  ≥ ε  +  j ≤ n : Aj (y) − L2  ≥ ε 
             
2 2
≤
n1−β n1−β
 j ≤ n : Aj (x) − L1  ≥ ε   j ≤ n : Aj (y) − L2  ≥ ε 
       
≤ 2
+ 2
. (3.1)
n1−β1 n1−β2
Now by taking the limit as n → ∞ in (3.1) and using the hypothesis, we conclude that
 j ≤ n :  Aj (x) − L1 ∓ Aj (y) − L2  ≥ ε 
     
lim =0
n→∞ n1−β
which completes the proof of (i). Since the proof of (ii) is similar, we omit it. 

Now we recall the concept of modulus of continuity. For f ∈ CM [a, b], the modulus of continuity of f , denoted by ω (f ; δ),
is defined to be
ω (f ; δ) := sup {|f (y) − f (x)| : y, x ∈ [a, b] , |y − x| ≤ δ} (δ > 0) .
It is also well known that, for any λ > 0 and for all f ∈ CM [a, b]
ω (f ; λδ) ≤ (1 + [λ]) ω (f ; δ)
where [λ] is defined to be the greatest integer less than or equal to λ.
Then we have the following result.

Theorem 3.2. Let A = {ajk } be a nonnegative regular matrix and let {Lk } be a sequence of positive linear operators acting from
CM [a, b] into B [a, b]. Assume that the following conditions hold:
(i) ‖Lk (e0 ) − e0 ‖B[a,b] = o n−β1 ((A)st ) on [a, b] ,
 
∑
(ii) ω f ; γj = o n ((A)st ) on [a, b] where γj :=  ∞ ajk Lk (ϕ) with ϕ (y) = (y − x)2 .
   −β2
 
k=1 B[a,b]

Then we have for all f ∈ CM [a, b]

‖Lk (f ) − f ‖B[a,b] = o n−β ((A)st ) on [a, b]

 

where β := min {β1 , β2 }.

 Author's personal copy

194 K. Demirci, S. Karakuş / Mathematical and Computer Modelling 53 (2011) 189–195

Proof. Let f ∈ CM [a, b] and x ∈ [a, b] be fixed. Using (2.3), the properties of ω, and the positivity and monotonicity of Lk ,
we get, for any δ > 0 and j ∈ N, that
   
− ∞  − ∞ − ∞ 
ajk Lk (f ; x) − f (x) ≤ ajk Lk (|f (y) − f (x)| ; x) + |f (x)|  ajk Lk (e0 ; x) − e0 (x)
   

 k=1  k =1  k=1 
 
∞     ∞
− |y − x| − 
≤ ajk Lk ω f ; δ ; x + |f (x)|  ajk Lk (e0 ; x) − e0 (x)
 
k=1
δ  k=1 
 
∞  [ ]  ∞
− |y − x| − 
≤ ajk Lk 1+ ; x ω (f ; δ) + H  ajk Lk (e0 ; x) − e0 (x)
 
k=1
δ  k=1

 
∞
(y − x)2 ∞
−    − 
≤ ajk Lk 1+ ; x ω ( f ; δ) + H a L ( e ; x ) − e ( x)
 
jk k 0 0
δ 2
 
k=1
 k=1

   
∞ ∞ ∞
1 −
− − 
≤ ajk Lk (e0 ; x) + 2 ajk Lk (ϕ; x) ω (f ; δ) + H  ajk Lk (e0 ; x) − e0 (x)
 
k =1
δ k=1  k=1

   
− ∞  − ∞  ω (f ; δ) −∞
≤ ajk Lk (e0 ; x) − e0 (x) ω (f ; δ) + ω (f ; δ) + H  ajk Lk (e0 ; x) − e0 (x) + ajk Lk (ϕ; x) .
   
 k =1   k=1  δ2 k=1

Then, taking the supremum over x ∈ [a, b], we get

     
− ∞  − ∞  − ∞ 
ajk Lk (f ) − f  ≤ ajk Lk (e0 ) − e0  ω (f ; δ) + 2ω (f ; δ) + H  ajk Lk (e0 ) − e0  (3.2)
     

 k=1   k=1   k=1 
B[a,b] B[a,b] B[a,b]
∑
where δ := γj :=  ∞ ajk Lk (ϕ) . Hence, given ε > 0, it follows from (3.2) and Lemma 3.1 that

k =1 B[a,b]
 ∞    ∞  
    ε 
 ajk Lk (f ) − f  ≥ε   ajk Lk (e0 ) − e0 
∑  ∑ 
 j≤n:  j≤n: ≥ 3 
    
k=1 B[a,b] k=1 B[a,b]
   
≤
n1−β n1−β1
 ∞  
 
ε
 ε   j ≤ n :  ajk Lk (e0 ) − e0 
∑ 
≥
  
  ε   3H 
 j ≤ n : ω f ; γj ≥   j ≤ n : ω f ; γj ≥  
    
3 6
k=1 B[a,b]
+ 1−β
+ 1−β
+ 1−β
(3.3)
n 2 n 2 n 1

where β := min {β1 , β2 }. Letting n → ∞ in (3.3), we conclude from (i) and (ii) that
 ∞  
 
( ) ε
 ∑ 
 j≤n: a L f − f ≥
 
jk k
 
 
k=1 B[a,b]
−  
lim =0
n→∞
n1−β

which means
‖Lk (f ) − f ‖B[a,b] = o n−β ((A)st ) on [a, b] .
 

The proof is completed. 

References

[1] H. Fast, Sur la convergence statistique, Colloq. Math. 2 (1951) 241–244.

[2] I. Niven, H.S. Zuckerman, H. Montgomery, An Introduction to the Theory of Numbers, 5th ed., Wiley, New York, 1991.
[3] J.A. Fridy, On statistical convergence, Analysis 5 (1985) 301–313.
[4] J. Boos, Classical and Modern Methods in Summability, Oxford University Press, New York, 2000.
[5] A.R. Freedman, J.J. Sember, Densities and summability, Pacific J. Math. 95 (1981) 293–305.
[6] J.A. Fridy, H.I. Miller, A matrix characterization of statistical convergence, Analysis 11 (1991) 59–66.
[7] E. Kolk, Matrix summability of statistically convergent sequences, Analysis 13 (1993) 77–83.
[8] F. Moricz, Tauberian conditions, under which statistical convergence follows from statistical summability (C , 1), J. Math. Anal. Appl. 275 (2002)
277–287.
[9] F. Moricz, Theorems related to statistical harmonic summability and ordinary convergence of slowly decreasing or oscillating sequences, Analysis 24
(2004) 127–145.
[10] F. Moricz, C. Orhan, Tauberian conditions under which statistical convergence follows from statistical summability by weighted means, Studia Sci.
Math. Hungar. 41 (4) (2004) 391–403.
 Author's personal copy

K. Demirci, S. Karakuş / Mathematical and Computer Modelling 53 (2011) 189–195 195

[11] O.H.H. Edely, M. Mursaleen, On statistical A-summability, Math. Comput. Modelling 49 (2009) 672–680.
[12] O. Duman, M.K. Khan, C. Orhan, A-statistical convergence of approximating operators, Math. Inequal. Appl. 6 (2003) 689–699.
[13] O. Duman, C. Orhan, Statistical approximation by positive linear operators, Studia Math. 161 (2004) 187–197.
[14] O. Duman, E. Erkus, V. Gupta, Statistical rates on the multivariate approximation theory, Math. Comput. Modelling 44 (2006) 763–770.
[15] E. Erkus, O. Duman, H.M. Srivastava, Statistical approximation of certain positive linear operators constructed by means of the Chan–Chyan–Srivastava
polynomials, Appl. Math. Comput. 182 (2006) 213–222.
[16] M.A. Özarslan, O. Duman, O. Doğru, Rates of A-statistical convergence of approximating operators, Calcolo 42 (2005) 93–104.
[17] R. Bojanic, F. Cheng, Estimates for the rate of approximation of functions of bounded variation by Hermite–Fejér polynomials, in: Proceedings of the
Conference of the Canadian Mathematical Society, vol. 3, 1983, pp. 5–17.
[18] R. Bojanic, M.K. Khan, Summability of Hermite–Fejér interpolation for functions of bounded variation, J. Natur. Sci. Math. 32 (1992) 5–10.
[19] P.P. Korovkin, Linear Operators and Approximation Theory, Hindustan Publ. Co., Delhi, 1960.
[20] A.D. Gadjiev, C. Orhan, Some approximation theorems via statistical convergence, Rocky Mountain J. Math. 32 (2002) 129–138.