Next Article in Journal
Information Theoretical Analysis of Quantum Mixedness in a Finite Model of Interacting Fermions
Next Article in Special Issue
Self-Normalized Moderate Deviations for Degenerate U-Statistics
Previous Article in Journal
Greedy Algorithm for Deriving Decision Rules from Decision Tree Ensembles
Previous Article in Special Issue
An Erdős-Révész Type Law for the Length of the Longest Match of Two Coin-Tossing Sequences
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

On the Komlós–Révész SLLN for Ψ-Mixing Sequences

by
Zbigniew S. Szewczak
Faculty of Mathematics and Computer Science, Nicolaus Copernicus University, ul. Chopina 12/18, 87-100 Toruń, Poland
Entropy 2025, 27(1), 36; https://doi.org/10.3390/e27010036
Submission received: 15 November 2024 / Revised: 28 December 2024 / Accepted: 30 December 2024 / Published: 4 January 2025
(This article belongs to the Special Issue The Random Walk Path of Pál Révész in Probability)

Abstract

:
The Komlós–Révész strong law of large numbers (SLLN) is proved for ψ -mixing sequences without a rate assumption.

1. Introduction and Result

Let { X k } k N ,   N = { 1 , 2 , 3 , } , be a sequence of random variables defined on a probability space ( Ω , F , P ) and X p p = E ( | X | p ) . It was proved by Komlós and Révész in [1] that for centered, independent random variables X k ,
k = 1 n X k 2 2 X k k = 1 n X k 2 2 n 0 , a . s .
provided that k = 1 X k 2 2 diverges (see also [2], p. 75 and Ex. 13 on p. 137 in [3]). If E ( X k ) = m ,   k 1 , then we have strong consistency:
k = 1 n X k m 2 2 X k k = 1 n X k m 2 2 n m , a . s .
and it was proved that this estimator has minimal variance. Nevertheless, as observed in the preface of [4], “For many phenomena in the real world, the observations are not independent…”, so the question on (1) is intriguing when there is a lack of independence. In [5], the Komlós–Révész theorem is investigated for pairwise independence by the method of subsequences. For martingale difference sequences in L p , p ( 1 , 2 ] , by the Doob theorem (see Th. 2 on p. 246 in [3]), it is obtained in [6]. In [7], the case p ( 0 , 1 ) and p > 2 is also discussed for negatively dependent and mixing sequences. In the latter, some rate on mixing is assumed, so this paper aims to remove this restriction.
Let σ -fields A , B satisfy A , B F . Recall the following strong measure of dependence:
ψ ( A , B ) = sup | P ( B A ) P ( B ) · P ( A ) 1 | ; P ( B ) · P ( A ) > 0 , A A , B B .
It is well known that (see p. 124 in Vol. I, [4])
ψ m = sup J 1 ψ ( F 1 J , F J + m ) = sup J 1 sup E ( g | F 1 J ) E ( g ) g 1 ,
where F k m denotes σ –field generated by X k , X k + 1 , , X m , m Z and the inner sup is taken over g L real 1 ( F J + m ) . We say that { X k } is ψ -mixing if lim m ψ m = 0 . It is worth noticing that Kesten and O’Brien and Bradley gave examples of ψ –mixing sequences with an arbitrary rate of mixing (see Ch. 3 and Ch. 26 in [4]).
The following is the main result in this paper.
Theorem 1. 
Suppose { X k } is ψ–mixing. Let { a k } be a sequence satisfying 0 < a k p 1 E | X k | p C < for k N .
(i) 
If 0 < p < 1 and k = 1 a k converges, then
k = 1 n a k X k k = n a k n 0 , a . s .
(ii) 
If 1 < p 2 and E ( X k ) = 0 ,   sup n ( k = 1 n a k ) 1 k = 1 n a k E | X k | < and k = 1 a k diverges, then
k = 1 n a k X k k = 1 n a k n 0 , a . s .
(iii) 
If p > 2 and E ( X k ) = 0 ,   sup n ( k = 1 n a k ) 1 k = 1 n a k E | X k | < and k = 1 a k diverges and 0 < a k E | X k | 2 C < for k N , then (3) holds.
Remark 1. 
(a) It follows from the proof that in case ( i ) , ψ–mixing is not required.
(b) In fact, sup n ( k = 1 n a k ) 1 k = 1 n a k E | X k | < if sup k E | X k | < . The latter is also used in the case of pairwise independent random variables (see Theorem 2 in [5]) and can be omitted if ψ m = 0 for some m 1 .
(c) For p = 1 , Theorem 1 does not hold. Suppose that { X k } is a stochastic sequence P ( X k = 1 ) = 1 P ( X k = 0 ) = 1 / 2 constructed in the proof of Theorem 1 in [8]. Thus, if { g n } is a non-increasing sequence, g n 0 , then ψ n g n . Now, let { ξ k } be a sequence of independent random variables with P ( ξ k = k ) = 1 2 k , P ( ξ k = k ) = 1 2 k , P ( ξ k = 0 ) = 1 1 k and independent of X k . Set Z k = ξ k cos ( X k π ) , k 1 so that E ( e i t Z k ) = E ( e i t ξ k ) . Thus, E Z k = 0 , E | Z k | = 1 . By Proposition 3.16 on p. 82 in Vol. I, [4] and by Theorem 6.2 on p. 193 in Vol. I, [4] we have that { Z k } is ψ-mixing (non-stationary) with rate g n . Let a k 1 . Now, (3) fails by the second Borel–Cantelli lemma (see [9] and p. 210 in [10]) since k 1 P ( | Z k | k ) = k 1 1 k = .
The following variant of the converse statement does not require strong mixing.
Proposition 1. 
Let { a k } be a sequence of positive reals. Suppose { X k } is an arbitrary dependent random sequence such that sup k E | X k | = C < and lim inf n E | k = 1 n a k X k | = c > 0 . If (3) holds, then k 1 a k diverges.
Set q = | p p 1 | . Theorem 1 applied with a k 1 = ( E | X k | p ) 1 p 1 and, in case ( i i i ) with a k 1 = E | X k | 2 ( E | X k | p ) 1 p 1 , yields
Corollary 1. 
Suppose { X k } is ψ-mixing.
(i) 
If 0 < p < 1 and k = 1 X k p q converges, then
k = 1 n X k p q X k n = k X k p q n 0 , a . s .
(ii) 
If 1 < p 2 ,   sup n ( k = 1 n X k p q ) 1 k = 1 n X k p q E | X k | < ,   E ( X k ) = 0 and k = 1 X k p q diverges, then
k = 1 n X k p q X k k = 1 n X k p q n 0 , a . s .
(iii) 
If p 2 ,
sup n ( k = 1 n X k 2 2 X k p q ) 1 k = 1 n ( X k 2 2 X k p q ) 1 E | X k | < ,
E ( X k ) = 0 and k = 1 ( X k 2 2 X k p q ) 1 diverges, then
k = 1 n ( X k 2 2 X k p q ) 1 X k k = 1 n ( X k 2 2 X k p q ) 1 n 0 , a . s .
We apply our result to the ψ -mixing Cramér model. Namely, we have sequence { η k } with P ( η k = 1 ) = 1 P ( η k = 0 ) = 1 / ln k ,   k 3 , which is ψ -mixing. Let ζ k = η k ln k . It is easy to see that σ k 2 = Var ( ζ k ) ln k ,   m k = E ( ζ k ) = 1 and k = 3 n 1 σ k 2 n ln n . Thus, for { ζ k } in the case p = 2 , we obtain by (5) that SLLN for Cramér model independence can be replaced by ψ -mixing (see e.g., Lemma A.1 in [11]).
Lemma 1. 
lim n ln n n k = 3 n η k = 1 , a . s .
The paper is organized as follows. In the next section, there are some auxiliary results. These results are required in the proof of Theorem 1 in the last section.

2. Auxiliary Results

The following result for p > 1 is in [12] and for p = 1 see [9] (see also Theorem 2.20 on p. 40 in [13]).
Theorem 2. 
Suppose { X k } is ψ-mixing and E ( X k ) = 0 ,   k Z . Let { b n } ,   b 0 > 0 be an increasing to infinity sequence of real numbers such that for some p 1 ,
(i) 
k = 1 b k 2 p E ( | X k | 2 p ) < ;
(ii) 
k = 1 b k 2 ( b k 2 b k 1 2 ) 1 p ( E ( X k 2 ) ) p < ;
(iii) 
sup n b n 1 k = 1 n E | X k | < .
Then, b n 1 S n 0 ,   n , almost surely (a.s.).
The next result improves Lemma 2.1 slightly in [14].
Proposition 2. 
Suppose { X k } is ψ-mixing and E ( X k ) = 0 ,   k Z . Let b n R ,   b 0 > 0 ,   b n , and sup n b n 1 k = 1 n E | X k | < ,   p ( 1 , 2 ] ,   k = 1 E | X k | p b k p < , then b n 1 S n 0 ,   n , a.s.
Proof of Proposition 2. 
Set X ¯ k = X k I ( | X k | b k ) E ( X k I ( | X k | b k ) ) . In view of Theorem 2, we can assume 1 < p < 2 .
k 1 P ( X k X k I ( | X k | b k ) ) C k 1 E | X k | p b k p < .
Further,
b n 1 | k = 1 n E ( X k I ( | X k | b k ) ) | b n 1 k = 1 n E ( | X k | I ( | X k | > b k ) )       b n 1 k = 1 n b k p + 1 E ( | X k | p I ( | X k | > b k ) ) b n 1 k = 1 n b k p + 1 E | X k | p 0 .
So that
k 1 E X ¯ k 2 b k 2 k 1 E ( X k 2 I ( | X k | b k ) ) b k 2 k 1 E | X k | p b k p < .
Now, apply Theorem 2 to { X ¯ k } with p = 1 . □
By Lemma 3.6″ on p. 284 in [15] and Kronecker’s lemma (see e.g., [16], p. 236) we have the following SLLN for arbitrary dependent random variables (see also [17]).
Lemma 2. 
Suppose { X k } is a sequence of random variables, p ( 0 , 1 ] ,   b n and k = 1 E | X k | p b k p < . Then, b n 1 S n 0 ,   n , a.s.

3. Proofs

Proof of Theorem 1. 
The key role in the proof of Theorem 1 is played by the Dini theorem (see e.g., [18], Theorem 4 on p. 127) and the Abel–Dini theorem (see e.g., [18], Theorem 1 on p. 125) (for a generalization, see Lemma 11 in [7]).
For p ( 0 , 1 ) by the Dini theorem,
n = 1 a n p E | X n | p ( k = n a k ) p C n = 1 a n ( k = n a k ) p <
since k = 1 a k < . Thus, (2) holds by Lemma 2.
Now, assume p ( 1 , 2 ] . By the Abel–Dini theorem,
n = 1 a n p E | X n | p ( k = 1 n a k ) p C n = 1 a n ( k = 1 n a k ) p <
since k = 1 a k = . Set b n = k = 1 n a k . Since sup n b n 1 k = 1 n a k E | X k | < ,
k = 1 n a k X k k = 1 n a k n 0 , a . s .
holds by Proposition 2.
In the case that p > 2 , set b 0 = 0.5 b 1 . By Theorem 2 and the previous point, it is enough to prove that
S = n = 1 ( b n 2 b n 1 2 ) 1 p 2 b n 2 a n p ( E ( X n 2 ) ) p 2 b n p < .
Let δ = 2 ϵ p , where ϵ ( 0 , 1 ) is fixed. In view of b n , we have that there exist C > 0 such that a n E ( X n 2 ) C b n 3 δ , for each n. Therefore,
a n p 1 E ( X n 2 ) p 2 C b n 3 p 2 ϵ a n p 2 1 .
But ( b n 2 b n 1 2 ) 1 p 2 = ( a n b n ( 2 a n b n ) ) 1 p 2 so that by the Abel–Dini theorem,
S C n = 1 a n ( a n b n ) 1 p 2 b n p + 2 b n 3 p 2 ϵ a n p 2 1 = C n = 1 a n b n 1 + ϵ < .
Since sup n b n 1 k = 1 n a k E | X k | < , Theorem 1 is proved. □
Proof of Proposition 1. 
Suppose k 1 a k < and (3) holds. Choose N 1 such that k n a k < c 3 C for n N 1 . Next, choose N 2 N 1 such that E | k = 1 n a k X k | > c / 2 for n N 2 . Now, for a fixed N N 2 , we have
k = N + 1 n a k X k n k = 1 N a k X k , a . s .
On the other hand, by Fatou’s lemma (cf. Theorem 5.1 on p. 218, [19]),
c / 2 < E | k = 1 N a k X k | lim inf n E | k = N + 1 n a k X k | < c / 3 .
Thus, c < 0 . Contradiction. □

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Data Availability Statement

Data is contained within the article.

Conflicts of Interest

The author declares no conflict of interest.

References

  1. Komlós, J.; Révész, P. On the weighted averages of independent random variables. Magyar Tud. Akad. Mat. Kutató Int. Közl. 1965, 9, 583–587. [Google Scholar]
  2. Révész, P. The Laws of Large Numbers; Academic Press: New York, NY, USA, 1968. [Google Scholar]
  3. Chow, Y.S.; Teicher, H. Probability Theory. Independence, Interchangebility, Martingales, 3rd ed.; Springer Texts in Statistics: New York, NY, USA, 2003. [Google Scholar]
  4. Bradley, R.C. Introduction to Strong Mixing Conditions; Kendrick Press: Heber City, UT, USA, 2007; Volume I–III. [Google Scholar]
  5. Rosalsky, A. A strong law for weighted averages of random variables and the Komlós–Révész estimation problem. Calcutta Statist. Assoc. Bull. 1986, 35, 59–66. [Google Scholar] [CrossRef]
  6. Cohen, G. On the Komlós–Révész estimation problem for random variables without variances. Acta Sci. Math. 2008, 74, 915–925. [Google Scholar]
  7. Szewczak, Z.S. On the Komlós–Révész SLLN for dependent variables. Acta Math. Hungar. 2018, 156, 47–55. [Google Scholar] [CrossRef]
  8. Kesten, H.; O’Brien, G.L. Examples of mixing sequences. Duke Math. J. 1976, 43, 405–415. [Google Scholar] [CrossRef]
  9. Blum, J.R.; Hanson, D.L.; Koopmans, L.H. On the Strong Law of Large Numbers for a Class of Stochastic Processes. Z. Wahr. Verw. Gebiete 1963, 2, 1–11. [Google Scholar] [CrossRef]
  10. Petrov, V.V. Limit Theorems of Probability Theory. Sequences of Independent Random Variables; Oxford Studies in Probability 4; Clarendon Pr: Oxford, UK, 1995. [Google Scholar]
  11. Weber, M.J.G. Critical probabilistic characteristics of the Cramér model for primes and arithmetical properties. Indian J. Pure Appl. Math. 2024. [Google Scholar] [CrossRef]
  12. Szewczak, Z.S. On the Wittmann strong law for mixing sequences. Period. Math. Hungar. 2024, 89, 257–264. [Google Scholar] [CrossRef]
  13. Hall, P.; Heyde, C.C. Martingale Limit Theory and Its Applications; Academic Press: New York, NY, USA, 1980. [Google Scholar]
  14. Hu, D.; Chen, P.; Sung, S.H. Strong laws of weighted sums of ψ–mixing random variables and applications in errors-in-variables regression models. TEST 2017, 26, 600–617. [Google Scholar] [CrossRef]
  15. Loève, M. On almost sure convergence. In Proceedings of the 2nd Berkeley Symposium on Mathematical Statistics and Probability, Berkeley, CA, USA, 31 July–12 August 1950; University of California Press: Berkeley, CA, USA, 1951; pp. 279–303. [Google Scholar]
  16. Ash, R.B. Probability and Measure Theory; Academic Press: New York, NY, USA, 2000. [Google Scholar]
  17. Petrov, V.V. On the order of growth and sums of dependent variables. Theory Probab. Appl. 1974, 18, 348–350. [Google Scholar] [CrossRef]
  18. Knopp, K. Infinite Sequences and Series; Dover Publication Inc.: New York, NY, USA, 1956. [Google Scholar]
  19. Gut, A. Probability: A Graduate Course, 2nd ed.; Springer: New York, NY, USA, 2013. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Szewczak, Z.S. On the Komlós–Révész SLLN for Ψ-Mixing Sequences. Entropy 2025, 27, 36. https://doi.org/10.3390/e27010036

AMA Style

Szewczak ZS. On the Komlós–Révész SLLN for Ψ-Mixing Sequences. Entropy. 2025; 27(1):36. https://doi.org/10.3390/e27010036

Chicago/Turabian Style

Szewczak, Zbigniew S. 2025. "On the Komlós–Révész SLLN for Ψ-Mixing Sequences" Entropy 27, no. 1: 36. https://doi.org/10.3390/e27010036

APA Style

Szewczak, Z. S. (2025). On the Komlós–Révész SLLN for Ψ-Mixing Sequences. Entropy, 27(1), 36. https://doi.org/10.3390/e27010036

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop