Almost sure limit theorems for U-statistics
Introduction
U-statistics generalize the concept of the sample mean of independent identically distributed (i.i.d.) random variables. The statistical interest in U-statistics stems from the fact that they form a class of unbiased estimators of a certain parameter with minimal variances. We begin by introducing some notation and recalling the concept of U-statistics.
Let X1,X2,… be i.i.d random variables with common distribution function F(x). Let m⩾1 and let be a measurable function symmetric in its arguments. The U-statistic with kernel h is defined byThe kernel h is called degenerate with respect to F(x) if for all 1⩽j⩽m,Letand for i=0,…,m letNote that and . Furthermore, the hi are degenerate for i=1,…,m (see Denker, 1985). If ζ1>0 the U-statistic Un(h) is called non-degenerate and degenerate otherwise. The smallest integer c for which ζc>0 is called the critical parameter of the U-statistic Un(h). Without loss of generality we will assume θ=0 throughout the rest of this article.
The theory of U-statistics started to develop intensively after Hoeffding's (1948) fundamental article. He showed asymptotic normality of non-degenerate U-statistics under the assumptionusing the following representation of U-statistics:where mUn(h1) is a sum of i.i.d. random variables and Un(h2),…,Un(hm) are U-statistics with degenerate kernels.
In the degenerate case nc/2Un(h) weakly converges to a multiple Wiener integral whenever h is square integrable (see e.g. Denker, 1985). Here c is the critical parameter of Un(h).
In the late 1980s, Brosamler (1988) and Schatte (1988) independently proved a new type of limit theorem. This type of statement extends the classical central limit theorem in the i.i.d. case to a pathwise version and is therefore called an almost sure central limit theorem (ASCLT). In the 1990s, many studies were done to prove almost sure limit theorems (ASLT) in different situations, for example in the case of independent but not necessarily identically distributed random variables (see Berkes and Dehling, 1993). Excellent surveys on this topic may be found in Atlagh and Weber (2000) as well as in Berkes (1998). Recently Berkes and Csaki (2001) obtained a general result in almost sure limit theory. They used it to prove almost sure versions of several classical limit theorems. In particular they stated the following theorem for U-statistics. Theorem 1.1 Let c be the critical parameter of the U-statistic Un(h). Under assumption (1),where G is the limit distribution of nc/2Un(h) and CG denotes the set of continuity points of G.
In the present note we relax the moment condition in Theorem 1.1 and extend the statement in two directions. First we will obtain an ASLT with stable limiting distribution for a non-degenerate U-statistic. Furthermore we extend Theorem 1.1 to a functional version.
Section snippets
Preliminaries
Let (Yn)n⩾1 be a sequence of random elements taking values in a Polish space and let G be a probability measure on the Borel σ-field in . We say that (Yn)n⩾1 satisfies the ASLT with limiting distribution G if with probability 1,Here δYk is the Dirac measure at Yk and “⇒” denotes weak convergence of measures. Throughout this note the following lemma will be of fundamental interest. Lemma 2.1 Let (Yn)n⩾1 be a sequence of -valued random elements which satisfies the ASLT with
Relaxing the moment assumption
In this section, we will relax the moment assumption of Theorem 1.1. For the weak convergence of nc/2Un(h) (where c denotes the critical parameter of Un(h)) Koroljuk and Borovskich (1994) weakened assumption (1) toWe are going to prove the validity of Theorem 1.1 under these assumptions: Theorem 3.1 Let c be the critical parameter of the U-statistic Un(h). If (4) is satisfied then the statement of Theorem 1.1 is true. Proof First of all note that, if c is the critical parameter of
Convergence to stable distributions
As we shall see in this section, under some mild moment conditions weak convergence of a sequence of non-degenerate U-statistics to a stable limit distribution implies the validity of the corresponding ASLT. Let Gα denote a stable law with characteristic exponent α. Theorem 4.1 Let for some α∈(1,2],where L(n) is a slowly varying function for which . Ifthen Remarks (1) Assumption (6)
The non-degenerate case
In the non-degenerate case the functional version of Theorem 1.1 can be deduced directly from Theorem 2 of Lacey and Philipp (1990), which deals with sums of i.i.d. random variables. Throughout this section we assume that (1) holds. Let D[0,1] denote the space of cadlag functions on [0,1] and let W denote the Wiener measure on D[0,1] (see Billingsley, 1999). We introduce the following D[0,1]-valued random functionswhere ⌊·⌋ denotes the integer
Acknowledgements
We are grateful to H. Dehling for drawing our attention to a helpful publication by E. Giné and J. Zinn.
References (16)
Results and problems related to the pointwise central limit theorem
- et al.
A universal result in almost sure limit theory
Stochastic Process. Appl.
(2001) - et al.
A note on the almost sure central limit theorem
Statist. Probab. Lett.
(1990) Functional limit theorems for U-statistics in the degenerate case
J. Multivariate Anal.
(1977)- et al.
Le théorème central limite presque sûr
Exposition. Math.
(2000) - et al.
Some limit theorems in log density
Ann. Probab.
(1993) Convergence of Probability Measures
(1999)An almost everywhere central limit theorem
Math. Proc. Cambridge Philos. Soc.
(1988)
Cited by (10)
On the almost sure convergence for the joint version of maxima and minima of stationary sequences
2019, Statistics and Probability LettersCitation Excerpt :The first result on Almost Sure Central Limit Theorem (ASCLT) presented independently by Brosamler (1988), Schatte (1988) and Lacey and Philipp (1990) extended the classical central limit theorem to an almost sure version. Following the above discovery, during the past two decades and half, there were many developments on the ASCLTs for some other functions of random variables, namely for U-statistics, local times and extremal statistics (Berkes and Csáki, 2001; Fahrner and Stadmüller, 1998; Holzmann et al., 2004; Peng et al., 2009). The exact analogue of the ASCLT (1.1) for extremal statistics was given in Cheng et al. (1998).
Change-point analysis using logarithmic quantile estimation
2019, Statistics and Probability LettersESTIMATION BY STABLE MOTIONS AND ITS APPLICATIONS
2023, Probability and Mathematical StatisticsA Monte Carlo algorithm for multiple stochastic integrals of stable processes
2022, Stochastics and DynamicsOn the almost sure central limit theorem for ARX processes in adaptive tracking
2019, International Journal of Adaptive Control and Signal Processing
- 1
Acknowledges the financial support of the graduate school Gruppen und Geometrie.
- 2
Acknowledges the financial support of the graduate school Strömungsinstabilitäten und Turbulenz.