Correlation cascades,ergodic properties and long memory of infinitely divisible processes

In this paper, we investigate the properties of the recently introduced measure of dependence called correlation cascade. We show that the correlation cascade is a promising tool for studying the dependence structure of infinitely divisible processes. We describe the ergodic properties (ergodicity, weak mixing, mixing) of stationary infinitely divisible processes in the language of the correlation cascade and establish its relationship with the codifference. Using the correlation cascade, we investigate the dependence structure of four fractional α$\alpha$-stable stationary processes. We detect the property of long memory and verify the ergodic properties of the discussed processes.     

Association and random measures

Summary Our point of departure is the result, due to Burton and Waymire, that every infinitely divisible random measure has the property variously known as association, positive correlations, or the FKG property. This leads into a study of stationary, associated random measures onR ^d . We establish simple necessary and sufficient conditions for ergodicity and mixing when second moments are present. We also study the second moment condition that is usually referrent to as finite susceptibility. As one consequence of these results, we can easily rederive some central limit theorems of Burton and Waymire. Using association techniques, we obtain a law of the iterated logarithm for infinitely divisible random measures under simple moment hypotheses. Finally, we apply these results to a class of stationary random measures related to measure-valued Markov branching processes.Research supported in part by NSF Grant DMS-8701212 at the University of Virginia     

On infinitely divisible semimartingales

Stricker’s theorem states that a Gaussian process is a semimartingale in its natural filtration if and only if it is the sum of an independent increment Gaussian process and a Gaussian process of finite variation, see Stricker (Z Wahrsch Verw Geb 64(3):303–312, 1983). We consider extensions of this result to non Gaussian infinitely divisible processes. First we show that the class of infinitely divisible semimartingales is so large that the natural analog of Stricker’s theorem fails to hold. Then, as the main result, we prove that an infinitely divisible semimartingale relative to the filtration generated by a random measure admits a unique decomposition into an independent increment process and an infinitely divisible process of finite variation. Consequently, the natural analog of Stricker’s theorem holds for all strictly representable processes (as defined in this paper). Since Gaussian processes are strictly representable due to Hida’s multiplicity theorem, the classical Stricker’s theorem follows from our result. Another consequence is that the question when an infinitely divisible process is a semimartingale can often be reduced to a path property, when a certain associated infinitely divisible process is of finite variation. This gives the key to characterize the semimartingale property for many processes of interest. Along these lines, using Basse-O’Connor and Rosiński (Stoch Process Appl 123(6):1871–1890, 2013a), we characterize semimartingales within a large class of stationary increment infinitely divisible processes; this class includes many infinitely divisible processes of interest, including linear fractional processes, mixed moving averages, and supOU processes, as particular cases. The proof of the main theorem relies on series representations of jumps of càdlàg infinitely divisible processes given in Basse-O’Connor and Rosiński (Ann Probab 41(6):4317–4341, 2013b) combined with techniques of stochastic analysis.     

Infinitely divisible stochastic processes

Summary It often happens that a stochastic process may be approximated by a sum of a large number of independent components no one of which contributes a significant proportion of the whole. For example the depth of water in a lake with many small rivers flowing into it from distant sources, or the point process of calls entering a telephone exchange (considered as the sum of a number of point processes of calls made by individual subscribers) may approximately fulfil these hypotheses. In the present work we formulate and solve the problem of characterizing stochastic processes all of whose finite-dimensional distributions are infinitely divisible. Although some of our results could be derived from known theorems on probabilities on general algebraic structures, many could not and it seems preferable to take the vector-valued infinitely divisible laws as our starting point. We emphasize that an infinitely divisible process (in our sense) on the real line is not necessarily a decomposable process in the sense of Lévy (cf. § 4) though decomposable processes are particular cases.In § 1 a representation theorem for non-negative (and possibly infinite) stochastic processes is derived, while an analogous theorem in the real-valued case is to be found in § 2. § 3 is devoted to the statement of a central limit theorem and the investigation of some of the properties of infinitely divisible processes. The investigation is continued in § 4 by an examination of processes on the real line giving, for example, a representation theorem under weak conditions for infinitely divisible processes which are a.s. sample continuous. Finally in § 5 a study is made of infinitely divisible point processes and random measures.The author is indebted to Professor J. F. C. Kingman for advice and encouragement.     

ON THE CONVERGENCE FOR NULL-ARRAYS OF THE POINT PROCESSES

The purpose of this paper is first to establish a representation of the Laplace transformfor the regular infinitely divisible point processes,and then to give a sufficient and neccesarycondition for convergence of the null-arrays toward a regular infinitely divisible pointprocess.     

Some recent results in infinite divisibility

In this paper, a survey is given of some recent developments in infinite divisibility. There are three main topics: (i) the occurrence of infinitely divisible distributions in applied stochastic processes such as queueing processes and birth-death processes, (ii) the construction of infinitely divisible distributions, mainly by mixing, and (iii) conditions for infinite divisibility in terms of distribution functions and densities.     

On Infinitely Divisible Distributions on Locally Compact Abelian Groups

Our aim in this paper is to characterize some classes of infinitely divisible distributions on locally compact abelian groups. Firstly infinitely divisible distributions with no idempotent factor on locally compact abelian groups are characterized by means of limit distributions of sums of independent random variables. We introduce semi-selfdecomposable distributions on topological fields, and in case of totally disconnected fields we give a limit theorem for them. We also give a characterization of semistable laws on p-adic field and show that semistable processes are constructed as scaling limits of sums of i.i.d.     

Random matrix models of stochastic integral type for free infinitely divisible distributions

The Bercovici-Pata bijection maps the set of classical infinitely divisible distributions to the set of free infinitely divisible distributions. The purpose of this work is to study random matrix models for free infinitely divisible distributions under this bijection. First, we find a specific form of the polar decomposition for the Lévy measures of the random matrix models considered in Benaych-Georges [6] who introduced the models through their laws. Second, random matrix models for free infinitely divisible distributions are built consisting of infinitely divisible matrix stochastic integrals whenever their corresponding classical infinitely divisible distributions admit stochastic integral representations. These random matrix models are realizations of random matrices given by stochastic integrals with respect to matrix-valued Lévy processes. Examples of these random matrix models for several classes of free infinitely divisible distributions are given. In particular, it is shown that any free selfdecomposable infinitely divisible distribution has a random matrix model of Ornstein-Uhlenbeck type ?? ₀ ^?? e ^?1 d?? _t ^d , d ?? 1, where ?? _t ^d is a d × d matrix-valued Lévy process satisfying an I _log condition.     

A random matrix model for the Gaussian distribution

Summary The Gaussian unitary ensemble is a random matrix model (RMM) for the Wigner law. While random matrices in this model are infinitely divisible, the Wigner law is infinitely divisible not in the classical but in the free sense. We prove that any variance mixture of Gaussian distributions -- whether infinitely divisible or not in the classical sense -- admits a RMM of non Gaussian infinitely divisible random matrices. More generally, it is shown that any mixture of the Wigner law admits a RMM. A key role is played by the fact that the Gaussian distribution is the mixture of Wigner law with the <InlineEquation ID=IE"1"><EquationSource Format="TEX"><![CDATA[<InlineEquation ID=IE"2"><EquationSource Format="TEX"><![CDATA[$]]></EquationSource></InlineEquation>]]></EquationSource></InlineEquation>2$-gamma distribution.     

Decomposition of infinitely divisible characteristic functions without Gaussian component

Summary The problem of characterizing the infinitely divisible characteristic functions which have only infinitely divisible factors is considered. Under the assumption that both the absolutely continuous and the singular (or the discrete) components exist in Poisson spectral measures, several necessary conditions for this problem are obtained. These conditions admit partial converses and new examples of infinitely divisible characteristic functions which have only infinitely divisible factors are given.     

The multitype branching diffusion

The multitype branching diffusion (MBD) is considered. A review of the general theory of multitype point processes is given in Section 2, and spatial central limit theorems for homogeneous infinitely divisible processes are proven in Section 3. In Section 4, the MBD is defined, and equations for its first four factorial moment density functions are found. The behaviour of the mean and covariance functionals as time approaches infinity is studied. The MBD with immigration (MBDI) is introduced in Section 5. The existence of a steady state is proven, and spatial central limit theorems are developed for the MBDI.     

Ordered thinnings of point processes and random measures

This is a study of thinnings of point processes and random measures on the real line that satisfy a weak law of large numbers. The thinning procedures have dependencies based on the order of the points or masses being thinned such that the thinned process is a composition of two random measures. It is shown that the thinned process (normalized by a certain function) converges in distribution if and only if the thinning process does. This result is used to characterize the convergence of thinned processes to infinitely divisible processes, such as a compound Poisson process, when the thinning is independent and nonhomogeneous, stationary, Markovian, or regenerative. Thinning by a sequence of independent identically distributed operations is also discussed. The results here contain Renyi's classical thinning theorem and many of its extensions.     

ON THE STRUCTURES OF RANDOM MEASURE AND POINT PROCESS CONVOLUTION SEMIGROUPS

§0.ＩｎｔｒｏｄｕｃｔｉｏｎＫｅｎｄａｌｌ[1]ｆｏｕｎｄｅｄｔｈｅＤｅｌｐｈｉｃｓｅｍｉｇｒｏｕｐｔｈｅｏｒｙａｎｄｕｓｅｄｔｈｉｓｔｈｅｏｒｙｔｏｓｔｕｄｙｔｈｅｓｔｒｕｃｔｕｒｅｓｏｆｔｈｅｒｅｎｅｗａｌｓｅｑｕｅｎｃｅｓｅｍｉｇｒｏｕｐａｎｄｔｈｅｓｔａｎｄａｒｄｐ-ｆｕｎｃｔｉｏｎｓｅｍｉｇｒｏｕｐ.Ｄａｖｉｄｓｏｎ[2,3],ＲｕｚｓａａｎｄＳｚ啨ｋｅｌｙ[4]ａｎｄＨｅ[5,6]ｇｅｎｅｒａｌｉｚｅｄｔｈｅＤｅｌｐｈｉｃｓｅｍｉｇｒｏｕｐｔｈｅｏｒｙａｎｄｕｓｅｄｔｈｅｉｒｒｅｓｕｌｔｓｔｏｓｔｕｄ…     

Thorin classes of Lévy processes and their transforms

We define and characterize Thorin classes {ie294-01}, of infinitely divisible distributions on R ₊. We investigate Poisson, Karlin, and Bessel transforms of Thorin classes and also consider extended Thorin classes {ie294-02}. Canonical representation and self-decomposability properties of Thorin subordinated Gaussian Lévy processes are discussed. As an example, a subordinated Cauchy process is considered in detail.     

Testing independence of variates in an infinitely divisible random vector

A method is given for testing the independence of variates in an infinitely divisible random vector and for testing the independence of several subsets of the variates. Applications to stochastic processes are indicated.     

Equivalence of measures induced by infinitely divisible processes

We find sufficient conditions for the equivalence of two measures on function space induced by infinitely divisible processes. The processes are not assumed to be stochastically continuous or to have independent increments. The theorem proved here is equivalent to known results in the special case of stochastically continuous processes with independent increments.     

Limit Theory for the Sample Autocovariance for Heavy-Tailed Stationary Infinitely Divisible Processes Generated by Conservative Flows

This study aim to develop limit theorems on the sample autocovariances and sample autocorrelations for certain stationary infinitely divisible processes. We consider the case where the infinitely divisible process has heavy tail marginals and is generated by a conservative flow. Interestingly, the growth rate of the sample autocovariances is determined not only by heavy tailedness of the marginals but also by the memory length of the process. Although this feature was first observed by Resnick et al. (Stoch Process Appl 85:321–339, 2000) for some very specific processes, we will propose a more general framework from the viewpoint of infinite ergodic theory. Consequently, the asymptotics of the sample autocovariances can be more comprehensively discussed.     

On the excursion theory for linear diffusions

We present a number of important identities related to the excursion theory of linear diffusions. In particular, excursions straddling an independent exponential time are studied in detail. Letting the parameter of the exponential time tend to zero it is seen that these results connect to the corresponding results for excursions of stationary diffusions (in stationary state). We characterize also the laws of the diffusion prior and posterior to the last zero before the exponential time. It is proved using Krein’s representations that, e.g. the law of the length of the excursion straddling an exponential time is infinitely divisible. As an illustration of the results we discuss the Ornstein–Uhlenbeck processes.     

Spectral representations of infinitely divisible processes

Summary The spectral representations for arbitrary discrete parameter infinitely divisible processes as well as for (centered) continuous parameter infinitely divisible processes, which are separable in probability, are obtained. The main tools used for the proofs are (i) a polar-factorization of an arbitrary Lévy measure on a separable Hilbert space, and (ii) the Wiener-type stochastic integrals of non-random functions relative to arbitrary infinitely divisible noise.The research of both authors was supported partially by the AFSOR Grant No. 87-0136; the second named author was also supported partially by a grant from the University of Tennessee     

A new characterization of Laplace functionals and probability generating functionals

Summary The well-known and widely used Laplace resp. probability generating functionals are characterized by properties of positive definiteness and continuity. The methods applied come from Harmonic Analysis on semigroups, and allow also intrinsic characterizations for the transforms of infinitely divisible random measures and point processes.