Generalized fractional Ornstein-Uhlenbeck processes
|
|
- Sheena Harrington
- 5 years ago
- Views:
Transcription
1 Generalized fractional Ornstein-Uhlenbeck processes Kotaro ENDO and Muneya MATSUI Department of Mathematics, Keio University Center for Mathematical Science, Munich University of Technology and Department of Mathematics, Keio University arxiv:87.11v1 [math.pr] 14 Jul 8 October 5, 18 Abstract We introduce an extended version of the fractional Ornstein-Uhlenbeck (FOU process where the integrand is replaced by the exponential of an independent Lévy process. We call the process the generalized fractional Ornstein-Uhlenbeck (GFOU process. Alternatively, the process can be constructed from a generalized Ornstein-Uhlenbeck (GOU process using an independent fractional Brownian motion (FBM as integrator. We show that the GFOU process is well-defined by checking the existence of the integral included in the process, and investigate its properties. It is proved that the process has a stationary version and exhibits long memory. We also find that the process satisfies a certain stochastic differential equation. Our underlying intention is to introduce long memory into the GOU process which has short memory without losing the possibility of jumps. Note that both FOU and GOU processes have found application in a variety of fields as useful alternatives to the Ornstein-Uhlenbeck (OU process. Keywords: Ornstein-Uhlenbeck processes, Lévy process, Stochastic integral, Long memory, Fractional Brownian motion. 1 Introduction The fractional Brownian motion (FBM is one of the most popular processes for constructing long-range dependent stochastic processes with continuous path and its fields of applications are very wide. To name just a few, we see FBM models in the fields of telecommunications, signal processes, environmental models and economics. A recent reference is e.g., Doukhan et al. (3. Statistical methods for FBM have also been studied (see e.g. Beran (1994. We review the definition and name properties of FBM. Definition 1.1 Let < H 1. A fractional Brownian motion B H := {Bt H} t R is a centered Gaussian process with B H = and Cov ( ( Bt H,Bs H = 1 t H + s H t s H, t,s R. Correspondence to Department of Mathematics, Keio University, Hiyoshi Kohoku-ku, Yokohama 3-85, Japan. mmuneya@gmail.com. This research is supported by JSPS Research Fellowships for Young Scientists. 1
2 From the definition FBM has stationary increments and is self-similar with index H, i.e., for c > {Bct H} d t R = {c H Bt H} t R where = d denotes equality of all finite dimensional distributions. While {Bt H } t R with H = 1/ is a two-sided Brownian motion (BM and has independent increments, {Bt H} t R with H (, 1 (1,1] has dependent increments. For < h < s and t R and N N, Γ h (s := Cov(B t+h B t,b t+s+h B t+s = Cov(B h,b s+h B s ( n 1 h n = (H k s H n (n! n=1 k= ( N n 1 h n = (H k s H n +O ( s H N, as s. (n! n=1 k= Thus, B H with H ( 1,1] has a long memory property, namely, n= Γ h(nh =. Finally B H is knowntohaveboundedp-variationfor1/h < p < (seeproposition.ofmikosch and Norvai sa (. For a more detailed theoretical treatment, we refer to Embrechts and Maejima ( or Samorodnitsky and Taqqu (1994. On the other hand, recently, extensions of the classical OU process have been suggested mainly on demand of applications. The generalized Ornstein-Uhlenbeck process given below is one with Lévy processes and plays important role in Economics (pricing of Asian options, perpetuities and risk theory. For its theories and applications, we refer to e.g, Carmona et al. (1, Erickson and Maller(5, Lindner and Maller(5 and Klüppelberg and Kostadinova(8. A multivariate extension is also considered in Kondo et al. (6 and Endo and Matsui (8. Let {(ξ t,η t } t be a bivariate Lévy process and V be an independent initial random variable. A generalized Ornstein-Uhlenbeck (GOU process is defined as V t = e (V ξt + e ξ s dη s, t. (1 The stationarity and the convergence or divergence property have been intensively studied. If {ξ t } t and {η t } t are independent and e ξ s dη s converges a.s. as t to a finite random variable, V := {V t } t has the stationary version (see e.g. Remark. of Lindner and Maller (5. The short memory property of V was also shown in Section 4 of Lindner and Maller (5. Another extension of the original OU process is the fractional Ornstein-Uhlenbeck process, where FBM is used as integrator. An advantage of using the process is to realize stationary long range dependent processes. Let λ > and an initial random variable X H L 1. A fractional Ornstein-Uhlenbeck (FOU process is defined as X H t = e λt ( X H + e λs dbs H. ( Here we need a non-semimartingale approach to construct stochastic integrals with FBM. We can find several useful theoretical tools in e.g., Lin (1995, Mikosch and Norvai sa ( or Pipiras and Taqqu (. Cheridito et al. (3 has shown that the FOU process is the unique continuous solution of a Langevin equation: X t = X H λ X sds+bt H, t and investigated its dependence properties. The main purpose of this paper is to construct a version of the GOU process which allows for long memory modeling by the use of a FBM.
3 as In order to define a generalized Ornstein-Uhlenbeck process we define a two-sided Lévy process ξ t := { ξ 1 t if t ξ t if t <, (3 where {ξ 1 t} t and {ξ t} t are independent copies of {ξ t } t. We work throughout with a bivariate complete probability space (Ω := Ω 1 Ω, F := F 1 F,P := P 1 P. (4 Let {ξ t } t R defined on (Ω 1, F 1,P 1 be a Lévy process and a FBM {Bt H} t R with Hurst index H (,1 defined on (Ω, F,P which is independent of {ξ t } t R. A generalized fractional Ornstein-Uhlenbeck (GFOU process with initial value Y L 1 (Ω is defined as Y t := e (Y ξt + e ξ s dbs H, t. (5 If the initial variable satisfies Y = then, for convenience, we sometimes replace Y = {Y t } t with Y t := e ξt e ξ s db H s, (6 e ξ s db H s. (7 The process Y is regarded as an extension of V given in (1 where the stochastic process of integration {η t } t is replaced with a {Bt H} t with H (,1 and also is regarded as an extended version of the FOU process where the integrand is replaced by the exponential of an independent Lévy process ξ t. We should remark that Y has jumps caused by the process e ξt. The paper is organized as follows. In Section.1 we recall the definition of Lévy processes and summarize properties needed. In Sections..1 and.. we review Riemann-Stieltjes integrals for functions with bounded p-variation and the stochastic integral in the L (Ω-sense respectively. In Section 3 we investigate the existence of the integral in the GFOU process in order to justify the definition of the GFOU process. The stationarity condition and the second order behavior of the GFOU process are discussed in Section 4, and we observe the long memory property. Here we also examine stochastic integrals constructed by a single FBM, where ξ in the process Y is replaced with B H used as the integrator. In Section 5 we obtain a stochastic differential equation, whose solution is given in form of the GFOU process. We use the following notations throughout. Write a.s. = if equality holds almost surely. We will take the expectations for a bivariate process {(Zt 1,Z t } t R. If the expectations only for a process {Zt} 1 t R is considered, we write its expectation as E Z 1. Preliminaries.1 Lévy processes In this subsection we introduce the setup for the Lévy process. Let ξ := {ξ t } t be a Lévy process on R with (a ξ,ν ξ,γ ξ generating triplet, where a ξ and γ ξ R are constants and a measure ν ξ 3
4 on R\{} satisfies R\{} (1 x ν ξ (dx <. We call ν ξ the Lévy measure of ξ. Then, the characteristic function of ξ t at time t = 1 is written as [ Ee izξ 1 = exp a ] ξ ( z +iγ ξ z + e izx 1 izx1 { x 1} νξ (dx, z R. (8 R\{} For more on Lévy processes and their properties, we refer to Sato (1999. In later sections we consider several examples related the α-stable Lévy motion with index < α <, denoted by ξ α := {ξt α} t. It is a Lévy process and its generating triplet is (,ν α,γ α where { c1 x ν α (dx := 1 α dx on (, c x 1 α dx on (, with c 1, c and c 1 +c. In order to define the in (5, the variation of the Lévy process plays an important role. We give a brief summary based mainly on Section 5.4 of Dudley and Norvaiša (1998 and p.48 of Mikosch and Norvai sa (. Define the p-variation for < p < of a process X := {X t } t R on [t 1,t ] for t 1 < t in R as v p (X := v p (X,[t 1,t ] := sup Xsi X si 1 p, (9 where is a partition t 1 = s < s 1 < < s n = t of [t 1,t ] and n 1. If v p (X,[t 1,t ] < for all t 1 < t, we say X has bounded p-variation, and if v 1 (X,[t 1,t ] < we say it is of bounded variation. Since every Lévy process is a semimartingale v p (ξ < for p (see Lépingle (1976. We will state three useful results which characterize p-variation of Lévy process in terms of the Lévy measure. Unfortunately we can not find a result which uniformly characterize the variation in terms of the Lévy measure. Assumptions and results are somewhat different from paper to paper. The first one is a well-known result (e.g. Theorem 1,9 (i of Sato ( Bounded variation A Lévy process {ξ t } t is of bounded variation if and only if a ξ = and (1 x ν ξ (dx <. R\{} The next result is a combination of Theorems 4.1 and 4. of Blumenthal and Getoor (1961 and Theorem of Monroe (197..(a p-variation of Lévy processes Define β = inf{α > : x <1 x α ν ξ (dx < }. We call β the Blumenthal and Getoor index. If the Lévy process {ξ t } t has no Gaussian component (a ξ =, then i=1 v p (ξ;[,1] < a.s. if p > β v p (ξ;[,1] = a.s. if p < β. (1 4
5 The result by Bretagnolle (197 is a sharpened version of.(a but with zero mean assumption..(b p-variation of Lévy processes Let 1 < p < and {ξ t } t be a Lévy process without Gaussian component (a ξ =. Then v p (ξ;[,1] < a.s. if and only if (1 x p ν ξ (dx <. Otherwise v p (ξ;[,1] = a.s. R\{} In particular for α-stable Lévy processes we have the result by Fristedt and Taylor (1973 which was stated in convenient form in Mikosch and Norvai sa (. 3. p-variation of α-stable Lévy processes Let {ξ α t } t be α-stable Lévy motion. Assume that γ α = for α < 1 and that the Lévy measure is symmetric for α = 1. Then v p (ξ α is finite or infinite with probability 1 according as p > α or p α. For the existence of the infinite interval integral in {Y t } t given in (7 we further need the a.s. behavior of {ξ t } t as t. Our assumption is that lim t ξ t = +. Doney and Maller ([Theorem 4.4] have obtained an equivalent condition of this in terms of the Lévy measure ν ξ. Since several paperswell explain equivalent conditions, (see p.7of Erickson and Maller (5 a.s. or p.174 of Lindner and Maller (5 we do not mention it. Actually if lim t ξ t = + holds, we can assert a stronger result, which is more useful for our aim. Lemma.1 Suppose lim t ξ t a.s. = +. Then for almost all ω 1 Ω 1 there exist δ > and t = t (ω 1 < such that ξ t > δt for t t. The proof is a combination of Theorem 4.3 and 4.4 in Doney and Maller (. Concerning the integral e ξ s dη s in the GFOUprocess given by (1, Erickson and Maller (5 have characterized theconvergence ofimproper integral e ξ s dη s intermsofthelévymeasure of{(ξ t,η t } t, a.s. in which the condition lim t ξ t = + was used.. Integrals with respect to functions with unbounded variation..1 Riemann-Stieltjes integrals with p-variation We review several useful definitions of integrals of functions which have unbounded variation but bounded p-variation. The excellent introduction to this area is given by Dudley and Norvaiša (1998. Let f and g be real functions on [a,b]. Define κ = {u 1,...,u n } to be an intermediate partition of = [a = s < s 1 < < s n = b] given as in (9, namely, s j 1 u i s j for i = 1,...,n. A Riemann-Stieltjes sum is defined as S RS (f,g,,κ := f(u i [g(s i g(s i 1 ]. i=1 5
6 Then we say that the Riemann Stieltjes integral exists and equals to I, if for every ǫ >, there exists δ > such that S RS (f,g,,κ I < ǫ for all partition with mesh max (s i s i 1 < δ and for all intermediate partitions κ of. The following theorem was proved by Young (1936. (See also Theorem.4 of Mikosch and Norvai sa ( or Theorem 4.6 of Dudley and Norvaiša (1998. Theorem.1 Assume f has bounded p-variation and g has bounded q-variation on [a, b] for some p,q > with p 1 +q 1 > 1. Then the integral b fdg exists in the Riemann-Stieltjes sense, and a the inequality b fdg C p,q(v p (f 1/p (v q (g 1/q a holds with C p,q = n 1 n (p 1 +q 1... Integral with respect to FBM in L -sense Another definition of the integral is in L (Ω -sense. Stochastic integrals with respect to FBM is sometimes defined as the L (Ω -limits of integrals of step functions (see e.g. Lin (1995. We see this when B H is the Brownian motion B 1/. If a function f(u : R R satisfies f(u L (R, there exist step functions f n (u := f i 1 {ui <u u i+1 }, < u 1 <... < u n+1 <, f i R, n N. i=1 such that f n converges to f in L (R. Then the integral R f(udb1/ u integrals of step functions R f n (udb 1/ u = i=1 ( f i Bu 1/ i+1 Bu 1/ i, is the L (Ω -limit of the since L (R and L (Ω are isometric and their inner products are equal, namely, [ ( ] E (f n (u f(udbu 1/ = (f n (u f(u du. R R Pipiras and Taqqu ( have investigated a similar characterization for the integral of B H when H 1. We apply this to the existence of the improper integral in the GFOU process {Y t} afterward. Define the linear space { } Sp(B H := X : f n (udbu H X in L (Ω for some sequence (f n n N (step functions. R Pipiras and Taqqu ( have analyzed the functional space of the integrand f(u in which it can be asymptotically approximated by f n (u and R f(udbh u is well-defined. For H (, 1 they succeeded in specified a Hilbert space of functions on the real line which is isometric to Sp(B H. However, for H ( 1,1 they had difficulty in finding the corresponding isometric space, and 6
7 as second best they analyzed inner product spaces in which the integral with B H (H ( 1,1 is well-defined. We give only one such inner product space and its inner product for B H with H ( 1,1. Other inner product spaces do not seem to work for our purpose since they require e.g. characteristic function of f or fractional derivative of f which do not exists in our case where f(u = e ξu. (See Section 7 of Pipiras and Taqqu (. The space is Λ H = { f : f, f Λ H < } ( 1, H,1, (11 where f,g Λ H = c H is the inner product with c H := H(H 1. 3 Existence of the integral R R f(ug(v u v H dudv In this Section we analyze the existence of the integral in the GFOU process given in (5. The definition of the GFOU process includes an integral with respect to FBM. Since paths of FBM are of infinite variation and since FBM with H 1 is not a semimartingale, the stochastic integral with respect to FBM (H 1 is not an Itô integral. Additionally, the integrand of the GFOU process is random and the infinite interval integral (6 is needed for its stationarity. We apply two approaches of the integral in Section. in order to cope with these problems. Proposition 3.1 Let B H := {Bt H} t R be a FBM with H (,1 and ξ := {ξ t } t R be an independent two-sided Lévy process. Assume that {ξ t } t R has bounded p-variation for < p <. Then the integral eξ s dbs H, < t < exists in the path-wise Riemann-Stieltjes sense if 1 +H > 1. p Furthermore with q > 1 and C H p,q = +q 1 n 1 we have n (p 1 ( e ξ s dbs H C p,q sup e ξs (v p (ξ 1/p( ( v q B H 1/q P a.s. s [,t] The proof of Proposition 3.1 is obvious from the continuity of the exponential function and we omit it. Remark 3.1 If p = 1 in Proposition 3.1, ξ has bounded variation and we can define path-wise integrals for all B H with H (,1. On the other hand if H > 1, we can define the path-wise integral for any integrands ξ since v p (ξ < for p >. The reason why we need the path-wise definition besides the L (Ωp-approach is that we want to define the integral for H (, 1, which we could not obtain in the L (Ω-approach. It also gives several useful tools easily, such as integration by parts or chain rule for analyzing stochastic differential equations in Section 5. Example 3.1 Set ξ := ξ α be an α-stable Lévy motion with index α defined in Subsection.1. Then under assumption in 3. p-variation of α-stable Lévy processes and the assumption 1 +H > 1. the integral α e ξα t eξα u db H u exists in the path-wise sense. Hence the GFOU process is well-defined. Both ξ α and B H have infinite variation and the former is known to be extremely heavy tailed process. 7
8 Recall that we write Y t = Y t (ω 1,ω to emphasis that it is a function from probability space (4 and that we write Y t if initial value Y satisfies (6. Define approximating step functions of the integrand e ξt+ξ u of Y t as f t,n (u;ω 1 := f i (u;ω 1 1 {u n i 1 <u u n i }, i=1 where f i (u;ω 1 := e ξt(ω 1+ξ u n (ω 1 i 1. Here u n i s are points in [ N n,t] such that N n = u n < u n 1 < < u n n = t and as n max (u n i u n i 1 and N n. By integrating f t,n with respect to Bt H, we also define approximating sequence as Z n t (ω 1,ω := = f t,n (u;ω 1 db H u (ω f i (u;ω 1 (Bu H n i (ω BH u (ω. n i 1 i=1 In the following theorem we define the integral e ξt ξ u db H u as the limit in probability of the Z n t as n. The reason why we need this approach is that with only path-wise definitions we find difficulty to treat improper integrals. For the existence of the improper integral we should consider long time (t behavior of both {B H t } t R and {ξ t } t R path-wisely, which seems to be not an easy task. Additionally, this approach is well-matched with the analysis of the second order behavior. Theorem 3.1 Let {Bt H} t R be a FBM with H ( 1,1 and {ξ t} t R be an independent two-sided a.s. Lévy process. If lim t ξ t = +, then for each t Zt n given in (1 converges in probability to a limit defined by Y t and which does not depend on the sequence u n i. If further E[e ξ 1 ] < 1, then Zt n converges to Y t in L (Ω and it follows that [ ] E[Y t ] = E e (ξt ξu dbu H = and for < s t, E[Y s Y t ] = s E ξ [e ξs+ξ u ξ t+ξ v ] u v H dudv. (1 Remark 3. (a For H (, 1 we could not validate the existence of the improper integral in Y t. a.s. (b Under the only condition that lim t ξ t = +, we cannot prove the L (Ω convergence. (c The condition E[e ξ 1 a.s. ] < 1 implies lim t ξ t = +, which is shown in Proposition 4.1 of Lindner and Maller (5. Proof of Theorem 3.1. We check that for each ω 1 the integrand e ξt(ω 1+ξ u (ω 1 Λ H given in (11. Since ξ t (ω 1 is constant, we drop it and only show e ξ u (ω 1 Λ H. The function e ξ u (ω 1 is càglàd and bounded 8
9 on any finite interval. Due to Lemma.1 and to the symmetry of two-sided Lévy processes there exists T(ω 1 < such that for all u T(ω 1, ξ u < δu where δ is some positive constant. Then for each ω 1, e ξ u (ω 1 +ξ v (ω 1 u v H dudv < is obvious. Hence we can utilize L (Ω -integral theory in Section... Namely, for each t and for each fixed ω 1, Z n t (ω 1, converges in L (Ω,P. Moreover Z n t converges in probability on (Ω,P for each t since sequence Z n t satisfies the Cauchy criterion, as seen by lim P n,m ( Zn t Zm t > ǫ = lim 1 { Z n n,m t Zt m >ǫ} dp dp 1 Ω 1 Ω = lim P ( Zt n (ω 1 Zt m (ω 1 > ǫdp 1 =. n,m Ω 1 The limit is called Y t and it is F 1 F measurable for each t. Now, with E[e ξ 1 ] < 1, we prove the L (Ω-convergence. We have E[e ξ 1 ] < 1 as well, hence θ 1 := loge[e ξ 1 ] > and θ := loge[e ξ 1 ] >. Then using the covariance of the FBM, we have [ ] E[(Zt n ] = E e ξt+ξ u n +ξ u i 1 n j 1 (B H u n i Bu H n (BH i 1 u n BH j u n j 1 = i=1 i=1 j=1 = c H j=1 [ ] {E ξ e ξt+ξ u n +ξ u i 1 n j 1 1 ( u n i u n j H + u n i 1 un j H + u n i un j 1 H u n i 1 un j 1 H} Γ n (u,v u v H dvdw, (13 where Γ n t (u,v = i=1 j=1 { 1 {u n i 1 <u u n i,un j 1 <v un j,i j} e θ (t u n i 1 θ 1(u n i 1 un j 1 which obviously converges point-wise to +1 {u n i 1 <u u n i,un j 1 <v un j,i<j} e θ (t u n j 1 θ 1(u n j 1 un i 1 }, 1 {u v} e θ (t u θ 1 (u v +1 {u<v} e θ (t v θ 1 (v u. 9
10 Since Γ n t(v,w < M for some M > uniformly in n, t, (13 is bounded by c H M t H. Furthermore according to usual Fubini s theorem, it also follows that E [ (Y t ] ( = Y t (ω 1,ω dp dp 1 Ω 1 Ω [ t ] = c H E ξ e ξt+ξ u +ξ v u v H dudv = c H e θ (t u θ 1 (v u 1 {u v} u v H dudv <. Observe that this integral is finite. Accordingly E[(Zt n ] E [ (Y t ] as n. Now we can apply Theorem of Chung (1 to Zt n and obtain the L (Ω-convergence. In consequence E[Y s Y t ] turns out to be finite and equation (1 follows from Fubini s theorem. Finally E[Zt n ] =, n N implies E[Y t ] =. Hence the proof is complete. The process {Y t } t R obtained in Theorem 3.1 is the GFOU process with initial value Y. We close this section with the following concluding Remarks. Remark 3.3 (a In both Proposition 3.1 and Theorem 3.1 ξ is independent of B H and we have b b e ξt e ξ u dbu H = e ξt+ξ u dbu H, a < b <. a a This is not allowed in usual theory of stochastic integrals related to semimartingale (Protter (4 since {ξ t } t R is not adapted. a.s. (b From Theorem 3.1 lim t ξ t = + implies Y = eξ s dbs H < a.s. Ω. Therefore together with Proposition 3.1 we can also treat {Y t } t = {Y t +Y } t path-wisely. (c We should mention that Erickson and Maller (5[p.81] gave another idea for improper integrals of a function of Lévy processes with respect to FBM: g(ξ t dbt H. Investigation of their idea is also our next concern. 4 Stationarity and Second order behavior of GFOU processes Here we investigate the strict stationarity and the second order behavior of the GFOU process Y := {Y t } t. Since we could not validate the existence of Y with Hurst parameter H (, 1 and since our main concern in this paper is the long memory case, we confine our results to the case H ( 1,1 throughout this section. The stationarity of Y is as follows. Proposition 4.1 If lim t ξ t a.s. = +, then Y t exists for all t and the process Y := {Y t } t is strictly stationary. Proof of Proposition 4.1. Let t 1 < t <... < t m <, m N and h >. We use the sequence Z n t given in (1. Since 1
11 both {ξ t } t R and {B H t } t R have stationary increments, does the pair {(ξ t,b H t } t R as well because of independence. Thus, Z n t i = d = e ξt+ξ u n j 1 (B H u n j Bu H, N n n = u n j 1 1 <... < un n = t i j=1 e ξ t+h+ξ u n +h j 1 (Bu H n j +h BH u n +h j 1 j=1 simultaneously in i and we have = Z n t i 1 +h, 1 i m, (Z n t 1,...,Z n t m d = (Z n t 1 +h,...,zn t m+h. Then by virtue of Theorem 3.1 as n, (Z n t 1,...,Z n t m converges in probability to (Y t1,...,y tn and (Z n t 1 +h,...,zn t m+h converges in probability to (Y t 1 +h,...,y tn+h. This yields (Y t1,...,y tn d = (Y t1 +h,...,y tn+h and the conclusion holds. Remark 4.1 (a The logic in the proof works even in the case H (, 1. If the assumptions of Proposition 3.1 are satisfied and if the integral Y = eξ s dbs H with H (, 1 exists, then Y with H (, 1 is defined and is strictly stationary. (b In connection with the GOU process {V} t given in (1. Proposition 4.1 corresponds to Theorem 3.1 of Carmona et al. (1, where ξ and η are independent, and under conditions of a.s. convergence of the integral e ξ a.s. s dη s and lim t ξ t = + the stationary version exists and equals in distribution to V. Next we investigate the second order behavior of Y and derive the auto-covariance function explicitly. What should be remarked is that while the auto-covariance function of the GOU process V given in (1 decreases exponentially (Theorem 4. of Lindner and Maller (5, that of the FOU process {X H t } t given in (1 decays like a power function (Theorem.4 and Corollary.5 in Cheridito et al. (3. Since Y is regarded as a version of GOU processes and FOU processes, this investigation is interesting. We utilize results in Theorem 3.1 and obtain Theorem 4.1 and Corollary 4.1 below. Note that even the existence of Y and the equation (1 are obtained several difficulties still lay in calculating the auto-covariance function. The integrand in (1 is regarded as exponential moment of of 4 dependent random variable, i.e. E ξ [e (ξs ξ u +ξ t ξ v ] and dependencies of these variables are different in the order of s,u,t and v. We also require that after E ξ [e (ξs ξ u +ξ t ξ v ] is calculated the double integral in (1 has a suitable representation for our purpose. In the proofs of Theorem 4.1 and Corollary 4.1 we will see how to get over these difficulties. Proofs of Theorem 4.1 and Corollary 4.1 are given in Appendix A since they require a lot of technical and tedious calculations. Theorem 4.1 Let {B H t } t R be a FBM with H ( 1,1 and {ξ t} t R be an independent two-sided Lévy process. Suppose that E[e ξ 1 ] < 1. Then the stationary version Y := {Y t } t exists and for 11
12 any s >,t, we have Cov(Y t,y t+s = Cov(Y,Y s ( e θ 1 s = c H Γ(H 1 e θ 1s Γ(H 1 θ θ1 H 1 θ1 H + e θ 1s s H 1 θ 1 (H 1 1 F 1 (H 1,H;θ 1 s+ eθ1s Γ(H 1,θ θ H 1 s [ e θ 1 s = c H Γ(H 1 e θ 1s Γ(H 1 θ θ1 H 1 θ1 H + sh 1 n+1 { (H k θ 1 (H 1 n= k=1 1 (θ 1 s (n+1 ( θ 1 s (n+1γ(n+1,θ 1s n! where θ 1 := log(e[e ξ 1 ] > and θ := log(e[e ξ 1 ] >. Here γ(, and Γ(, are incomplete gamma functions in 8.35 of Gradshteyn and Ryzhik ( and 1 F 1 (, ; is the confluent hyper-geometric function in 9.1 of Gradshteyn and Ryzhik (. Note that since γ(n+1,sθ 1 Γ(n+1 = n! as s, we have Cov(Y t,y t+s = H n=1 n 1 k=1 = O(s H. (H kθ n 1 s H n +O(e θ 1s This conclude that Y with H ( 1,1 is a long memory process. While we obtained Cov(Y t,y t+s using special functions in Theorem 4.1, it is mainly for numerical purpose since for such functions useful softwares are available. Next we investigate long time dependence of {Y t } t with the initial value Y := X L (Ω where X is independent of {ξ t } t and {B H t } t. Corollary 4.1 Let Y := {Y t } t, H ( 1,1 be a GFOU process with the initial value X L (Ω, where X is independent of ξ := {ξ t } t and B H := {B H t } t. Then for fixed t as s. Cov(Y t,y t+s = H n=1 n 1 k=1 = O(s H. }], (H kθ n 1 {s H n e θ 1t (t+s H n }+O(s H N We see what happens to the second order behavior of the process Y if ξ in Y is replaced with B H which is the same process as the variable of integration. Although we expect long memory this does not hold. Note that we need only the probability space (Ω, F,P here. For H ( 1,1 and the initial random variable X L (Ω independent of {Bt H} t R, define W t = e BH t ( X + e BH u db H u. To analyze {W t } t we use the path-wise integral theory (see Subsection..1. Let f be continuous differentiable and F(x = F(+ x f(ydy. Then with H (1,1 it follows that F(Bt H F(BH = f(bt H dbh u 1 a.s.
13 By setting f = e t in above we obtain W t = 1+e BH t (X 1. Proposition 4. Let H ( 1,1 and t,s. Define M 1 := (E[X] 1 and M := E[(X 1 ]. The process {W t } t has the following auto-covariance and correlation functions. } Cov(W t,w t+s = e 1 {th +(t+s H } {M e 1 {th +(t+s H s H} M 1 ( = O e 1 {(t+sh +s H 1 } as s. Corr(W t,w t+s = M e 1 (th s H M 1 e 1 (t+sh M e th M 1 M M 1 e (t+sh We also consider the drift added process = O(e 1 sh as s. Ŵ t = e (BH t +at (X + e BH u +au d(b H u +au, where a > and X L (Ω is independent of {B H t } t. Even if a drift is added, usual path-wise integral works and Ŵ t = 1+(X 1e (BH t +at holds. Then the auto-covariance and correlation functions of {Ŵt} t are calculated is a similar manner and become Cov(Ŵt,Ŵt+s = e a(t+s Cov(W t,w t+s, Corr(Ŵt,Ŵt+s = Corr(W t,w t+s. Thus our conclusion here is that even if a drift is added it does not have long memory. 5 Stochastic differential equation related with GFOU processes We analyze a stochastic differential equation of which a solution is given by the GFOU process. Let U := {U t } t bealévyprocesswithgeneratingtriplet(a U,ν U,γ U. AssumethattheLévymeasure ν U has no mass on (, 1]. The Doléans-Dade exponential of U t is written as E(U t = e ξt where ξ t = U t + a ξ t (log(1+ U s U s. <s t See Section. of Erickson and Maller (5. Here ξ t is the Lévy processes. 13
14 Proposition 5.1 Under the assumption in Proposition 3.1, GFOU {Y t } t with the initial value Y L 1 (Ω satisfies the stochastic differential equation; where E(U t = e ξt. dy t = Y t du t +db H t, (14 Proof of Proposition 5.1. Since the condition of Theorem.1 is satisfied, the integral BH s also exists in the Riemann- deξs Stieltjes path-wise sense. We use the integration by parts formula to Y t and obtain Y t = e ξt (Y B H s deξs +B H t. If we put Q t := Y t B H t, the equation above becomes Q t = e ξt (Q B H s deξs, where Q = Y. Since {e ξt } t is a semimartingale and {Bt H} t is continuous and adapted, the process {Q t } t is also semimartingale. We set R t := e ξt and S t := Q BH s. Then the deξs integration by parts formula for semimartingales (e.g. Corollary, II of Protter (4 yields Q t Q = R t S t R S = = = R s ds s + + e ξ s B s de ξs + Q s du s Observe the relation between e ξt and U t ; + S s dr s +[R,S] t R S + Q s e ξ s de ξs + B H s ( e ξ s de ξs d[e ξ,e ξ ] s. B H s d[e ξ,e ξ ] s Using this we obtain which is equivalent to The proof is now complete. 1 = e ξs e ξs = = + + e ξ s de ξs + du s + Q t Q = Y t Y = e ξ s de ξs +[e ξ,e ξ ] t e ξ s de ξ s +[e ξ,e ξ ] t. + ( Qs +Bs H dus, Y s du s +B H t. 14
15 Remark 5.1 The Lévy measure of {ξ t } t is obtained from that of {U t } t ; ν ξ ((x, = ν U ((,e x 1 and ν ξ ((, x = ν U ((e x 1,. (15 SeeagainSection.ofErickson and Maller (5. Henceifν U isconcretelygiven, using criterion of p-variation in Section.1 we can check the condition of Proposition 3.1. The following technical Lemma is not difficult but useful for the existence of {Y t } t which is directly constructed from the stochastic differential equation (14. The poof is only a calculation and we omit it. Lemma 5.1 Assume that {U t } t is a Lévy process and that {ξ t } t satisfies E(U t = e ξt. Then for < δ <, convergence and divergence of x δ ν ξ (dx and x δ ν U (dx are equivalent. x <1 Example 5.1 As an example we consider the stochastic differential equation (14, where U t is given by an α-stable Lévy motion ξt α (see Section.1. From remark above the Lévy measure ν ξ is given by { c (1 e ν ξ (dx = x 1 α e x dx on (, c 1 (e x 1 1 α e x dx on (,. Observe that x <1 ν ξ (dx x 1 α dx as x and hence variation property of ξ t is the same as that of U t = ξ α t. As a result v p (ξ is finite if p > α. A Proofs of Section 4 Proof of Proposition 4.1. From the stationary version Y is definable. By virtue of the stationarity of Y and Fubini s theorem, we have Cov(Y t,y t+s = Cov(Y t,y t+s = c H E ξ [e (ξs ξ u ξ v ] u v H dudv +c H := I+II. s E ξ [e (ξs ξ u ξ v ] u v H dudv First we consider the integral I. The independent increments property of {ξ t } t R gives E[e (ξs ξ u ξ v ] = E[e (ξs ξu ξv ] = E[1 {u v} e (ξs ξ (ξ ξ u (ξ u ξ v ] +E[1 {u<v} e (ξs ξ (ξ ξ v (ξ v ξ u ] = 1 {u v} E[e (ξs ξ ]E[e (ξ ( u ξ ]E[e (ξ u v ξ ] +1 {u<v} E[e (ξs ξ ]E[e (ξ ( v ξ ]E[e (ξ v u ξ ]. 15
16 The integrand E[e (ξs ξ u ξ v ] is symmetric with respect to u and v, and hence I = c H e θ 1s Then further calculation shows I = c H e θ 1s 1 {u v} e θ u+θ 1 (v u u v H dudv. 1 {u v} e θ u+θ 1 (v u u v H dudv (By change of variables;x = u v = c H e θ 1s e θ u du = c He θ 1s Γ(H 1. θ θ1 H 1 Next we consider II. A similar conclusion as above gives and we have II = c H e θ 1s E[e (ξs ξ v (ξ ξ u ] = E[e (ξs ξv (ξ ξ u ] s e θ 1u (By change of variables;x = v u = c H e θ 1s u e θ 1u s u e θ 1x x H dx (16 = e θ 1(s v+θ 1 (u+v = e θ 1s+θ 1 (u+v (u v H dudv (17 e θ 1(u+x x H dudx (18 = c H e θ 1s 1 { x<u<s x} e θ 1u+θ 1 x x H dx ( s s x = c H e θ 1s e θ1x x H dx e θ1u du+ e θ1x x H dx e θ1u du x s x { s ( = c H e θ 1s 1 1 e θ 1x e θ1x x H dx θ 1 θ 1 ( 1 + e θ1(s x 1 } s θ 1 θ e θ 1x e θ1x x H dx ( = c H e θ 1s 1 s e θ1x x H dx 1 e θ1x x H dx (19 θ 1 θ 1 + eθ 1s e θ1x x H dx θ 1 s ( = c H e θ s 1s H+1 θ 1 (H 1 1 F 1 (H 1,H;θ 1 s 1 Γ(H 1 θ1 H Γ(H 1,θ 1 s + eθ 1s θ1 H 16
17 Combining I and II, we obtain the first assertion. The series representation of the incomplete Gamma function, i.e. gives Γ(α,x = x α 1 e x (1+ Γ(H 1,θ 1 s = We apply the binomial expansion (1 u/s H = to the representation n= n=1 s H 1 θ 1 (H 1 1 x n(α 1 (α n n+1 (H k(θ 1 s (n+1 ( n= k=1 (H (H 3 (H n 1 n! ( u s n, ( < u < s e θ1s s H 1 θ 1 (H 1 1 F 1 (H 1,H;θ 1 s = e θ1s e θ1x x H dx θ 1 (By change of variables;u = s x = sh θ 1 s s e θ 1u (1 u/s H du. Then we exchange the infinite sum and integral by usual Fubuni s theorem and obtain e θ 1s s H 1 θ 1 (H 1 1 F 1 (H 1,H;θ 1 s = sh 1 θ 1 (H 1 n+1 (H k( sθ 1 (n+1γ(n+1,θ 1s. (1 n! n= k=1 Thus substituting these expansions ( and (1 in the previous representation of covariance we obtain the result. Proof of Corollary 4.1. Since X is independent it follows that We divide E[Y t,y t+s ] into piece as follows. E[Y t ]E[Y t+s ] = (E[X] e θ 1(t+s. ( E[Y t Y t+s ] [{ }{ +s }] = E Xe ξt + e (ξt ξu dbu H Xe ξ t+s + e (ξ t+s ξ v dbv H [ +s ] = E[X e (ξt+ξt+s ]+E Xe ξ t+s + e (ξ t+s ξ v dbv H [ ] [ +s +E Xe ξ t+s e (ξt ξu dbu H +E e (ξt ξ u +ξ t+s ξ v dbu H dbh v ]. 17
18 The first term is calculated as E[X e (ξt+ξ t+s ] = E[X ]E[e (ξ t+s ξ t+ξ t ] = E[X ]E[e ξs ]E[e ξt ] = E[X ]e θ 1s θ t. (3 From Theorem 4.1, the second and the fourth terms are. We only need the last term, which is calculated as [ +s ] Cov(Y t,y t+s E e (ξt ξu dbu H e (ξ t+s ξ v dbv H t [ +s ] E e (ξt ξu dbu H e (ξ t+s ξ v dbv H [ ] E e (ξt ξu dbu H e (ξ t+s ξ v dbv H = Cov(Y t,y t+s { e θ1t Cov(Y,Y t+s e θ1(t+s Cov(Y,Y t } e θ 1s E [ e ξt Y Y t ] e θ 1 s E[e ξt Y Y t ], (4 where Y t is Y t with initial value. By adding up (, (3 and (4, we have Cov(Y t,y t+s = Cov(Y t,y t+s e θ1t Cov(Y,Y t+s ( +e θ 1s E[X ]e θ1t (E[X] e θ1t +e θ1t Cov(Y,Y t E[e ξt Y Y t ] E[e ξt Y Yt ]. Hence via Theorem 4.1 the conclusion follows. Acknowledgment: The authors are grateful to Prof. Makoto Maejima for very helpful comments. We give deep thanks to Prof. Claudia Klüppelberg for careful reading our paper and for useful comments. Prof. Alexander Lindner suggested the topic of Section 5. We also thankful to Prof. Jean Jacod for pointing out several theoretical mistakes and for giving us nice suggestions. We shall never forget their kindness. References J. Beran. Statistics for Long-Memory Processes (Monographs on Statistics and Applied Probability. Chapman & Hall, USA, R.M. Blumenthal and R.K. Getoor. Sample functions of stochastic processes with stationary independent increments. J. Math. and Mech., 1: , J. Bretagnolle. p-variation de fonctons aléatoires. Séminaire de Probabilités VI;Lect. Notes in Math., 58:51 71,
19 P. Carmona, F. Petit, and M. Yor. Exponential functionals of Lévy processes. In O.E. Barndorff- Nielsen, T. Mikosch, and S.I. Resnick, editors, Lévy Processes: Theory and Applications, pages Birkhäuser, Boston, 1. P. Cheridito, H. Kawaguchi, and M. Maejima. Fractional Ornstein-Uhlenbeck processes. Electron. J. Probab., 8:1 14, 3. K.L. Chung. A Course in Probability Theory. Academic press, San Diego, 3 edition, 1. R. Doney and R.A. Maller. Stability and attraction to normality for Lévy processes at zero and at infinity. J. Theoret. Probab., 15:751 79,. P. Doukhan, G. Oppenheim, and M.S. Taqqu. Long-Range Dependence: Theory and Applications. Birkhäuser, Boston, 3. R.M. Dudley and R. Norvaiša. An introduction to p-variation and young integrals: with emphasis on sample functions of stochastic process. Lecture Notes P. Embrechts and M. Maejima. Selfsimilar Processes. Princeton Univ. Press, New Jersey,. K. Endo and M. Matsui. The stationarity of multidimensional generalized ornstein-uhlenbeck processes. Statist. Probab. Lett., 8. to appear. K.B. Erickson and R.A. Maller. Generalised Ornstein-Uhlenbeck processes and the convergence of Lévy integrals. Lecture Note in Math., 1857:7 94, 5. Springer. B. Fristedt and S.J. Taylor. Strong variation for the sample functions of a stable process. Duke Math. J., 4:59 78, I.S. Gradshteyn and I.M. Ryzhik. Table of Integrals, Series, and Products. Academic Press Inc., San Diego, 6 edition,. C. Klüppelberg and R. Kostadinova. Integrated insurance risk models with exponential Lévy investment. Insurance: Math & Economics, 4(:56 577, 8. H. Kondo, M. Maejima, and K. Sato. Some properties of exponential integrals of Lévy processes and examples. Electron. Comm. Probab., 11:91 33, 6. D. Lépingle. La variation d ordre p des semi-martingales. Z. Wahrscheinlichkeitstheorie Verw. Geb., 36:95 316, S.J. Lin. Stochastic analysis of fractional Brownian motions. Stochastics and Stochastics Reports, 55:11 14, A. Lindner and R.A. Maller. Lévy integrals and the stationarity of generalised Ornstein-Uhlenbeck processes. Stochastic Processes. Appl., 115:171 17, 5. T. Mikosch and R. Norvai sa. Stochastic integral equations without probability. Bernoulli, 6: ,. I. Monroe. On the γ-variation of processes with stationary independent increments. Ann. Math. Statist., 43:113 1,
20 V. Pipiras and M.S. Taqqu. Integration questions related to fractional Brownian motion. Probab. Theory Related Fields, 118:51 91,. P.E. Protter. Stochastic integration and differenctial equations. Springer, Berlin, edition, 4. G. Samorodnitsky and M. S. Taqqu. Stable Non-Gaussion Random Processes: Stochastic Models with Infinite Variance. Stochastic modeling. Chapman & Hall, London, K. Sato. Lévy Processes and Infinitely Divisible Distributions. Cambridge University Press, Cambridge, U.K., L.C. Young. An inequality of the Hölder type, connected with Stieltjes integration. Acta Math., 67:51 8, 1936.
Multivariate Generalized Ornstein-Uhlenbeck Processes
Multivariate Generalized Ornstein-Uhlenbeck Processes Anita Behme TU München Alexander Lindner TU Braunschweig 7th International Conference on Lévy Processes: Theory and Applications Wroclaw, July 15 19,
More information16. M. Maejima, Some sojourn time problems for strongly dependent Gaussian processes, Z. Wahrscheinlichkeitstheorie verw. Gebiete 57 (1981), 1 14.
Publications 1. M. Maejima, Some limit theorems for renewal processes with non-identically distributed random variables, Keio Engrg. Rep. 24 (1971), 67 83. 2. M. Maejima, A generalization of Blackwell
More informationOn pathwise stochastic integration
On pathwise stochastic integration Rafa l Marcin Lochowski Afican Institute for Mathematical Sciences, Warsaw School of Economics UWC seminar Rafa l Marcin Lochowski (AIMS, WSE) On pathwise stochastic
More informationGARCH processes continuous counterparts (Part 2)
GARCH processes continuous counterparts (Part 2) Alexander Lindner Centre of Mathematical Sciences Technical University of Munich D 85747 Garching Germany lindner@ma.tum.de http://www-m1.ma.tum.de/m4/pers/lindner/
More informationExponential functionals of Lévy processes
Exponential functionals of Lévy processes Víctor Rivero Centro de Investigación en Matemáticas, México. 1/ 28 Outline of the talk Introduction Exponential functionals of spectrally positive Lévy processes
More informationThe multidimensional Ito Integral and the multidimensional Ito Formula. Eric Mu ller June 1, 2015 Seminar on Stochastic Geometry and its applications
The multidimensional Ito Integral and the multidimensional Ito Formula Eric Mu ller June 1, 215 Seminar on Stochastic Geometry and its applications page 2 Seminar on Stochastic Geometry and its applications
More informationJoint Parameter Estimation of the Ornstein-Uhlenbeck SDE driven by Fractional Brownian Motion
Joint Parameter Estimation of the Ornstein-Uhlenbeck SDE driven by Fractional Brownian Motion Luis Barboza October 23, 2012 Department of Statistics, Purdue University () Probability Seminar 1 / 59 Introduction
More informationPower Variation and Time Change
Power Variation and Time Change Ole E. Barndorff-Nielsen Department of Mathematical Sciences, University of Aarhus, Ny Munkegade, DK-8 Aarhus C, Denmark oebn@imf.au.dk Neil Shephard Nuffield College, University
More informationTopics in fractional Brownian motion
Topics in fractional Brownian motion Esko Valkeila Spring School, Jena 25.3. 2011 We plan to discuss the following items during these lectures: Fractional Brownian motion and its properties. Topics in
More informationI forgot to mention last time: in the Ito formula for two standard processes, putting
I forgot to mention last time: in the Ito formula for two standard processes, putting dx t = a t dt + b t db t dy t = α t dt + β t db t, and taking f(x, y = xy, one has f x = y, f y = x, and f xx = f yy
More informationOn the martingales obtained by an extension due to Saisho, Tanemura and Yor of Pitman s theorem
On the martingales obtained by an extension due to Saisho, Tanemura and Yor of Pitman s theorem Koichiro TAKAOKA Dept of Applied Physics, Tokyo Institute of Technology Abstract M Yor constructed a family
More informationRegular Variation and Extreme Events for Stochastic Processes
1 Regular Variation and Extreme Events for Stochastic Processes FILIP LINDSKOG Royal Institute of Technology, Stockholm 2005 based on joint work with Henrik Hult www.math.kth.se/ lindskog 2 Extremes for
More informationPROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS
PROBABILITY: LIMIT THEOREMS II, SPRING 218. HOMEWORK PROBLEMS PROF. YURI BAKHTIN Instructions. You are allowed to work on solutions in groups, but you are required to write up solutions on your own. Please
More informationBernardo D Auria Stochastic Processes /10. Notes. Abril 13 th, 2010
1 Stochastic Calculus Notes Abril 13 th, 1 As we have seen in previous lessons, the stochastic integral with respect to the Brownian motion shows a behavior different from the classical Riemann-Stieltjes
More informationAdditive functionals of infinite-variance moving averages. Wei Biao Wu The University of Chicago TECHNICAL REPORT NO. 535
Additive functionals of infinite-variance moving averages Wei Biao Wu The University of Chicago TECHNICAL REPORT NO. 535 Departments of Statistics The University of Chicago Chicago, Illinois 60637 June
More informationBernardo D Auria Stochastic Processes /12. Notes. March 29 th, 2012
1 Stochastic Calculus Notes March 9 th, 1 In 19, Bachelier proposed for the Paris stock exchange a model for the fluctuations affecting the price X(t) of an asset that was given by the Brownian motion.
More informationA Class of Fractional Stochastic Differential Equations
Vietnam Journal of Mathematics 36:38) 71 79 Vietnam Journal of MATHEMATICS VAST 8 A Class of Fractional Stochastic Differential Equations Nguyen Tien Dung Department of Mathematics, Vietnam National University,
More informationOn Distributions Associated with the Generalized Lévy s Stochastic Area Formula
On Distributions Associated with the Generalized Lévy s Stochastic Area Formula Raouf Ghomrasni Abstract A closed-form ression is obtained for the conditional probability distribution of t R s ds given
More informationPathwise volatility in a long-memory pricing model: estimation and asymptotic behavior
Pathwise volatility in a long-memory pricing model: estimation and asymptotic behavior Ehsan Azmoodeh University of Vaasa Finland 7th General AMaMeF and Swissquote Conference September 7 1, 215 Outline
More informationLARGE DEVIATION PROBABILITIES FOR SUMS OF HEAVY-TAILED DEPENDENT RANDOM VECTORS*
LARGE EVIATION PROBABILITIES FOR SUMS OF HEAVY-TAILE EPENENT RANOM VECTORS* Adam Jakubowski Alexander V. Nagaev Alexander Zaigraev Nicholas Copernicus University Faculty of Mathematics and Computer Science
More informationOrnstein-Uhlenbeck Processes and Extensions
Ornstein-Uhlenbeck Processes and Extensions Ross A. Maller 1, Gernot Müller 2, and Alex Szimayer 3 1 The Australian National University Ross.Maller@anu.edu.au 2 Munich University of Technology mueller@ma.tum.de
More informationPROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS
PROBABILITY: LIMIT THEOREMS II, SPRING 15. HOMEWORK PROBLEMS PROF. YURI BAKHTIN Instructions. You are allowed to work on solutions in groups, but you are required to write up solutions on your own. Please
More informationGeneralized fractional Lévy processes 1
Generalized fractional Lévy processes 1 Claudia Klüppelberg Center for Mathematical Sciences and Institute for Advanced Study Technische Universität München Sandbjerg, January 2010 1 1 joint work with
More information2 Makoto Maejima (ii) Non-Gaussian -stable, 0 << 2, if it is not a delta measure, b() does not varnish for any a>0, b() a = b(a 1= )e ic for some c 2
This is page 1 Printer: Opaque this Limit Theorems for Innite Variance Sequences Makoto Maejima ABSTRACT This article discusses limit theorems for some stationary sequences not having nite variances. 1
More informationFiltrations, Markov Processes and Martingales. Lectures on Lévy Processes and Stochastic Calculus, Braunschweig, Lecture 3: The Lévy-Itô Decomposition
Filtrations, Markov Processes and Martingales Lectures on Lévy Processes and Stochastic Calculus, Braunschweig, Lecture 3: The Lévy-Itô Decomposition David pplebaum Probability and Statistics Department,
More informationMulti-Factor Lévy Models I: Symmetric alpha-stable (SαS) Lévy Processes
Multi-Factor Lévy Models I: Symmetric alpha-stable (SαS) Lévy Processes Anatoliy Swishchuk Department of Mathematics and Statistics University of Calgary Calgary, Alberta, Canada Lunch at the Lab Talk
More informationLimit theorems for multipower variation in the presence of jumps
Limit theorems for multipower variation in the presence of jumps Ole E. Barndorff-Nielsen Department of Mathematical Sciences, University of Aarhus, Ny Munkegade, DK-8 Aarhus C, Denmark oebn@imf.au.dk
More informationApplications of Ito s Formula
CHAPTER 4 Applications of Ito s Formula In this chapter, we discuss several basic theorems in stochastic analysis. Their proofs are good examples of applications of Itô s formula. 1. Lévy s martingale
More informationGAUSSIAN PROCESSES; KOLMOGOROV-CHENTSOV THEOREM
GAUSSIAN PROCESSES; KOLMOGOROV-CHENTSOV THEOREM STEVEN P. LALLEY 1. GAUSSIAN PROCESSES: DEFINITIONS AND EXAMPLES Definition 1.1. A standard (one-dimensional) Wiener process (also called Brownian motion)
More informationRepresentations of Gaussian measures that are equivalent to Wiener measure
Representations of Gaussian measures that are equivalent to Wiener measure Patrick Cheridito Departement für Mathematik, ETHZ, 89 Zürich, Switzerland. E-mail: dito@math.ethz.ch Summary. We summarize results
More informationConvergence at first and second order of some approximations of stochastic integrals
Convergence at first and second order of some approximations of stochastic integrals Bérard Bergery Blandine, Vallois Pierre IECN, Nancy-Université, CNRS, INRIA, Boulevard des Aiguillettes B.P. 239 F-5456
More informationMaximum Likelihood Drift Estimation for Gaussian Process with Stationary Increments
Austrian Journal of Statistics April 27, Volume 46, 67 78. AJS http://www.ajs.or.at/ doi:.773/ajs.v46i3-4.672 Maximum Likelihood Drift Estimation for Gaussian Process with Stationary Increments Yuliya
More informationFinite-time Ruin Probability of Renewal Model with Risky Investment and Subexponential Claims
Proceedings of the World Congress on Engineering 29 Vol II WCE 29, July 1-3, 29, London, U.K. Finite-time Ruin Probability of Renewal Model with Risky Investment and Subexponential Claims Tao Jiang Abstract
More informationLAN property for sde s with additive fractional noise and continuous time observation
LAN property for sde s with additive fractional noise and continuous time observation Eulalia Nualart (Universitat Pompeu Fabra, Barcelona) joint work with Samy Tindel (Purdue University) Vlad s 6th birthday,
More informationGaussian Processes. 1. Basic Notions
Gaussian Processes 1. Basic Notions Let T be a set, and X : {X } T a stochastic process, defined on a suitable probability space (Ω P), that is indexed by T. Definition 1.1. We say that X is a Gaussian
More informationStationary distributions of non Gaussian Ornstein Uhlenbeck processes for beam halos
N Cufaro Petroni: CYCLOTRONS 2007 Giardini Naxos, 1 5 October, 2007 1 Stationary distributions of non Gaussian Ornstein Uhlenbeck processes for beam halos CYCLOTRONS 2007 Giardini Naxos, 1 5 October Nicola
More informationBrownian Motion. 1 Definition Brownian Motion Wiener measure... 3
Brownian Motion Contents 1 Definition 2 1.1 Brownian Motion................................. 2 1.2 Wiener measure.................................. 3 2 Construction 4 2.1 Gaussian process.................................
More informationGeneralized Gaussian Bridges of Prediction-Invertible Processes
Generalized Gaussian Bridges of Prediction-Invertible Processes Tommi Sottinen 1 and Adil Yazigi University of Vaasa, Finland Modern Stochastics: Theory and Applications III September 1, 212, Kyiv, Ukraine
More informationLecture 21 Representations of Martingales
Lecture 21: Representations of Martingales 1 of 11 Course: Theory of Probability II Term: Spring 215 Instructor: Gordan Zitkovic Lecture 21 Representations of Martingales Right-continuous inverses Let
More informationBernstein-gamma functions and exponential functionals of Lévy processes
Bernstein-gamma functions and exponential functionals of Lévy processes M. Savov 1 joint work with P. Patie 2 FCPNLO 216, Bilbao November 216 1 Marie Sklodowska Curie Individual Fellowship at IMI, BAS,
More informationOn the quantiles of the Brownian motion and their hitting times.
On the quantiles of the Brownian motion and their hitting times. Angelos Dassios London School of Economics May 23 Abstract The distribution of the α-quantile of a Brownian motion on an interval [, t]
More informationStruktura zależności dla rozwiązań ułamkowych równań z szumem α-stabilnym Marcin Magdziarz
Politechnika Wrocławska Instytut Matematyki i Informatyki Struktura zależności dla rozwiązań ułamkowych równań z szumem α-stabilnym Marcin Magdziarz Rozprawa doktorska Promotor: Prof. dr hab. Aleksander
More informationMULTIDIMENSIONAL WICK-ITÔ FORMULA FOR GAUSSIAN PROCESSES
MULTIDIMENSIONAL WICK-ITÔ FORMULA FOR GAUSSIAN PROCESSES D. NUALART Department of Mathematics, University of Kansas Lawrence, KS 6645, USA E-mail: nualart@math.ku.edu S. ORTIZ-LATORRE Departament de Probabilitat,
More informationContinuous Time Approximations to GARCH(1, 1)-Family Models and Their Limiting Properties
Communications for Statistical Applications and Methods 214, Vol. 21, No. 4, 327 334 DOI: http://dx.doi.org/1.5351/csam.214.21.4.327 Print ISSN 2287-7843 / Online ISSN 2383-4757 Continuous Time Approximations
More informationHarnack Inequalities and Applications for Stochastic Equations
p. 1/32 Harnack Inequalities and Applications for Stochastic Equations PhD Thesis Defense Shun-Xiang Ouyang Under the Supervision of Prof. Michael Röckner & Prof. Feng-Yu Wang March 6, 29 p. 2/32 Outline
More informationA Short Introduction to Diffusion Processes and Ito Calculus
A Short Introduction to Diffusion Processes and Ito Calculus Cédric Archambeau University College, London Center for Computational Statistics and Machine Learning c.archambeau@cs.ucl.ac.uk January 24,
More informationOctavio Arizmendi, Ole E. Barndorff-Nielsen and Víctor Pérez-Abreu
ON FREE AND CLASSICAL TYPE G DISTRIBUTIONS Octavio Arizmendi, Ole E. Barndorff-Nielsen and Víctor Pérez-Abreu Comunicación del CIMAT No I-9-4/3-4-29 (PE /CIMAT) On Free and Classical Type G Distributions
More informationJump-type Levy Processes
Jump-type Levy Processes Ernst Eberlein Handbook of Financial Time Series Outline Table of contents Probabilistic Structure of Levy Processes Levy process Levy-Ito decomposition Jump part Probabilistic
More informationLecture 22 Girsanov s Theorem
Lecture 22: Girsanov s Theorem of 8 Course: Theory of Probability II Term: Spring 25 Instructor: Gordan Zitkovic Lecture 22 Girsanov s Theorem An example Consider a finite Gaussian random walk X n = n
More informationAsymptotic inference for a nonstationary double ar(1) model
Asymptotic inference for a nonstationary double ar() model By SHIQING LING and DONG LI Department of Mathematics, Hong Kong University of Science and Technology, Hong Kong maling@ust.hk malidong@ust.hk
More informationOn Optimal Stopping Problems with Power Function of Lévy Processes
On Optimal Stopping Problems with Power Function of Lévy Processes Budhi Arta Surya Department of Mathematics University of Utrecht 31 August 2006 This talk is based on the joint paper with A.E. Kyprianou:
More informationThe Codimension of the Zeros of a Stable Process in Random Scenery
The Codimension of the Zeros of a Stable Process in Random Scenery Davar Khoshnevisan The University of Utah, Department of Mathematics Salt Lake City, UT 84105 0090, U.S.A. davar@math.utah.edu http://www.math.utah.edu/~davar
More informationOn Integration-by-parts and the Itô Formula for Backwards Itô Integral
11 On Integration-by-parts and the Itô Formula for... On Integration-by-parts and the Itô Formula for Backwards Itô Integral Jayrold P. Arcede a, Emmanuel A. Cabral b a Caraga State University, Ampayon,
More informationSupermodular ordering of Poisson arrays
Supermodular ordering of Poisson arrays Bünyamin Kızıldemir Nicolas Privault Division of Mathematical Sciences School of Physical and Mathematical Sciences Nanyang Technological University 637371 Singapore
More informationBranching Processes II: Convergence of critical branching to Feller s CSB
Chapter 4 Branching Processes II: Convergence of critical branching to Feller s CSB Figure 4.1: Feller 4.1 Birth and Death Processes 4.1.1 Linear birth and death processes Branching processes can be studied
More information1. Stochastic Processes and filtrations
1. Stochastic Processes and 1. Stoch. pr., A stochastic process (X t ) t T is a collection of random variables on (Ω, F) with values in a measurable space (S, S), i.e., for all t, In our case X t : Ω S
More informationEULER MARUYAMA APPROXIMATION FOR SDES WITH JUMPS AND NON-LIPSCHITZ COEFFICIENTS
Qiao, H. Osaka J. Math. 51 (14), 47 66 EULER MARUYAMA APPROXIMATION FOR SDES WITH JUMPS AND NON-LIPSCHITZ COEFFICIENTS HUIJIE QIAO (Received May 6, 11, revised May 1, 1) Abstract In this paper we show
More informationFRACTIONAL BROWNIAN MOTION WITH H < 1/2 AS A LIMIT OF SCHEDULED TRAFFIC
Applied Probability Trust ( April 20) FRACTIONAL BROWNIAN MOTION WITH H < /2 AS A LIMIT OF SCHEDULED TRAFFIC VICTOR F. ARAMAN, American University of Beirut PETER W. GLYNN, Stanford University Keywords:
More informationGARCH processes probabilistic properties (Part 1)
GARCH processes probabilistic properties (Part 1) Alexander Lindner Centre of Mathematical Sciences Technical University of Munich D 85747 Garching Germany lindner@ma.tum.de http://www-m1.ma.tum.de/m4/pers/lindner/
More informationSolution for Problem 7.1. We argue by contradiction. If the limit were not infinite, then since τ M (ω) is nondecreasing we would have
362 Problem Hints and Solutions sup g n (ω, t) g(ω, t) sup g(ω, s) g(ω, t) µ n (ω). t T s,t: s t 1/n By the uniform continuity of t g(ω, t) on [, T], one has for each ω that µ n (ω) as n. Two applications
More informationStochastic Calculus and Black-Scholes Theory MTH772P Exercises Sheet 1
Stochastic Calculus and Black-Scholes Theory MTH772P Exercises Sheet. For ξ, ξ 2, i.i.d. with P(ξ i = ± = /2 define the discrete-time random walk W =, W n = ξ +... + ξ n. (i Formulate and prove the property
More informationStochastic Calculus (Lecture #3)
Stochastic Calculus (Lecture #3) Siegfried Hörmann Université libre de Bruxelles (ULB) Spring 2014 Outline of the course 1. Stochastic processes in continuous time. 2. Brownian motion. 3. Itô integral:
More informationLecture 12. F o s, (1.1) F t := s>t
Lecture 12 1 Brownian motion: the Markov property Let C := C(0, ), R) be the space of continuous functions mapping from 0, ) to R, in which a Brownian motion (B t ) t 0 almost surely takes its value. Let
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 7 9/25/2013
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.65/15.070J Fall 013 Lecture 7 9/5/013 The Reflection Principle. The Distribution of the Maximum. Brownian motion with drift Content. 1. Quick intro to stopping times.
More informationLower Tail Probabilities and Related Problems
Lower Tail Probabilities and Related Problems Qi-Man Shao National University of Singapore and University of Oregon qmshao@darkwing.uoregon.edu . Lower Tail Probabilities Let {X t, t T } be a real valued
More informationIndependence of some multiple Poisson stochastic integrals with variable-sign kernels
Independence of some multiple Poisson stochastic integrals with variable-sign kernels Nicolas Privault Division of Mathematical Sciences School of Physical and Mathematical Sciences Nanyang Technological
More informationPoisson Cluster process as a model for teletraffic arrivals and its extremes
Poisson Cluster process as a model for teletraffic arrivals and its extremes Barbara González-Arévalo, University of Louisiana Thomas Mikosch, University of Copenhagen Gennady Samorodnitsky, Cornell University
More informationComment on Weak Convergence to a Matrix Stochastic Integral with Stable Processes
Comment on Weak Convergence to a Matrix Stochastic Integral with Stable Processes Vygantas Paulauskas Department of Mathematics and Informatics, Vilnius University Naugarduko 24, 03225 Vilnius, Lithuania
More informationIntegral representations in models with long memory
Integral representations in models with long memory Georgiy Shevchenko, Yuliya Mishura, Esko Valkeila, Lauri Viitasaari, Taras Shalaiko Taras Shevchenko National University of Kyiv 29 September 215, Ulm
More informationA NEW PROOF OF THE WIENER HOPF FACTORIZATION VIA BASU S THEOREM
J. Appl. Prob. 49, 876 882 (2012 Printed in England Applied Probability Trust 2012 A NEW PROOF OF THE WIENER HOPF FACTORIZATION VIA BASU S THEOREM BRIAN FRALIX and COLIN GALLAGHER, Clemson University Abstract
More informationLIMIT THEOREMS FOR SHIFT SELFSIMILAR ADDITIVE RANDOM SEQUENCES
Watanabe, T. Osaka J. Math. 39 22, 561 63 LIMIT THEOEMS FO SHIFT SELFSIMILA ADDITIVE ANDOM SEQUENCES TOSHIO WATANABE eceived November 15, 2 1. Introduction We introduce shift selfsimilar random sequences,
More informationHSC Research Report. The Lamperti transformation for self-similar processes HSC/97/02. Krzysztof Burnecki* Makoto Maejima** Aleksander Weron*, ***
HSC Research Report HSC/97/02 The Lamperti transformation for self-similar processes Krzysztof Burnecki* Makoto Maeima** Aleksander Weron*, *** * Hugo Steinhaus Center, Wrocław University of Technology,
More information(2m)-TH MEAN BEHAVIOR OF SOLUTIONS OF STOCHASTIC DIFFERENTIAL EQUATIONS UNDER PARAMETRIC PERTURBATIONS
(2m)-TH MEAN BEHAVIOR OF SOLUTIONS OF STOCHASTIC DIFFERENTIAL EQUATIONS UNDER PARAMETRIC PERTURBATIONS Svetlana Janković and Miljana Jovanović Faculty of Science, Department of Mathematics, University
More informationMOMENT CONVERGENCE RATES OF LIL FOR NEGATIVELY ASSOCIATED SEQUENCES
J. Korean Math. Soc. 47 1, No., pp. 63 75 DOI 1.4134/JKMS.1.47..63 MOMENT CONVERGENCE RATES OF LIL FOR NEGATIVELY ASSOCIATED SEQUENCES Ke-Ang Fu Li-Hua Hu Abstract. Let X n ; n 1 be a strictly stationary
More informationSAMPLE CHAPTERS UNESCO EOLSS STATIONARY PROCESSES. E ξ t (1) K.Grill Institut für Statistik und Wahrscheinlichkeitstheorie, TU Wien, Austria
STATIONARY PROCESSES K.Grill Institut für Statistik und Wahrscheinlichkeitstheorie, TU Wien, Austria Keywords: Stationary process, weak sense stationary process, strict sense stationary process, correlation
More informationfor all f satisfying E[ f(x) ] <.
. Let (Ω, F, P ) be a probability space and D be a sub-σ-algebra of F. An (H, H)-valued random variable X is independent of D if and only if P ({X Γ} D) = P {X Γ}P (D) for all Γ H and D D. Prove that if
More informationLeast Squares Estimators for Stochastic Differential Equations Driven by Small Lévy Noises
Least Squares Estimators for Stochastic Differential Equations Driven by Small Lévy Noises Hongwei Long* Department of Mathematical Sciences, Florida Atlantic University, Boca Raton Florida 33431-991,
More informationFunctional Limit theorems for the quadratic variation of a continuous time random walk and for certain stochastic integrals
Functional Limit theorems for the quadratic variation of a continuous time random walk and for certain stochastic integrals Noèlia Viles Cuadros BCAM- Basque Center of Applied Mathematics with Prof. Enrico
More informationA Note on the Central Limit Theorem for a Class of Linear Systems 1
A Note on the Central Limit Theorem for a Class of Linear Systems 1 Contents Yukio Nagahata Department of Mathematics, Graduate School of Engineering Science Osaka University, Toyonaka 560-8531, Japan.
More information9 Brownian Motion: Construction
9 Brownian Motion: Construction 9.1 Definition and Heuristics The central limit theorem states that the standard Gaussian distribution arises as the weak limit of the rescaled partial sums S n / p n of
More informationShort-time expansions for close-to-the-money options under a Lévy jump model with stochastic volatility
Short-time expansions for close-to-the-money options under a Lévy jump model with stochastic volatility José Enrique Figueroa-López 1 1 Department of Statistics Purdue University Statistics, Jump Processes,
More informationThe Uniform Integrability of Martingales. On a Question by Alexander Cherny
The Uniform Integrability of Martingales. On a Question by Alexander Cherny Johannes Ruf Department of Mathematics University College London May 1, 2015 Abstract Let X be a progressively measurable, almost
More informationNumerical Methods with Lévy Processes
Numerical Methods with Lévy Processes 1 Objective: i) Find models of asset returns, etc ii) Get numbers out of them. Why? VaR and risk management Valuing and hedging derivatives Why not? Usual assumption:
More informationLecture 4: Introduction to stochastic processes and stochastic calculus
Lecture 4: Introduction to stochastic processes and stochastic calculus Cédric Archambeau Centre for Computational Statistics and Machine Learning Department of Computer Science University College London
More informationStrictly Stationary Solutions of Autoregressive Moving Average Equations
Strictly Stationary Solutions of Autoregressive Moving Average Equations Peter J. Brockwell Alexander Lindner Abstract Necessary and sufficient conditions for the existence of a strictly stationary solution
More informationUniversity Of Calgary Department of Mathematics and Statistics
University Of Calgary Department of Mathematics and Statistics Hawkes Seminar May 23, 2018 1 / 46 Some Problems in Insurance and Reinsurance Mohamed Badaoui Department of Electrical Engineering National
More informationSTOCHASTIC GEOMETRY BIOIMAGING
CENTRE FOR STOCHASTIC GEOMETRY AND ADVANCED BIOIMAGING 2018 www.csgb.dk RESEARCH REPORT Anders Rønn-Nielsen and Eva B. Vedel Jensen Central limit theorem for mean and variogram estimators in Lévy based
More informationJae Gil Choi and Young Seo Park
Kangweon-Kyungki Math. Jour. 11 (23), No. 1, pp. 17 3 TRANSLATION THEOREM ON FUNCTION SPACE Jae Gil Choi and Young Seo Park Abstract. In this paper, we use a generalized Brownian motion process to define
More informationSIMULATION OF WEAKLY SELF-SIMILAR STATIONARY INCREMENT Sub ϕ (Ω)-PROCESSES: A SERIES EXPANSION APPROACH
SIMULATION OF WEAKLY SELF-SIMILAR STATIONARY INCREMENT Sub ϕ Ω-PROCESSES: A SERIES EXPANSION APPROACH YURIY KOZACHENKO, TOMMI SOTTINEN, AND OLGA VASYLYK Abstract. We consider simulation of Sub ϕω-processes
More informationItô s formula. Samy Tindel. Purdue University. Probability Theory 2 - MA 539
Itô s formula Samy Tindel Purdue University Probability Theory 2 - MA 539 Mostly taken from Brownian Motion and Stochastic Calculus by I. Karatzas and S. Shreve Samy T. Itô s formula Probability Theory
More informationOn Reflecting Brownian Motion with Drift
Proc. Symp. Stoch. Syst. Osaka, 25), ISCIE Kyoto, 26, 1-5) On Reflecting Brownian Motion with Drift Goran Peskir This version: 12 June 26 First version: 1 September 25 Research Report No. 3, 25, Probability
More informationPathwise uniqueness for stochastic differential equations driven by pure jump processes
Pathwise uniqueness for stochastic differential equations driven by pure jump processes arxiv:73.995v [math.pr] 9 Mar 7 Jiayu Zheng and Jie Xiong Abstract Based on the weak existence and weak uniqueness,
More informationPotentials of stable processes
Potentials of stable processes A. E. Kyprianou A. R. Watson 5th December 23 arxiv:32.222v [math.pr] 4 Dec 23 Abstract. For a stable process, we give an explicit formula for the potential measure of the
More informationA REFINED FACTORIZATION OF THE EXPONENTIAL LAW P. PATIE
A REFINED FACTORIZATION OF THE EXPONENTIAL LAW P. PATIE Abstract. Let ξ be a (possibly killed) subordinator with Laplace exponent φ and denote by I φ e ξs ds, the so-called exponential functional. Consider
More informationMoment Properties of Distributions Used in Stochastic Financial Models
Moment Properties of Distributions Used in Stochastic Financial Models Jordan Stoyanov Newcastle University (UK) & University of Ljubljana (Slovenia) e-mail: stoyanovj@gmail.com ETH Zürich, Department
More informationPathwise Construction of Stochastic Integrals
Pathwise Construction of Stochastic Integrals Marcel Nutz First version: August 14, 211. This version: June 12, 212. Abstract We propose a method to construct the stochastic integral simultaneously under
More informationKrzysztof Burdzy University of Washington. = X(Y (t)), t 0}
VARIATION OF ITERATED BROWNIAN MOTION Krzysztof Burdzy University of Washington 1. Introduction and main results. Suppose that X 1, X 2 and Y are independent standard Brownian motions starting from 0 and
More informationA NOTE ON STOCHASTIC INTEGRALS AS L 2 -CURVES
A NOTE ON STOCHASTIC INTEGRALS AS L 2 -CURVES STEFAN TAPPE Abstract. In a work of van Gaans (25a) stochastic integrals are regarded as L 2 -curves. In Filipović and Tappe (28) we have shown the connection
More informationObstacle problems for nonlocal operators
Obstacle problems for nonlocal operators Camelia Pop School of Mathematics, University of Minnesota Fractional PDEs: Theory, Algorithms and Applications ICERM June 19, 2018 Outline Motivation Optimal regularity
More informationFrom Fractional Brownian Motion to Multifractional Brownian Motion
From Fractional Brownian Motion to Multifractional Brownian Motion Antoine Ayache USTL (Lille) Antoine.Ayache@math.univ-lille1.fr Cassino December 2010 A.Ayache (USTL) From FBM to MBM Cassino December
More information