On the exit from a finite interval for the risk processes with stochastic premiums

We consider the almost semicontinuous step-process ξ(t). The conditional characteristic functions of the jumps of ξ(t) have the form E [eiαξk /ξk > 0] = c(c − iα)−1. For such processes, the boundary functionals related to the exit from a finite interval are investigated.

Saved in:
Bibliographic Details
Date:2005
Main Authors: Gusak, D.V., Karnaukh, E.V.
Format: Article
Language:English
Published: Інститут математики НАН України 2005
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/4427
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:On the exit from a finite interval for the risk processes with stochastic premiums / D.V. Gusak, E.V. Karnaukh // Theory of Stochastic Processes. — 2005. — Т. 11 (27), № 3-4. — С. 71–81. — Бібліогр.: 11 назв.— англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859778949737873408
author Gusak, D.V.
Karnaukh, E.V.
author_facet Gusak, D.V.
Karnaukh, E.V.
citation_txt On the exit from a finite interval for the risk processes with stochastic premiums / D.V. Gusak, E.V. Karnaukh // Theory of Stochastic Processes. — 2005. — Т. 11 (27), № 3-4. — С. 71–81. — Бібліогр.: 11 назв.— англ.
collection DSpace DC
description We consider the almost semicontinuous step-process ξ(t). The conditional characteristic functions of the jumps of ξ(t) have the form E [eiαξk /ξk > 0] = c(c − iα)−1. For such processes, the boundary functionals related to the exit from a finite interval are investigated.
first_indexed 2025-12-02T09:16:04Z
format Article
fulltext Theory of Stochastic Processes Vol. 11 (27), no. 3–4, 2005, pp. 71–81 UDC 519.21 D. V. GUSAK AND E. V. KARNAUKH ON THE EXIT FROM A FINITE INTERVAL FOR THE RISK PROCESSES WITH STOCHASTIC PREMIUMS We consider the almost semicontinuous step-process ξ(t). The conditional character- istic functions of the jumps of ξ(t) have the form E eiαξk /ξk > 0 = c(c − iα)−1. For such processes, the boundary functionals related to the exit from a finite interval are investigated. The problems on the exit from a finite interval for the process ξ(t) (t ≥ 0, ξ(0) = 0) with stationary independent increments were considered by many authors (see, for example [1, ch. IV, § 2]). In [1], the joint distributions of extrema and the distributions of values of the process up to the exit from the interval were expressed in terms of rather complicate series of the ”convolutions” of Γ±(s, x, y) = E [ e−sτ±(±x), γ±(±x) ≤ y ] , where τ±(±x) = inf {t > 0 : ±ξ(t) > x} , γ±(±x) = ±ξ(τ±(±x)) ∓ x, x > 0. Simpler relations for the Wiener processes are established in [1, p. 463] and in [2, § 27]. In [3] - [6], the mentioned problems were investigated for semicontinuous processes ξ(t) (ξ(t) have jumps of one sign). For these processes, the distribution density of ξ(t) up to the exit from the interval was represented [7], [8] in terms of the resolvent functions Rs(x) (introduced by V.S. Korolyuk in [3]). We consider the compound Poisson process ξ(t) = ∑ k≤ν(t) ξk, where ν(t) is the Poisson process with rate λ > 0. The distributions of ξk satisfy the next condition (F (x) is a cumulative distribution function) (1) P {ξk < x} = qF (x)I {x ≤ 0} + (1 − pe−cx)I {x > 0} , c > 0, p + q = 1. The process ξ(t) is the almost upper semicontinuous piecewise constant process. We can represent ξ(t) as the claim surplus process ξ(t) = C(t) − S(t) with the stochastic premium function C(t) = ∑ k≤ν1(t) ηk, ηk > 0, E eiαηk = c c − iα , c > 0, and with the process of claims S(t) = ∑ k≤ν2(t) ξ′k, ξ′k > 0. Here, ν1(t), ν2(t) are the independent Poisson processes with rates λ1, λ2 > 0, λ1 + λ2 = λ (for details, see [8] ). 2000 AMS Mathematics Subject Classification. Primary 60G50; Secondary 60K10. Key words and phrases. Almost semicontinuous processes, risk process with stochastic premiums, functionals connected with the exit from an interval. 71 72 D. V. GUSAK AND E. V. KARNAUKH Note that C(t) → 0 and ξ(t) → −S(t) as c → ∞, where −S(t) is a non-increasing process. Let Cc(t) be the process with the cumulant ψc(α) = λc ( c c − iα − 1 ) , λc = ac, a > 0, then ψc(α) −→ c→∞ iαa, consequently Cc(t) −→ c→∞ at, and ξc(t) = Cc(t) − S(t) → ξ0(t) = at−S(t), where the limit process ξ0(t) is the classical upper semicontinuous risk process with the non-stochastic premium function C(t) = at. Let θs be the exponentially distributed random variable (P{θs > t} = e−st; s, t > 0). Then the randomly stopped process ξ(θs) has the characteristic function (ch.f.) ϕ(s, α) = Eeiαξ(θs) = s s − ψ(α) , where (2) ψ(α) = λp(c(c − iα)−1 − 1) + λq(ϕ(α) − 1), ϕ(α) = ∫ 0 −∞ eiαxdF (x). Let us denote the first exit time from the interval (x − T, x), 0 < x < T , T > 0: τ(x, T ) = inf {t > 0 : ξ(t) /∈ (x − T, x)} , and the events A+(x) = {ω : ξ(τ(x, T )) ≥ x} , A−(x) = {ω : ξ(τ(x, T )) ≤ x − T } . Then τ(x, T )=̇ { τ+(x, T ) = τ+(x), ω ∈ A+(x); τ−(x, T ) = τ−(x − T ), ω ∈ A−(x). Overshoots at the moments of the exit from the interval are denoted by the following relations: γ− T (x) = x − T − ξ(τ−(x, T )), γ+ T (x) = ξ(τ+(x, T )) − x. The main task of our paper is the finding of the following moment generating functions (m.g.f.) of the functionals connected with the exit from the interval: Q(T, s, x) = E e−sτ(x,T ), QT (s, x) = E [ e−sτ+(x,T ), A+(x) ] , QT (s, x) = E [ e−sτ−(x,T ), A−(x) ] , V ±(s, α, x, T ) = E [ eiαγ± T (x)−sτ±(x,T ), A±(x) ] , V±(s, α, x, T ) = E [ eiαξ(τ±(x,T ))−sτ±(x,T ), A±(x) ] , V (s, α, x, T ) = E [ eiαξ(θs), τ(x, T ) > θs ] , Let us denote the extrema ξ±(t) = sup 0≤s≤t (inf)ξ(s), ξ± = sup 0≤s<∞ (inf)ξ(s), the joint distribution of {ξ(θs), ξ+(θs), ξ−(θs)}: Hs(T, x, y) = P { ξ(θs) < y, ξ+(θs) < x, ξ−(θs) > x − T } = P {ξ(θs) < y, τ(x, T ) > θs} , ON THE EXIT FROM THE INTERVAL FOR THE RISK PROCESS 73 and P±(s, x) = P { ξ±(θs) < x } , x ≷ 0, p±(s) = P { ξ±(θs) = 0 } , q±(s) = 1 − p±(s); ϕ±(s, α) = ± ∫ ±∞ 0 eiαxdP±(s, x), T±(s, x) = E [ e−sτ±(x), τ±(x) < ∞ ] , x ≷ 0. Lemma 1. For the process ξ(t) with cumulant (2), the main factorization identity is represented by the relations (3) ϕ(s, α) = ϕ+(s, α)ϕ−(s, α), �α = 0; (4) ϕ+(s, α) = p+(s)(c − iα) ρ+(s) − iα , where ρ+(s) = cp+(s) is the positive root of Lundberg’s equation ψ(−ir) = s, s > 0. (5) P { ξ+(θs) > x } = T +(s, x) = q+(s)e−cρ+(s)x, x > 0. If m > 0 : (6) lim s→0 ρ+(s)s−1 = ρ′+(0) = m−1, lim s→0 P−(s, x) = P { ξ− < x } , x < 0. If m < 0 : (7) lim s→0 ρ+(s) = ρ+ > 0; lim s→0 s−1P { ξ−(θs) > x } = E τ−(x), x < 0. If σ2 1 = Dξ(1) < ∞ and m = λ ( pc−1 − qF̃ (0) ) = 0 ( F̃ (0) = ∫ 0 −∞ F (x)dx ) , then (8) lim s→0 ρ+(s)s−1/2 = √ 2 σ1 ; lim s→0 s−1/2P ′ −(s, x) = f0(x), x < 0, f0(x) = k0 ∂ ∂x (∫ ∞ 0 P { ξ̃0(t) < x } dt ) = −k0 ∂ ∂x E τ0(x), x < 0; where k0 = cσ1 (√ 2 )−1 , τ0(x) = inf { t > 0 : ξ̃0(t) < x } , x < 0; ξ̃0(t) is the decreasing process with the spectral measure Π0(dx) = λq (cF (x)dx + dF (x)) , x < 0. Proof. Relations (3) - (7) were proved in [7] - [8]. If m = 0 ( p = cqF̃ (0) ) , then ϕ(s, α) = s(c − iα) s(c − iα) − iαλ(p − qF̃ (α)(c − iα)) , F̃ (α) = ∫ 0 −∞ eiαxF (x)dx. On the basis of the factorization identity (3) as s → 0, we get 1√ s ϕ−(s, α) = √ s p+(s) ρ+(s) − iα s(c − iα) − iαλ ( p − qF̃ (α)(c − iα) ) → f̃0(α), f̃0(α) = cσ1√ 2 1 −λq [( F̃ (α) − F̃ (0) ) c + ϕ(α) − 1 ] = cσ1√ 2 1 −ψ̃0(α) , ψ̃0(α) = ∫ 0 −∞ ( eiαx − 1 ) Π0(dx), Π0(dx) = λq (cF (x)dx + dF (x)) , x < 0. 74 D. V. GUSAK AND E. V. KARNAUKH Let’s denote ϕ0(s, α) = E eiαξ0(θs) = s s − ψ̃0(α) , where ξ̃0(t) is the decreasing process with the cumulant ψ̃0(α). Since cσ1√ 2 ϕ0(s, α)s−1 → f̃0(α) = ∫ 0 −∞ eiαxf0(x)dx, s → 0, we get that f0(x) = k0 ∂ ∂x (∫ ∞ 0 P { ξ̃0(t) < x } dt ) , or −f0(x) = k0 ∂ ∂x ∫ ∞ 0 P {τ0(x) > t} dt = k0 ∂ ∂x E τ0(x), x < 0. Let’s introduce the set of boundary functions on the interval I ⊂ (−∞,∞) L(I) = { G(x) : ∫ I |G(x)|dx < ∞ } and the set of integral transforms R0(I) = { g0(α) : g0(α) = C + ∫ I eiαxG(x)dx } . Let’s denote the projection operations on R0((−∞,∞)) by the following relations:[ g0(α) ] I = ∫ I eiαxG(x)dx, [ g0(α) ]0 I = C + ∫ I eiαxG(x)dx,[ g0(α) ] − = [ g0(α) ] (−∞,0) , [ g0(α) ] + = [ g0(α) ] (0,∞) . The main results of our paper are included in the following two assertions. Theorem 1. For the process ξ(t) with cumulant (2), QT (s, x) has the form (0 < x < T ) (9) QT (s, x) = q+(s)e−ρ+(s)x ∫ 0 x−T eρ+(s)ydP−(s, y)× × [ e−ρ+(s)T ∫ −T −∞ ec(T+y)dP−(s, y) + ∫ 0 −T eρ+(s)ydP−(s, y) ]−1 . Theorem 2. For the process ξ(t) with cumulant (2), the joint distributions of{ τ+(x, T ), γ+ T (x) } and {τ+(x, T ), ξ(τ+(x, T ))} are determined by the relations (10) ⎧⎪⎨⎪⎩ V +(s, α, x, T ) = c c − iα QT (s, x), 0 < x < T, V+(s, α, x, T ) = eiαxV +(s, α, x, T ) = c eiαx c − iα QT (s, x). The ch.f. of ξ(θs) before the exit time from the interval has the form V (s, α, x, T ) = ϕ+(s, α) [ϕ−(s, α) (1 − V+(s, α, x, T ))][x−T,∞) = ϕ+(s, α) [ ϕ−(s, α) ( 1 − c eiαx(c − iα)−1QT (s, x) )] [x−T,∞) , (11) ON THE EXIT FROM THE INTERVAL FOR THE RISK PROCESS 75 the corresponding distribution has the density (x − T < z < x, z �= 0) (12) hs(T, x, z) = ∂ ∂z Hs(T, x, z) = = ( p+(s)P ′ −(s, z) − q+(s)ρ+(s) ∫ 0 z eρ+(s)(y−z)dP−(s, y) ) I {z < 0}+ + ρ+(s)QT (s, x) ∫ 0 z−x eρ+(s)(y−(z−x))dP−(s, y), and the following atomic probability P {ξ(θs) = 0, τ(x, T ) > θs} = P {ξ(θs) = 0} = p−(s)p+(s) = s s + λ . Proof of Theorem 1. From the stochastic relations for τ+(x, T ), γ+ T (x) (ξ = ξ1 has the cumulative distribution function F1(x), ζ is the moment of the first jump of ξ(t)), τ+(x, T )=̇ { ζ, ξ > x, ζ + τ+(x − ξ, T ), x − T < ξ < x, γ+ T (x)=̇ { ξ − x, ξ > x, γ+ T (x − ξ), x − T < ξ < x, we have the following equation for V +(s, α, x) = V +(s, α, x, T ): (13) (s + λ)V +(s, α, x) = λpc c − iα e−cx + λ ∫ x x−T V +(s, α, x − z)dF1(z), 0 < x < T. If α = 0, then, from (13), we obtain the equation for QT (s, x) (14) (s + λ)QT (s, x) = λpe−cx + λ ∫ x x−T QT (s, x − z)dF1(z), 0 < x < T. Since P (A+(x)) = 1 for x < 0, we have the boundary conditions QT (s, x) = { 0, x > T, 1, x < 0. After the replacement Q T (s, x) = 1 − QT (s, x) , relation (14) yields the equation for Q T (s, x) (0 < x < T ) (s + λ)Q T (s, x) = s + λF (x − T ) + λ ∫ T 0 Q T (s, z)F ′ 1(x − z)dz, which, after prolonging for x > 0, has the form (15) (s + λ)Q T (s, x) = sC(x) + λ ∫ ∞ −∞ Q T (s, z)F ′ 1(x − z)dz + C> T (s, x), C(x) = I {x > 0} , C> T (s, x) = CT (s)e−cx, x > 0, (16) CT (s) = λp [ ecT − cQ ∗ s(T ) ] , Q ∗ s(T ) = ∫ T 0 ecxQ T (s, x)dx. Let’s introduce the function Cε(x) = e−εxC(x), x > 0, and consider, instead of (15), the equation for Yε(T, s, x) (ε > 0): (17) (s + λ)Yε(T, s, x) = sCε(x) + λ ∫ ∞ −∞ Yε(T, s, x − z)dF1(z) + C> T (s, x), x > 0. 76 D. V. GUSAK AND E. V. KARNAUKH Denote yε(T, s, α) = ∫ ∞ 0 eiαxYε(T, s, x)dx, C̃ε(α) = ∫ ∞ 0 eiαxCε(x)dx, C̃T (s, α) = ∫ ∞ 0 eiαxC> T (s, x)dx. By performing the integral transformation of (17), we obtain the equation (s − ψ(α))yε(T, s, α) = sC̃ε(α) + C̃T (s, α) − [yε(α)ϕ(α)]− or (18) syε(T, s, α)ϕ−1(s, α) = sC̃ε(α) + C̃T (s, α) − [yε(α)ϕ(α)]− . After using the factorization decomposition (3) and the projection operation [ ]+, relation (18) yields syε(T, s, α)ϕ−1 + (s, α) = [ ϕ−(s, α) ( sC̃ε(α) + C̃T (s, α) )] + or (19) syε(T, s, α) = ϕ+(s, α) [ ϕ−(s, α) ( sC̃ε(α) + C̃T (s, α) )] + . By inverting relation (19), we obtain (20) sYε(T, s, x) = s ∫ x 0 Bε(x − y)dP+(s, y) + ∫ x 0 B(s, x − y, T )dP+(s, y), Bε(x) = ∫ x −∞ e−ε(x−y)dP−(s, y) = ∫ 0 −∞ e−ε(x−y)dP−(s, y) = = e−εxE eεξ−(θs), B(s, x, T ) = CT (s) ∫ x−T −∞ e−c(x−y)dP−(s, y), x > 0. Taking into account that Cε(x) → I {x > 0} as ε → 0, Yε(T, s, x) → Q T (s, x) as ε → 0, 0 < x < T . So Eq. (20) yields sQ T (s, x) = sP+(s, x) + p+(s)B(s, x, T ) + ∫ x +0 B(s, x − z, T )P ′ +(s, z)dz. Taking into account that q+(s)ρ+(s) ∫ x 0 ∫ z−T −∞ e−c(z−y)dP−(s, y)e−ρ+(s)(x−z)dz = =q+(s)ρ+(s) ∫ x−T −∞ e−ρ+(s)x+cydP−(s, y) ∫ x max(0,y+T ) e−cq+(s)zdz =p+(s) [∫ −T −∞ ecy−ρ+(s)xdP−(s, y)+ + ∫ x−T −T eρ+(s)(y+T−x)−cTdP−(s, y) − ∫ x−T −∞ e−c(x−y)dP−(s, y) ] , we have sQ T (s, x) = sP+(s, x) + p+(s)CT (s)e−ρ+(s)x× × [∫ −T −∞ ecydP−(s, y) + ∫ x−T −T e−cT+ρ+(s)(y+T )dP−(s, y) ] . ON THE EXIT FROM THE INTERVAL FOR THE RISK PROCESS 77 From the last equation, we can find CT (s) and Q ∗ s(T ) and then get (9). Let’s note that QT (s, x) → P+(s, x) as T → ∞ and QT (s, x) → 0 as c → ∞. If we consider, instead of ξ(t), the process ξc(t) = Cc(t) − S(t), then relation (9) yields QT c (s, x) = qc +(s)E [ eρc +(s)(ξ− c (θs)+T−x), ξ−c (θs) + T − x > 0 ] × × ( E [ ec(ξ− c (θs)+T ), ξ−c (θs) + T < 0 ] + E [ eρc +(s)(ξ− c (θs)+T ), ξ−c (θs) + T > 0 ])−1 . Taking into account that, for x > 0, P {ξ+ c (θs) > x} = qc +(s)e−ρc +(s)x −→ c→∞ e−ρ+ 0 (s)x, where ρ+ 0 (s) is the positive solution of the equation ψ0(−ir) := ar − λ2 (∫ 0 −∞ erxdF (x) − 1 ) = 0, we get QT c (s, x) → QT ∞(s, x) as c → ∞. If we denote ξ0 ±(t) = sup 0≤u≤t (inf)ξ0(u), then QT ∞(s, x) = E [ eρ+ 0 (s)(ξ0 −(θs)+T−x), ξ0 −(θs) + T − x > 0 ] × × ( E [ eρ+ 0 (s)(ξ0 −(θs)+T ), ξ0 −(θs) + T > 0 ])−1 = ∫ T−x 0 eρ+ 0 (s)(T−x−y)dP {−ξ0 −(θs) < y }×(∫ T 0 eρ+ 0 (s)(T−y)dP {−ξ0 −(θs) < y })−1 = Rs(T − x)R−1 s (T ), where the last relation is the well-known formula (see [3]) for the upper semicontinuous processes. Proof of Theorem 2. The first relation in (10) follows from Eqs. (13) and (14). The second relation follows from the first one. The first equality in (11) was proved in [9]. After inverting (11), we get hs(T, x, z) =p+(s) ∂ ∂z P−(s, z)I {z < 0} + q+(s)ρ+(s) ∫ min{z,0} x−T e−ρ+(s)(z−y)dP−(s, y)− − QT (s, x) [ p+(s) ∂ ∂z P { ξ−(θs) + θ′c + x ≤ z } + + q+(s)ρ+(s) ∫ z x−T e−ρ+(s)(z−y)dP { ξ−(θs) + θ′c + x < z }] . (21) Using the integral transformation of (21) with respect to the distribution of θ′c, we get formula (12). Corollary 1. For the joint distribution {τ−(x, T ), ξ(τ−(x, T ))}, we have (22) sE [ e−sτ−(x,T ), ξ(τ−(x, T )) < z, A−(x) ] = ∫ x x−T Π−(z−y)dHs(T, x, y), z ≤ x−T, where Hs(T, x, y) is determined by its density (12) and Π−(x) = ∫ x −∞ Π(dy), x < 0. 78 D. V. GUSAK AND E. V. KARNAUKH The probability of the lack of exit (non-exit) from the interval (x − T, x) has the form (23) P {τ(x, T ) > θs} = P { ξ−(θs) > x − T }− − QT (s, x) [∫ −T −∞ ec(z+T )dP−(s, z) + P { ξ−(θs) > −T }] . The m.g.f. for τ(x, T ) and τ−(x, T ) are determined in the following way: (24) { Q(T, s, x) = 1 − P {τ(x, T ) > θs} , 0 < x < T, QT (s, x) = Q(T, s, x) − QT (s, x), 0 < x < T. Proof. Formula (22) follows from [6, Theorem 7.3]. By substituting (12) in (22), we obtain the relation in terms of QT (s, x) and the truncated distribution of ξ−(θs) + θ′c. Taking into account that P {τ(x, T ) > θs} = ∫ x x−T dHs(T, x, z) = = P { ξ−(θs) > x − T }− q+(s) ∫ 0 x−T eρ+(s)(y−(x−T ))dP−(s, z)+ + QT (s, x) [∫ 0 −T eρ+(s)(z+T )dP−(s, z) − P { ξ−(θs) > −T }] , and using formula (9), we obtain (23) after some simple transformations. Substitut- ing (23) into the first relation of (24), we find the m.g.f. of τ(x, T ), and then we can get the m.g.f. of τ−(x, T ) (see the second relation in (24)). On the basis of formulas (6) - (8), we can get the following statement about the limit behavior of QT (s, x) and hs(T, x, z) as s → 0. Corollary 2. The function h′ 0(T, x, z) = lims→0 s−1hs(T, x, z) (x − T < z < x, z �= 0, 0 < x < T ) according to the sign of m has the following forms: if m > 0 (25) h′ 0(T, x, z) = 1 m ( c−1 ∂ ∂z P { ξ− < z }− P { ξ− > z }) I {z < 0}+ + 1 m QT (x)P { ξ− > z − x } ; if m < 0 (26) h′ 0(T, x, z) = ( −p+ ∂ ∂z E τ−(z) + q+ρ+ ∫ 0 z eρ+(y−z)dE τ−(y) ) I {z < 0}− − QT (x)ρ+ ∫ 0 z−x eρ+(y−(z−x))dE τ−(y); if m = 0 (27) h′ 0(T, x, z) = ( − ∂ ∂z E τ0(z) − cλ−1 + c ∫ 0 z ∂ ∂y E τ0(y)dy ) I {z < 0}+ + cQT (x) ( λ−1 − ∫ 0 z−x ∂ ∂y E τ0(y)dy ) . The ruin probability QT (x) = lim s→0 QT (s, x) ON THE EXIT FROM THE INTERVAL FOR THE RISK PROCESS 79 (according to the sign of m) is determined from (9) in the following way: (28) QT (x) = ⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩ ∫ 0 x−T dP { ξ− < y }× × [∫ −T −∞ ec(T+y)dP { ξ− < y } + ∫ 0 −T dP { ξ− < y }]−1 , m > 0, q+e−ρ+x ( 1 λp+ − ∫ 0 x−T eρ+y ∂ ∂y E τ−(y)dy ) × × [ 1 λp+ − e−ρ+T ∫ −T −∞ ec(T+y) ∂ ∂y E τ−(y)dy− − ∫ 0 −T eρ+y ∂ ∂y E τ−(y)dy ]−1 , m < 0, ( λ−1 − ∫ 0 x−T ∂ ∂y E τ0(y)dy ) × × [ λ−1 − ∫ −T −∞ ec(T+y) ∂ ∂y E τ0(y)dy − ∫ 0 −T ∂ ∂y E τ0(y)dy ]−1 , m = 0. The distribution of ξ(τ−(x, T )) has the form (29) P { ξ(τ−(x, T )) < z, A−(x) } = 1 λ Π−(z) + ∫ 0− x−T Π−(z − y)h′ 0(T, x, y)dy+ + ∫ x 0+ Π−(z − y)h′ 0(T, x, y)dy, z < x − T. Corollary 3. For the process ξ(t) with the cumulant function (30) ψ(α) = λp(c(c − iα)−1 − 1) + λq(b(b + iα)−1 − 1), QT (x) is represented in the following way (0 < x < T ): (31) QT (x) = ⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎪⎩ ( 1 − q−eρ−(x−T ) )( 1 − q−c (c + ρ−)−1 e−ρ−T )−1 , m > 0, q+e−ρ+x ( 1 − b(ρ+ + b)−1eρ+(x−T ) ) ( 1 − b(ρ+ + b)−1q+e−ρ+T )−1 , m < 0, c(1 + b(T − x)) b + c + bcT , m = 0. If ξ(t) is a symmetric process (p = q = 1/2, b = c), then QT (x) = 1 + c(T − x) 2 + cT , QT (x) = 1 + cx 2 + cT , (0 < x < T ). Proof. Let’s note that the process with cumulant (30) is the almost upper and lower semicontinuous process. Then, in addition to relations (4) - (5), we have (32) ϕ−(s, α) = p−(s)(b + iα) ρ−(s) + iα , where −ρ−(s) = −bp−(s) is the negative root of the equation ψ(−ir) = s, s > 0, (33) P { ξ−(θs) < x } = T−(s, x) = q−(s)eρ−(s)x, x < 0. 80 D. V. GUSAK AND E. V. KARNAUKH If m > 0, then (34) P { ξ−(θs) < x } −→ s→0 P { ξ− < x } = q−ebp−x, x < 0, p−(s) −→ s→0 p− > 0. Taking into account that p+(s)p−(s) = s (s + λ)−1, we have, for m < 0, q′−(s) = −p′−(s) → −(λp+)−1 as s → 0. Hence, (35) E τ−(x) = − ∂ ∂s T−(s, x)|s=0 = 1 − bx λp+ , x < 0. If m = 0, then we have, for ξ̃0(t), Π0(dx) = λ0be bxdx, x < 0, λ0 = λq(c + b)b−1. Moreover, ξ̃−0 (t) = ξ̃0(t), p0 −(s) = P { ξ̃0(θs) = 0 } = s s + λ0 . Hence, the m.g.f. of τ0(x) has the form T− 0 (s, x) = E e−sτ0(x) = q0 −(s)ebp0 −(s)x, x < 0. Since (p0 −)′(s) = −(q0 −)′(s) → λ−1 0 as s → 0, we get (36) E τ0(x) = − ∂ ∂s T− 0 (s, x) ∣∣ s=0 = 1 − bx λ0 , x < 0. Substituting formulas (34) - (36) into the corresponding relations of (28), we get (31). Remark. We should note that it is easy to get the representation of the m.g.f. of the functionals related to the exit from the interval for the almost lower semicontinuous process η(t) (with the parameter b > 0, by considering that ξ(t) = −η(t)). Particularly, QT (s, x) = q−(s) ∫ x 0 eρ−(s)(x−y)dP+(s, y)× × [∫ ∞ T eb(T−y)dP+(s, y) + ∫ T 0 eρ−(s)(T−y)dP+(s, y) ]−1 . Let ξ(t) be the almost upper semicontinuous piecewise constant process. Then ξ1(t) = at + ξ(t), a < 0, is the almost upper semicontinuous piecewise linear process. For the process ξ1(t) on the basis of the stochastic relations for τ+(x, T ), τ+(x, T )=̇ { ζ, ξ + aζ > x, ζ + τ+(x − ξ − aζ), x − T < ξ + aζ < x, we have the integro-differential equation for QT (s, x) a ∂ ∂x QT (s, x) = λ ∫ x x−T QT (s, x − z)dF1(z) − (s + λ)QT (s, x) + λpe−cx, 0 < x < T. Introducing the function Q T (s, x) = 1 − QT (s, x) and following the reasoning analo- gous to that for the piecewise constant process ξ(t), we can get the representation of the functionals related to the exit from the interval (x − T, x) for the piecewise linear processes. Two boundary-value problems for the integer-valued random walks are considered in [10] and, for the process with stationary independent increments, are treated in [11]. ON THE EXIT FROM THE INTERVAL FOR THE RISK PROCESS 81 Bibliography 1. I.I. Gikhman, A.V. Skorokhod, Theory of Stochastic Processes, vol. 2, Nauka, Moscow, 1973, pp. 620. (in Russian) 2. A.V. Skorokhod, Stochastic Processes with Independent Increments, Nauka, Moscow, 1964, pp. 280. (in Russian) 3. V.S. Korolyuk, Boundary Problems for Compound Poisson Processes, Naukova Dumka, Kyiv, 1975, pp. 140. (in Russian) 4. V.M. Shurenkov, Limit distribution of the exit time from an expanding interval and of the position at exit time of a process with independent increments and one-signed jumps, Prob. Theory Appl. 23 (1978), no. 2, 402-407. 5. V.S. Korolyuk, N.S. Bratiichuk, B. Pirdjanov, Boundary-Value Problems for Random Walks, Ilim, Ashkhabad, 1987, pp. 250. (in Russian) 6. N.S. Bratiichuk, D.V. Gusak, Boundary Problems for Processes with Independent Increments, Naukova Dumka, Kyiv, 1990, pp. 264. (in Russian) 7. D.V. Gusak, Compound Poisson processes with two-sided reflection, Ukr. Math. J. 53 (2002), no. 12, 1616-1625. 8. D.V. Gusak, Risk process with stochastic premiums and distributions of their functionals, The- ory of Stoch. Process. 11(27) (2005), no. 1-2, 29-39. 9. E.A. Pecherskiy, Some identities related to the exit of a random walk out of a segment and a semiinterval, Probab. Theory Appl. 19 (1974), no. 1, 104-119. (in Russian) 10. T.V. Kadankova, Two-sided boundary problems for the random walks with geometrically dis- tributed negative jumps, Teor. Imov. and Mat. Stat. 68 (2003), 49-60. (in Ukrainian) 11. V.F. Kadankov, T.V. Kadankova, On the distribution of the moment of the first exit from an interval and the value of overshoot through the boundary for the processes with independent increments and random walks, Ukr. Math. J. 57 (2005), no. 10, 1359-1385. E-mail : random@imath.kiev.ua E-mail : kveugene@univ.kiev.ua
id nasplib_isofts_kiev_ua-123456789-4427
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 0321-3900
language English
last_indexed 2025-12-02T09:16:04Z
publishDate 2005
publisher Інститут математики НАН України
record_format dspace
spelling Gusak, D.V.
Karnaukh, E.V.
2009-11-09T15:33:32Z
2009-11-09T15:33:32Z
2005
On the exit from a finite interval for the risk processes with stochastic premiums / D.V. Gusak, E.V. Karnaukh // Theory of Stochastic Processes. — 2005. — Т. 11 (27), № 3-4. — С. 71–81. — Бібліогр.: 11 назв.— англ.
0321-3900
https://nasplib.isofts.kiev.ua/handle/123456789/4427
519.21
We consider the almost semicontinuous step-process ξ(t). The conditional characteristic functions of the jumps of ξ(t) have the form E [eiαξk /ξk > 0] = c(c − iα)−1. For such processes, the boundary functionals related to the exit from a finite interval are investigated.
en
Інститут математики НАН України
On the exit from a finite interval for the risk processes with stochastic premiums
Article
published earlier
spellingShingle On the exit from a finite interval for the risk processes with stochastic premiums
Gusak, D.V.
Karnaukh, E.V.
title On the exit from a finite interval for the risk processes with stochastic premiums
title_full On the exit from a finite interval for the risk processes with stochastic premiums
title_fullStr On the exit from a finite interval for the risk processes with stochastic premiums
title_full_unstemmed On the exit from a finite interval for the risk processes with stochastic premiums
title_short On the exit from a finite interval for the risk processes with stochastic premiums
title_sort on the exit from a finite interval for the risk processes with stochastic premiums
url https://nasplib.isofts.kiev.ua/handle/123456789/4427
work_keys_str_mv AT gusakdv ontheexitfromafiniteintervalfortheriskprocesseswithstochasticpremiums
AT karnaukhev ontheexitfromafiniteintervalfortheriskprocesseswithstochasticpremiums