Distribution of the maximum of the Chentsov random field

Let D = [0, 1]^2 and X(s, t), (s, t) belongs D, be a two-parameter Chentsov random field. The aim of this paper is to find the probability distribution of the maximum of X(s, t) on a class of polygonal lines.

Збережено в:
Бібліографічні деталі
Дата:2008
Автор: Kruglova, N.
Формат: Стаття
Мова:Англійська
Опубліковано: Інститут математики НАН України 2008
Онлайн доступ:https://nasplib.isofts.kiev.ua/handle/123456789/4538
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Цитувати:Distribution of the maximum of the Chentsov random field / N. Kruglova // Theory of Stochastic Processes. — 2008. — Т. 14 (30), № 1. — С. 76–81. — Бібліогр.: 8 назв.— англ.

Репозитарії

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859668784165421056
author Kruglova, N.
author_facet Kruglova, N.
citation_txt Distribution of the maximum of the Chentsov random field / N. Kruglova // Theory of Stochastic Processes. — 2008. — Т. 14 (30), № 1. — С. 76–81. — Бібліогр.: 8 назв.— англ.
collection DSpace DC
description Let D = [0, 1]^2 and X(s, t), (s, t) belongs D, be a two-parameter Chentsov random field. The aim of this paper is to find the probability distribution of the maximum of X(s, t) on a class of polygonal lines.
first_indexed 2025-11-30T12:30:39Z
format Article
fulltext Theory of Stochastic Processes Vol. 14 (30), no. 1, 2008, pp. 76–81 UDC 519.21 NATALIA KRUGLOVA DISTRIBUTION OF THE MAXIMUM OF THE CHENTSOV RANDOM FIELD Let D = [0, 1]2 and X(s, t), (s, t) ∈ D, be a two-parameter Chentsov random field. The aim of this paper is to find the probability distribution of the maximum of X(s, t) on a class of polygonal lines. 1. Introduction Let {X(s, t) : s, t ≥ 0} be a standard Chentsov field of two parameters that is a sepa- rable real Gaussian stochastic process such that 1) X(0, t) = X(s, 0) = 0 for all s, t ∈ [0, 1]; 2) E[X(s, t)] = 0 for all s, t ≥ 0; 3) E[X(s, t)X(s1, t1)] = min{s, s1}min{t, t1} for all (s, t) and (s1, t1) ∈ D. This definition is given by Yeh [6] in 1960. Another (equivalent) definition is given by Chentsov [7] in 1955 in terms of the probability density of X(s, t). Yeh showed that the sample paths of this field are continuous with probability one andX(s, t) has independent stationary increments in the plane. The probability distributions of functionals of a Chentsov random field like M = max(s,t)∈D X(s, t) are not yet known. Some trivial probability distribution theory for X(s, t) can be obtained by using the known results about the standard Wiener process. The distribution of the supremum of a Chentsov random field on the curve f(s), where f(s) is a non-decreasing function of s, can be obtained, since a transformation of X(s, f(s)) is equivalent to a one-dimensional standard Wiener process. The probability distribution of the supremum of X(s, t) on the boundary of a unit square is obtained by Paranjape and Park [1]. This probability is of its own interest, and it gives a nice lower bound for the probability distribution of the supremum of X(s, t) over the whole unit square D, which is unknown yet. Park and Skoug [5] have found the probability that X(s, t) crosses a barrier of the type ast+ bs+ ct+ d on the boundary ∂Λ, where Λ = [0, S]× [0, T ] is a rectangle. Later on, I. Klesov [3] considered a probability of the form (1) P (L, g) = P { sup L X(s, t)− g(s, t) < 0 } , whereX is a Chentsov random field on D = [0, 1]2, L ⊂ D, and g is an almost everywhere Lebesgue continuous function onD. He presented results, where g(s, t) is a linear function and L is a polygonal line with one point of break. Klesov and Kruglova [8] considered a probability of the form (1), where L is a polygonal line with two points of break. The main purpose of this paper is to evaluate the probability distribution of the form (1), where g(s, t) = λ and L is a polygonal line with several points of break. we can Key words and phrases. Two-parameter Chentsov random field, probability distribution, standard Wiener process. 76 MAXIMUM OF THE CHENTSOV RANDOM FIELD 77 express this distribution in a very useful form: as an expression of the ”tail” of the two-dimensional Gauss process. 2. Auxiliary results Lemma 1. (Doob’s Transformation Theorem) [2]. Let X(t) be any Gaussian process with covariance function R(s, t) = u(s)v(t), s ≤ t, if the ratio a(t) = u(t)/v(t) is con- tinuous and strictly increasing with inverse a1(t), then w(t) and Y (a1(t))/v(a1(t)) are stochastically equivalent processes. Lemma 2. (Malmquist’s Theorem 1) [4]. For a standard Wiener process w(t) and for b > 0, a ≥ 0, s1 ≤ at′ + b, P { w(t) ≤ at+ b, 0 < t < t ′ |w(t ′ ) = s1 } = = P { w(t) ≤ bt+ (at ′ + b− s1)/t′ , 0 < t <∞ } = = 1− exp { −2b(at ′ + b− s1/t′ } . Lemma 3. (Malmquist’s Theorem 2) [4]. For a standard Wiener process w(t) and for b > 0, a ≥ 0, P {w(t) ≤ at+ b, x < t ≤ y|w(x) = s1, w(y) = s2} = = 1− exp { − 2R 1−R2 · P1 − s1√ x · P2 − s2√ y } , where R = √ x y , s1 ≤ P1 = ax+ b, s2 ≤ P2 = ay + b. Let L be a line as shown in Fig. 1 and given by the formula t s A B 0 1 1 Q Figure 1 (2) L = { (s, t) : sa−1 + t = 1, s ≤ k; s+ tb−1 = 1, s > k, (s, t) ∈ D} , where tanα = a, tanβ = b, k = a(b−1) ab−1 , α, β > π 4 . 78 NATALIA KRUGLOVA Theorem 1. (Paranjape and Park)[1]. Let {X(s, t) : s, t ≥ 0} be a standard Chentsov field. Then (3) P { sup (s,t)∈L X(s, t) ≤ λ } = Φ ( λ(a+ c) a √ c ) − exp {−2λ2 a } Φ ( λ(c− a) a √ c ) − − exp {−2λ2 b } Φ { λ(1 − bc) b √ c } + exp {−2λ2(a−1 + b−1 − 2) } × Φ { λc−1/2(b−1 − c− 2) } . 3. Main results and proofs Let L be a line as shown in Fig. 2 and given by the formula t s A B 0 1 1 Figure 2 Q1(x1, y1) Q2(x2, y2) (4) L = ⎧⎪⎪⎨⎪⎪⎩ t = 1− s(1−y1) x1 , s ∈ [0, x1] t = − s(y1−y2) x2−x1 + x2y1−x1y2 x2−x1 , s ∈ (x1, x2] t = − sy2 1−x2 + y2 1−x2 , s ∈ (x2, 1]. Theorem 2. Let X(s, t) be a standard Chentsov random field on a unit square. Let the polygonal line L have two points of break Q1(x1, y1) and Q2(x2, y2) and be given by formula (4). Let the coordinates of Q1 and Q2 satisfy the conditions 1) y2 < y1; 2) x2 y2 > x1 y1 . Then P2 = P { sup (s,t)∈L X(s, t) < λ } × ∫ λ y1 −∞ ∫ λ y2 −∞ 1 2π √ x1 y1 ( x2 y2 − x1 y1 ) exp { − u2 1 2x1 y1 } exp ⎧⎨⎩− (u2 − u1)2 2 ( x2 y2 − x1 y1 ) ⎫⎬⎭ × ( 1− exp { −2λy1 x1 ( λ y1 − u1 )})( 1− exp { −2λ ( λ y2 − u2 )}) MAXIMUM OF THE CHENTSOV RANDOM FIELD 79 (5) × ( 1− exp { −2(λ− u1y1)(λ − u2y2) (x2y1 − x1y2) }) du1du2 Corollary 1. Passing to the limit as Q1 −→ Q2 and using (5), we obtain a result which agrees with Park’s result for a polygonal line with a single point of break (Theorem 1). Let us denote that x0 = 0, xn+1 = 1, y0 = 1, yn+1 = 0. Let L be a line given by the formula (6) L = {(s, t) : t = v(s), s ∈ [0, 1]} . For which (x1, y1), . . . , (xn, yn) are the points of break where v(s) = n+1∑ i=1 ( −s(yi−1 − yi) xi − xi−1 + xiyi−1 − xi−1yi xi − xi−1 ) I(xi−1;xi](s). Let us denote Δ0 = 0,Δi = xi yi , i = 1, n,Δn+1 =∞. The following theorem is a generalization of Theorem 2. Theorem 3. Let X(s, t) be a standard Chentsov random field on a unit square. Let u0 = un+1 = 0. Let the polygonal line L have n points of break and be given by formula (6). Let the coordinates of these points satisfy the conditions 1) y1 > · · · > yn; 2) x1 y1 < · · · < xn yn . Then Pn = P { sup (s;t)∈L X (s; t) < λ } = ∫ λ y1 −∞ . . . ∫ λ yn −∞ n∏ i=1 ϕ0,Δi−Δi−1 (ui − ui−1) × n+1∏ i=1 ⎛⎝1− exp ⎧⎨⎩−2 ( λ yi−1 − ui−1 )( λ yi − ui ) (Δi −Δi−1) ⎫⎬⎭ ⎞⎠du1 . . . dun where ϕ0,Δ(u) is the density of the Gaussian random variable with variance Δ. Proof. Let the restriction of X(s, t) over L be denoted by w1(s). Then w1(s) = X (s, v(s)) Let us find the derivation of v(s). v′(s) = n+1∑ i=1 − (yi−1−yi) xi−xi−1 I(xi−1;xi](s) < 0 because of conditions over coordinates of points. This means that v(s) is monotone decreasing function. Using the covariance property of X(s, t), we can write cov(w1(s1), w1(s2)) = cov (X(s1, v(s1), X(s2, v(s2)) = s1v(s2), 0 < s1 � s2 � 1 a(s) = s v(s) is continuous monotone increasing function. We can write a(s) in an explicit form: a(s) = n+1∑ i=1 s − s(yi−1−yi) xi−xi−1 + xiyi−1−xi−1yi xi−xi−1 I(xi−1,xi](s) It is enough to prove a continuity a(s) in the points xi, i = 1, n: a(xi) = xi −xi(yi−1−yi) xi−xi−1 + xiyi−1−xi−1yi xi−xi−1 = xi yi . 80 NATALIA KRUGLOVA a(xi+) = xi −xi(yi−yi+1) xi+1−xi + xi+1yi−xiyi+1 xi+1−xi = xi yi(xi+1−xi) xi+1−xi = xi yi . That is a(xi) = a(xi+) continuous in the point xi. That is a(s) continuous in (0; 1). s is a monotone increasing function and v(s) is a monotone decreasing function. That is why a(s) is a monotone increasing function. For a(s) the inverse will be the function: a−1(s) = n+1∑ i=1 s(xiyi−1 − xi−1yi) s(yi−1 − yi) + xi − xi−1 I[Δi−1,Δi). It is necessary to notice that v(0) = −x0(y0−y1) x1−x0 + x1y0−x0y1 x1−x0 = y0 = 1 and v(1) = −xn+1(yn−yn+1) xn+1−xn + xn+1yn−xnyn+1 xn+1−xn = yn+1 = 0. That is why a(0) = 0 and lim t→1 a(t) =∞. 1 v(a−1(s)) = n+1∑ i=1 ( s(yi−1 − yi) + xi − xi−1 xiyi−1 − xi−1yi ) I[Δi−1,Δi)(s) The functions a(s) and v(·) satisfy the conditions of Doob’s transformation theorem. Thus, w∗(s) = n+1∑ i=1 ( s(yi−1 − yi) + xi − xi−1 xiyi−1 − xi−1yi ) ×w1 ( s(xiyi−1 − xi−1yi) s(yi−1 − yi) + xi − xi−1 ) I[Δi−1,Δi)(s) and w(t) are stochastically equivalent processes. Pn(λ) = P { sup (s;t)∈L X (s; t) < λ } = P { sup s∈[0,1] X (s; v(s)) < λ } = P { sup s∈[0,1] w1(s) < λ } = P { sup s∈[0,∞) w1 ( a−1(s) ) < λ } = P ⎛⎝⋂ s�0 { w1 ( a−1(s) ) < λ }⎞⎠ = P (⋂ s>0 { w1 ( a−1(s) ) < λ }⋂{ X ( a−1(0), v ( a−1(0) )) < λ }) = P (⋂ s>0 { w1 ( a−1(s) ) < λ }⋂ Ω ) Because X ( a−1(0), v ( a−1(0) )) = X(0, 1) = 0 and that is why{ X ( a−1(0), v ( a−1(0) )) < λ } = Ω Pn = P (⋂ s>0 { w1 ( a−1(s) ) < λ }) = P {⋂ s>0 w1 ( a−1(s) ) v (a−1(s)) − λ v (a−1(s)) < 0 } = = P { sup s∈(0,∞) w1 ( a−1(s) ) v (a−1(s)) − λ v (a−1(s)) < 0 } = P { sup s∈(0,∞) w(t) − λ v (a−1(s)) < 0 } = = P { w(t) < λ(xi − xi−1) xiyi−1 − xi−1yi + λt(yi−1 − yi) xiyi−1 − xi−1yi ; t ∈ (Δi−1; Δi] , i = 1, n+ 1 } MAXIMUM OF THE CHENTSOV RANDOM FIELD 81 = ∫ λ y1 −∞ . . . ∫ λ yn −∞ 1 (2π)n/2 P { w(s) < λ+ (1− y1)sλ x1 , s ∈ (0; Δ1] |w(Δ1) = u1 } × × n∏ i=2 P { w(t) < λ(xi − xi−1) xiyi−1 − xi−1yi + λt(yi−1 − yi) xiyi−1 − xi−1yi ; t ∈ (Δi−1; Δi]∣∣∣ w(Δi−1) = ui−1, w(Δi) = ui } ×P { w(t) < λ(1− xn) yn + λt, t > Δn|w(Δn) = un } × n∏ i=1 ϕ0,Δi−Δi−1(ui − ui−1)√ Δi −Δi−1 dui. Then, by using Lemma 2 and Lemma 3, we get Pn = P { sup (s;t)∈L X (s; t) � λ } = ∫ λ y1 −∞ . . . ∫ λ yn −∞ n+1∏ i=1 ⎛⎝1− exp ⎧⎨⎩−2 ( λ yi−1 − ui−1 )( λ yi − ui ) (Δi −Δi−1) ⎫⎬⎭ ⎞⎠ × n∏ i=1 ϕ0,Δi−Δi−1 (ui − ui−1)du1 . . . dun Bibliography 1. S.R. Paranjape and C. Park, Distribution of the supremum of the two-parameter Yeh-Wiener process on the boundary, Appl. Probab. 10 (1973), no. 4, 875–880. 2. J.L. Doob, Heuristic approach to Kolmogorov–Smirnov theorems, Ann. Math. Statist. 20 (1949), 393–403. 3. I.I. Klesov, On the probability of attainment of a curvilinear level by a Wiener field, Probab. and Math. Statist. 51 (1995), 63–67. 4. S. Malmquist, On certain confidence contours for distribution functions, Ann. Math. Statist. 25 (1954), 523–533. 5. C. Park and D.L. Skoug, Distribution estimates of barrier-crossing probabilities of the Yeh- Wiener process, Pacific Journal of Mathematics 78 (1978), 455–466. 6. J. Yeh, Wiener Measure in a Space of Functions of Two Variables, Transactions of the American Mathematical Society 95 (1960), 433–450. 7. N.N. Chentsov, Wiener random fields depending on several parameters, Doklady AN SSSR 46 (1956), no. 4, 607–609. (Russian) 8. O.I. Klesov and N.V. Kruglova, Distribution of the maximum of the two-parameter Chentsov random field, Naukovi visti NTUU ”KPI” 4 (2007), 136–141. (Ukrainian) ��� ��� ����� ��� $� %��� �� $��� �� 0:&/�<1 E-mail : natahak@ukr.net
id nasplib_isofts_kiev_ua-123456789-4538
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 0321-3900
language English
last_indexed 2025-11-30T12:30:39Z
publishDate 2008
publisher Інститут математики НАН України
record_format dspace
spelling Kruglova, N.
2009-11-25T11:04:54Z
2009-11-25T11:04:54Z
2008
Distribution of the maximum of the Chentsov random field / N. Kruglova // Theory of Stochastic Processes. — 2008. — Т. 14 (30), № 1. — С. 76–81. — Бібліогр.: 8 назв.— англ.
0321-3900
https://nasplib.isofts.kiev.ua/handle/123456789/4538
519.21
Let D = [0, 1]^2 and X(s, t), (s, t) belongs D, be a two-parameter Chentsov random field. The aim of this paper is to find the probability distribution of the maximum of X(s, t) on a class of polygonal lines.
en
Інститут математики НАН України
Distribution of the maximum of the Chentsov random field
Article
published earlier
spellingShingle Distribution of the maximum of the Chentsov random field
Kruglova, N.
title Distribution of the maximum of the Chentsov random field
title_full Distribution of the maximum of the Chentsov random field
title_fullStr Distribution of the maximum of the Chentsov random field
title_full_unstemmed Distribution of the maximum of the Chentsov random field
title_short Distribution of the maximum of the Chentsov random field
title_sort distribution of the maximum of the chentsov random field
url https://nasplib.isofts.kiev.ua/handle/123456789/4538
work_keys_str_mv AT kruglovan distributionofthemaximumofthechentsovrandomfield