Manual for SOA Exam MLC.

Transcription

1 Chapter 10. Poisson processes. Section Nonhomogenous Poisson processes Extract from: Arcones Fall 2009 Edition, available at 1/14

2 Nonhomogenous Poisson processes Definition 1 The counting process {N(t) : t 0} is said to be a nonhomogenous Poisson process with intensity function λ(t), t 0, if (i) N(0) = 0. (ii) For each t > 0, N(t) has a Poisson distribution with mean m(t) = t 0 λ(s) ds. (iii) For each 0 t 1 < t 2 < < t m, N(t 1 ), N(t 2 ) N(t 1 ),..., N(t m ) N(t m 1 ) are independent r.v. s. 2/14

3 Nonhomogenous Poisson processes Definition 1 The counting process {N(t) : t 0} is said to be a nonhomogenous Poisson process with intensity function λ(t), t 0, if (i) N(0) = 0. (ii) For each t > 0, N(t) has a Poisson distribution with mean m(t) = t 0 λ(s) ds. (iii) For each 0 t 1 < t 2 < < t m, N(t 1 ), N(t 2 ) N(t 1 ),..., N(t m ) N(t m 1 ) are independent r.v. s. An nonhomogeneous Poisson process with λ(t) = λ, for each t 0, is a regular Poisson process. 3/14

4 Nonhomogenous Poisson processes Definition 1 The counting process {N(t) : t 0} is said to be a nonhomogenous Poisson process with intensity function λ(t), t 0, if (i) N(0) = 0. (ii) For each t > 0, N(t) has a Poisson distribution with mean m(t) = t 0 λ(s) ds. (iii) For each 0 t 1 < t 2 < < t m, N(t 1 ), N(t 2 ) N(t 1 ),..., N(t m ) N(t m 1 ) are independent r.v. s. An nonhomogeneous Poisson process with λ(t) = λ, for each t 0, is a regular Poisson process. m(t) is the mean value function of the non homogeneous Poisson process. 4/14

5 The increments of an nonhomogeneous Poisson process are independent, but not necessarily stationary. 5/14

6 The increments of an nonhomogeneous Poisson process are independent, but not necessarily stationary. A nonhomogeneous Poisson process is a Markov process. 6/14

7 The increments of an nonhomogeneous Poisson process are independent, but not necessarily stationary. A nonhomogeneous Poisson process is a Markov process. For each 0 s < t, N(t) N(s) has Poisson distribution with mean m(t) m(s) = t s λ(x) dx. 7/14

8 The increments of an nonhomogeneous Poisson process are independent, but not necessarily stationary. A nonhomogeneous Poisson process is a Markov process. For each 0 s < t, N(t) N(s) has Poisson distribution with mean m(t) m(s) = t s λ(x) dx. For each 0 t 1 < t 2 < < t m and each integers k 1,..., k m 0, P{N(t 1 ) = k 1, N(t 2 ) = k 2,..., N(t m ) = k m } = P{N(t 1 ) = k 1, N(t 2 ) N(t 1 ) = k 1 k 1,..., N(t m ) N(t m 1 ) = k m k m 1 } = e m(t 1 ) (m(t 1 )) k 1 e (m(t 2 ) m(t 1 )) (m(t 2 ) (m(t 1 )) k 2 k 1 k 1! (k 2 k 1 )! e (m(tm) m(t m 1 )) (m(t m) m(t m 1 )) km k m 1 (k m k m 1 )!, 8/14

9 Example 1 For a nonhomogenous Poisson process the intensity function is given by { 5 if t is in (1, 2], (3, 4],... λ(t) = 3 if t is in (0, 1], (2, 3],... Find the probability that the number number of observed occurrences in the time period (1.25, 3] is more than two. 9/14

10 Example 1 For a nonhomogenous Poisson process the intensity function is given by { 5 if t is in (1, 2], (3, 4],... λ(t) = 3 if t is in (0, 1], (2, 3],... Find the probability that the number number of observed occurrences in the time period (1.25, 3] is more than two. Solution: N(3) N(1.25) has a Poisson distribution with mean m(3) m(1.25) = Hence, λ(t) dt = dt dt = P{N(3) N(1.25) > 2} = 1 e 6.75 ( (6.75) 2 /2) = /14

11 Let S n be the time of the n th occurrence. S n is an extended r.v. with values in [0, ]. Then, P{S n > t} = P{N(t) n 1} = n 1 j=0 = P{Gamma(n, 1) m(t)}. If lim t m(t) =, then e m(t) (m(t)) j j! P{S n = } = lim t P{S n > t} = P{Gamma(n, 1) lim t m(t)} = 0 and S n is a r.v. The density of S n is f Sn (t) = e m(t) (m(t))n 1 m (t) (n 1)! If lim t m(t) <, then S n is a mixed r.v. with = e m(t) (m(t))n 1 λ(t), t 0. (n 1)! P{S n = } = P{Gamma(n, 1) lim t m(t)} > 0 and density of its continuous part f Sn (t) = e m(t) (m(t)) n 1 λ(t) (n 1)!, t 0. 11/14

12 Let T n = S n S n 1 be the n th interarrival time. For a non homogeneous Poisson process {T n } n=1 are not necessarily independent r.v. s. For 0 s t, P{T n+1 > t S n = s} = P{S n+1 > s + t S n = t} =P{N(s + t) = n + 1 S n = t} = P{N(s + t) N(s) = 1 S n = t} =P{N(s + t) N(s) = 1} = e (m(s+t) m(s)). Notice S n = t depends on the non homogeneous Poisson process until time t. Hence, {N(s + t) N(s) = 1} and {S n = t} are independent. 12/14

13 Example 2 For a non homogenous Poisson process, the intensity function is given by { t for 0 < t 4, λ(t) = 4 for 10 < t. If S 5 = 2, calculate the probability that S 6 > 5. 13/14

14 Example 2 For a non homogenous Poisson process, the intensity function is given by { t for 0 < t 4, λ(t) = 4 for 10 < t. If S 5 = 2, calculate the probability that S 6 > 5. Solution: We have that P{S 6 > 5 S 5 = 2} = P{N(5) = 5 S 5 = 2} = P{N(5) N(2) = 0 S 5 = 2} = P{N(5) N(2) = 0} and m(5) m(2) = 5 2 λ(t) dt = 4 2 t dt dt = 10. Hence, P{S 6 > 5 S 5 = 2} = P{N(5) N(2) = 0} = e /14