STAT3603

Jeff Yao

STAT3603 Stochastic Processes

Jeff Yao
STAT3603

Jeff Yao
Part I
Introduction
to
Probability
theory

Random Probability tools

variables

Conditional
Probabili-
ties and
Conditional
Expectation
1 Introduction to Probability theory

2 Random variables

3 Conditional Probabilities and Conditional Expectation

STAT3603

Jeff Yao
Part II
Introduction

Chapman-
Kolmogorov
Equation

Classification
Markov Chains
of states

Limiting
Probabili- 1 Introduction
ties

Mean Time
Spent in 2 Chapman-Kolmogorov Equations
Transient
States

Applications
3 Classification of States

4 Limiting Probabilities

5 Mean Time Spent in Transient States

6 Applications
STAT3603

Jeff Yao Part III

The Poisson
Process

Interarrival
and Waiting
The Poisson Process
Time Distri-
butions

Conditional
Distribution 1 The Exponential Distribution
of the
Arrival
Times 2 Independent exponential variables
Further
Properties
of Poisson 3 The Poisson Process
Process

Compound
Poisson
4 Interarrival and waiting time distributions
Process

Non- 5 Conditional Distribution of the Arrival Times

homogeneous
Poisson
process 6 Further Properties of Poisson Processes

7 Compound Poisson Process

 §3.1 The Exponential Distribution

STAT3603

Jeff Yao
Basic properties
The Poisson
Process
I X ∼ E(λ), exponential distribution with rate parameter λ, has density
Interarrival
and Waiting function
Time Distri-
butions

Conditional
f (x) = λe −λx 1{x≥0} .
Distribution
of the
Arrival
Times

Further
Properties
I Survival function: F (x) = P(X > x) = e −λx , x ≥ 0.
of Poisson
Process 1 1
I Moments: E (X ) = , var(X ) = ;
Compound
λ λ2
Poisson
Process I The m.g.f.
Non-
homogeneous λ
Poisson φ(t) = E (e tX ) = , t < λ.
process λ−t
 Memoryless property

STAT3603

Jeff Yao
I Nonnegative r.v. are often used as models for lifetimes.

The Poisson
Process
I Definition: the survival function of a nonnegative variable (exponential,
Interarrival
gamma, χ2 , Weibull,...) is
and Waiting
Time Distri-
butions F (x) = P(X > x), x ≥ 0.
Conditional
Distribution
of the
Arrival
Times Memoryless property
Further
Properties I For a exponential variable X , we have
of Poisson
Process

Compound P{X > s + t|X > t} = P{X > s} for all s, t > 0.
Poisson
Process

Non-
homogeneous Meaning: if the instrument has survived until time t, then it will survive
Poisson
process as it was a completely new item. (damageless, memoryless).

Conversely, this memoryless property characterizes the exponenyial

distributions: no other distribution has such property.
 §3.2 Independent exponential variables

STAT3603

Jeff Yao

The Poisson
I Recall a γ(α, λ) distribution has density
Process

Interarrival (λx)α−1
and Waiting λe −λx , x ≥ 0,
Time Distri- Γ(α)
butions

Conditional with shape parameter α > 0 and scale parameter λ > 0.

Distribution
of the
Arrival
Times
I If X ∼ γ(α, λ) and X ∼ γ(α0 , λ) are independent, X + Y ∼ γ(α + α0 , λ).
Further
Properties I Obviously, E(λ) = γ(1, λ).
of Poisson
Process

Compound I If X1 , . . . , Xn are i.i.d. E(λ), then Sn = X1 + · · · + Xn is a γ(n, λ) variable

Poisson
Process with density function
Non-
homogeneous (λx)n−1
Poisson λe −λx , x > 0.
process (n − 1)!
 Independent exponential variables

STAT3603

Jeff Yao
We discuss two important properties of a sequence
The Poisson X1 , . . . , Xn of independent exponential r.v.
Process
with respective rates λ1 , . . . , λn (so not i.d.).
Interarrival
and Waiting
Time Distri-
butions Important property I
Conditional
Distribution Their minimum is still a exponential r.v.:
of the
Arrival
Times min{X1 , . . . , Xn } ∼ E(λ1 + · · · + λn ).
Further
Properties
of Poisson
Process

Compound
Proof.
Poisson
Process P(min{X1 , . . . , Xn } > t) = P(X1 > t, . . . , Xn > t)
Non-
homogeneous = P(X1 > t) · · · P(Xn > t)
Poisson
process = e −(λ1 +···+λn )t .

Warning. Do not confuse with max(X1 , . . . , Xn ).

STAT3603
Cont’d.
Jeff Yao
A lemma
The Poisson
Process Suppose X1 and X2 are independent exponential random variables with
Interarrival respective rates λ1 and λ2 . Then
and Waiting
Time Distri- λ1
butions 1 P{X1 < X2 } = .;
λ1 + λ2
Conditional
Distribution
of the   λ1
Arrival 2 E X1 1{X1 <X2 } = .
Times (λ1 + λ2 )2
Further
Properties
of Poisson Proof.
Process

Compound
Poisson
Process

Non-
homogeneous
Poisson
process

1
Trivial case: if λ1 = λ2 , P{X1 < X2 } = P{X2 < X1 } = 2
which is trivial by
symmetry.
STAT3603
Cont’d.
Jeff Yao
Important property II: index of the minimum
The Poisson
Process Let Rn be the index of the observation realizing the minimum of the sequence,
Interarrival i.e.
and Waiting
Time Distri-
butions Rn = i if Xi = min{X1 , . . . , Xn } .
Conditional
Distribution
of the Then
Arrival
Times λi
P(Rn = i) = , 1 ≤ i ≤ n.
Further
Properties
λ1 + · · · + λn
of Poisson
Process

Compound
Poisson
Proof.
Process

Non-
homogeneous
Poisson
process
 Application 1

STAT3603
Let X1 , X2 , X3 be independent exponential random variables with respective
Jeff Yao
parameters λi , i = 1, . . . , 3. Show that
The Poisson
Process λ1 λ2
Interarrival
P{X1 < X2 < X3 } =
and Waiting
(λ2 + λ3 )(λ1 + λ2 + λ3 )
Time Distri-
butions

Conditional I Classical method:

Distribution
of the
Arrival
Times

Further
Properties
of Poisson
Process

Compound
Poisson I Conditioning method:
Process

Non-
homogeneous
Poisson
process
 Application 2

STAT3603
I post office run by two clerks whose service time are both E(λ)-distributed;
Jeff Yao

The Poisson I Smith enters the system when both Jones and Brown are being served by a
Process
clerk;
Interarrival
and Waiting
Time Distri- I The serive to Smith will begin as soon as either Jones or Brown leaves.
butions

Conditional Question. What is the probability that, of the three customers, Mr. Smith is
Distribution
of the the last to leave the post office?
Arrival
Times

Further
Jones Smith
Properties s
of Poisson
Process

Compound Brown Brown

Poisson
Process

Non- Solution: Assume J leaves first at time s, then, from

homogeneous
Poisson
process
P{B > s + t|B > s} = P{B > t}, the additional time T that B requires has
the same (and independent) distribution as S. So P{S − T > 0} = 1/2.
 Application 3

STAT3603
I You arrive at a post office having two clerks;
Jeff Yao

The Poisson I their service times are E(λi ), i = 1, 2;

Process

Interarrival
and Waiting I The clerks are busy but no one else is waiting in line, and you will enter service
Time Distri-
butions
when either clerk becomes free.
Conditional Question. Find E [T ], where T = amount of time that you spend in the office.
Distribution
of the
Arrival
Times
Solution: let Xi be the service time of clerk i. X1 X'1
Further I T = (X1 + X10 )1{X1 <X2 } + (X2 + X20 )1{X2 <X1 } ,
Properties
of Poisson where all the r.v. are independent. X2
Process
I By symmetry, it suffices to compute
Compound
Poisson
Process λ1 1
E [first term] = ·|· {z
· · ·}· = + .
Non- (λ1 + λ2 )2 λ1 + λ2
homogeneous using the lemma
Poisson
process

3
I Finally, E [T ] = .
λ1 + λ2
 §3.3 Poisson Process

STAT3603
Introductory Example: Phone Calls
Jeff Yao
t
The Poisson
Process

Interarrival 0 t
and Waiting I Consider the arrivals of telephone calls at a telephone exchange. We
Time Distri-
butions assume that we count the arrivals of calls, from the beginning of some
Conditional time and we view this beginning time as t = 0.
Distribution
of the
Arrival
Times
I Let Nt denote the number of calls arrived by time t.
Further
Properties I Then Nt is a random variable and all the possible values of this random
of Poisson
Process variable are {0, 1, 2, . . .}, i.e. non-negative integer valued random
Compound variable.
Poisson
Process

Non-
I Also for each t ≥ 0, we have a random variable Nt and thus we have a
homogeneous
Poisson
random process {Nt , t ≥ 0}.
process
For this process we see that
1 The time parameter t ∈ [0, ∞) is continuous;

2 The state space E = {0, 1, 2, . . .} is discrete.

 Counting process

STAT3603

Jeff Yao
A stochastic process {Nt : t ≥ 0} is called a counting process if
The Poisson
Process 1 For t ≥ 0, Nt has nonnegative integer values;
Interarrival
and Waiting
Time Distri- 2 For s < t, Nt − Ns equals the number of events that have occurred in the
butions
interval (s, t].
Conditional
Distribution
of the Note that Property ○
2 implies that t 7→ Nt is increasing: if s ≤ t, then
Arrival
Times Ns ≤ Nt .
Further
Properties
of Poisson
Process

Compound Examples of counting process:

Poisson
Process 1 Nt = # of phone calls in the time interval (0, t].
Non-
homogeneous
Poisson 2 Nt = # of claims in an insurance company in the time interval (0, t].
process

3 Nt = # of people who were born by time t.

 Two Important Concepts

STAT3603

Jeff Yao
Independent Increments

The Poisson
The numbers of events which occur in disjoint time intervals are independent.
Process
I Reasonable for Example ○.1 Phone calls are independently made by
Interarrival
and Waiting individuals.
Time Distri-
butions
I Not reasonable for Example ○.
3 The number of persons who are alive at
Conditional
Distribution t may affect the number of persons who will be born in (t, t + s].
of the
Arrival
Times

Further
Properties
of Poisson
Process
Stationary Increments
Compound Nt+u − Ns+u has the same distribution as Nt − Ns for all t > s > 0, u ≥ 0.
Poisson
Process
That is, the distribution of the number of events which occur in any interval
Non-
homogeneous of time depends only on the length of the interval.
Poisson
process I Reasonable for Example ○ 1 if we consider day as unit of time (not hour
as the unit of time).

I Not reasonable for Example ○

3 even the birth rate is kept to be constant.
 Definition of Poisson Process

STAT3603

Jeff Yao Definition. The counting process {Nt : t ≥ 0} is said to be a Poisson

The Poisson
process having rate λ, if
Process
1 N0 = 0;
Interarrival
and Waiting
Time Distri-
butions
2 The process has independent increments;
Conditional
Distribution 3 The number of events in any interval of length t is Poisson distributed
of the
Arrival with mean λt. That is, for all s, t ≥ 0,
Times

Further (λt)n
Properties P{Nt+s − Ns = n} = e −λt , n = 0, 1, . . .
of Poisson n!
Process

Compound (which is independent of s.)

Poisson
Process

Non-
homogeneous I It follows from ○
3 that for all t ≥ 0
Poisson
process
Nt = Nt − N0 ∼ P(λt) .

In particular, ENt = λt.

 A notation: Laudau’s o(·)

STAT3603

Jeff Yao Definition. A function f (x) is said to be o(g (x)) (small o of g (x)) in the
The Poisson
neighbourhood of a, denoted as x ∼ a, if
Process
f (x)
Interarrival
and Waiting
lim =0.
Time Distri-
x→a g (x)
butions

Conditional In other words, f (x) is much (infinitely) smaller than g (x) when x ∼ a.
Distribution
of the
Arrival
Times Examples. Most of time, g (x) = x n , a power of x.
Further I 4x 4 = o(x 3 ) when x ∼ 0 ;
Properties
of Poisson
Process
I log(x) = o(x) and x k = o(e x ), when x ∼ ∞.
Compound
Poisson
Process I Link to Taylor expansions:
Non-
1
homogeneous
Poisson
e x = 1 + x + x 2 + o(x 2 ) , x ∼ 0 ,
process
2
s s(s − 1) 2
(1 + x) = 1 + sx + x + o(x 2 ), x ∼0.
2
 Properties of Poisson process

STAT3603

Jeff Yao
Theorem 1. Let Nh be the number of points in (0, h] of a Poisson process
{Nt : t ≥ 0} with rate λ. Then for small h ∼ 0,
The Poisson
Process 1 P{Nh = 0} = 1 − λh + o(h);
Interarrival
and Waiting
Time Distri- 2 P{Nh = 1} = λh + o(h);
butions

Conditional
Distribution
3 P{Nh ≥ 2} = o(h).
of the
Arrival That is, on a interval of very small length h, there will be almost no points;
Times
the probability for 1 point is of order h and the one for at least 2 points is of
Further
Properties much smaller order o(h).
of Poisson
Process In other words, up to an error of o(h), Nh could be considered as a b(λh) r.v..
Compound
Poisson
Process Proof. Note that Nh ∼ P(λh).
Non-
homogeneous I P(Nh = 0) = e −λh = 1 − λh + 1
2!
(λh)2 + · · · = 1 − λh + o(h).
Poisson
process
I P(Nh = 1) = e −λh λh = λh + o(h).

I P(Nh ≥ 2) = 1 − [P(Nh = 0) + P(Nh = 1)] = 1 − [1 + o(h)] = o(h).

 Link with “Poisson approximation”

STAT3603

Jeff Yao Poisson approximation

The Poisson Theorem 2. Let X1 , . . . , Xn be a i.i.d. sequence of b(p) r.v’s and assume
Process
that p is related to n such that pn → λ > 0. Then when n → ∞, the
Interarrival
and Waiting distribution of Sn = X1 + · · · + Xn converges to P(λ).
Time Distri-
butions

Conditional
Distribution
of the
Proof. Use m.g.f’s.
Arrival
Times

Further
Properties
of Poisson
Process

Compound
Poisson
Process

Non-
homogeneous
Poisson
process
 Link with “Poisson approximation”: cont’d

STAT3603

Jeff Yao Now Consider a P.P. (Nt ) and for a fixed t > 0,
The Poisson
Process I a partition of (0, t] on k equal bins ((i − 1)h, ih] with bin-width h = t/k,
Interarrival 1 ≤ i ≤ k.
and Waiting
Time Distri-
butions

Conditional t 2t 3t kt
Distribution 0
of the k k k k
Arrival
Times

Further
Properties
I The bin-width h = t/k is small for large k, so that the counts in the bins
of Poisson
Process
Xi = Nih − N(i−1)h , 1 ≤ i ≤ k,
Compound
Poisson
Process could be considered as i.i.d. b(λh);
Non-
homogeneous
Poisson I Clearly, Nt = X1 + · · · + Xk ;
process

I Letting k → ∞ and k · λh = k · λt/k → λt, by the theorem,

Nt ∼ P(λt) .
 §3.4 Interarrival and Waiting Time Distributions

STAT3603 I let T1 denote the time of the first event occurred;

Jeff Yao
I Tn denote the elapsed time between the (n − 1)th and the nth event,
The Poisson
Process n > 1.
Interarrival
and Waiting
Time Distri-
I The sequence {Tn , n = 1, 2, . . . , } is called the sequence of interarrival
butions
times.
Conditional
Distribution E.g. If T1 = 5, T2 = 10 (for a particular trial), then the first event of the
of the
Arrival
Poisson process occurred at time 5 and the second at time 15.
Times

Further I Sn = T1 + . . . + Tn is the arrival time of the nth event, also called the
Properties
of Poisson waiting time until the nth event arrives.
Process

Compound
Poisson T1 T2 Tn
Process
1st 2nd (n-1)th nth
Non- S1
homogeneous
Poisson S2
process

Sn-1
Sn
STAT3603

Jeff Yao

The Poisson
Process

Interarrival
and Waiting
Time Distri-
butions
Theorem 3
Conditional For a Poisson point process with intensity parameter λ, the interarrival times
Distribution
of the Tn , n = 1, 2, . . . are i.i.d. exponential random variables with rate parameter λ.
Arrival
Times
Proof in 2 steps:
Further
Properties
of Poisson
1 use “small-h” argument to find the joint density of (S1 , . . . , Sn );
Process

Compound 2 by a change of variables, find the joint density of (T1 , . . . , Tn ) is found.

Poisson
Process

Non-
homogeneous
Poisson
process
STAT3603 Proof of Theorem 3: step 1
I For small enough hj > 0, 1 ≤ j ≤ n, it holds
Jeff Yao
0 < s1 < s1 + h1 < s2 < s2 + h2 < · · · < sn < sn + hn .
The Poisson
Process

Interarrival
and Waiting I It is enough to show that for small positive h1 , . . . , hn ∼ 0
Time Distri-
butions
P(s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn ) ∼ f (s1 , s2 , ..., sn ) × h1 h2 · · · hn .
Conditional
Distribution for some function f (s1 , . . . , sn )
of the
Arrival the event (s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn ) proba
Times
0 event in (0, s1 ], 1 event in (s1 , s1 + h1 ] e −λs1 · λh1 e −λh1
Further 0 event in (s1 + h1 , s2 ] 1 event in (s2 , s2 + h2 ] e −λ(s2 −s1 −h1 ) · λh2 e −λh2
Properties
of Poisson .
Process
.
.
−λ(sn −sn−1 −hn−1 )
Compound 0 event in in (sn−1 + hn−1 , sn ], 1 event in (sn , sn + hn ] e · λhn e −λhn
Poisson
Process

Non-
I
e −λsn e −λhn λn × h1 · · · hn
homogeneous
Poisson P(s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn ) =
process −λsn n
∼e λ × h1 · · · hn .
The joint density of (S1 , . . . , Sn ) is
f (s1 , . . . , sn ) = λn e −λsn · 1{0<s1 <···<sn } .
STAT3603

Jeff Yao
Proof of Theorem 3: step 2
The Poisson
Process I We have the joint density of (S1 , . . . , Sn ):
Interarrival
and Waiting f (s1 , . . . , sn ) = λn e −λsn 1{0<s1 <···<sn } .
Time Distri-
butions

Conditional
Distribution
of the
Arrival I Now consider the change of variables:
Times

Further
Properties
T1 = S1 , T2 = S2 − S1 , ..., Tn = Sn − Sn−1 .
of Poisson
Process
s The corresponding Jacobian is 1, so the joint density for (T1 , . . . , Tn ) is
Compound
Poisson
just
Process
n
Non-
Y
homogeneous g (t1 , . . . , tn ) = λn e −λ(t1 +···+tn ) 1{t1 >0,t2 >0,...,tn >0} = λe λtj 1{tj >0} .
Poisson
process j=1

The conclusion follows.

STAT3603

Jeff Yao

The Poisson
Process

Interarrival
and Waiting
Corollary. The r.v. Sn = T1 + · · · , +Tn , waiting time until the nth event
Time Distri- arrives, has a Gamma distribution γ(n, λ) with density function
butions

Conditional
(λt)n−1
Distribution
of the
fSn (t) = λe −λt , t≥0
Arrival
(n − 1)!
Times

Further
Properties Note. The m.g.f. of Sn is
of Poisson
Process  λ n
Compound E [e uSn ] = , u < λ.
Poisson λ−u
Process

Non-
homogeneous
Poisson
process
 Example 3.13

STAT3603

Jeff Yao

The Poisson
Process
Suppose that people immigrate into a territory at a Poisson rate λ = 1 per
Interarrival
and Waiting day.
Time Distri-
butions 1 What is the expected time until the 10th immigrant arrives?
Conditional
Distribution
of the
2 What is the probability that the elapsed time between the 10th and the
Arrival
Times
11th arrivals exceeds two days?
Further
Properties
of Poisson Solution:
Process

Compound 1 By independent increments E (S10 ) = E (T1 + · · · + T10 ) = 10/λ = 10

Poisson
Process
(days);
Non-
homogeneous 2 P{T11 > 2} = e −2λ = e −2 ≈ 0.1353.
Poisson
process
 §3.5 Conditional Distribution of the Arrival Times

STAT3603

Jeff Yao

The Poisson
Process

Interarrival
and Waiting
Time Distri- I Suppose that in the interval (0, t], one event has taken place, Nt = 1;
butions

Conditional
Distribution
I Want conditional distribution of the event time: T1 |Nt = 1;
of the
Arrival
Times I Of cause, this distribution has support on (0, t]: given Nt = 1, T1 ∈ (0, t].
Further
Properties
of Poisson
Process Theorem 4
Compound
Poisson
Conditional to Nt = 1, T1 is distributed uniformly on (0, t].
Process

Non-
homogeneous
Poisson
process
 Proof of Theorem 4 (by the h-method

STAT3603 Looking for g (s) = f (s|Nt = 1), conditional density function of T1 , such that
Jeff Yao for small h ∼ 0 and 0 < s < t,
The Poisson
Process
P(s < T1 ≤ s + h | Nt = 1) ∼ h · g (s).
Interarrival
and Waiting
Time Distri- the event (s < T1 ≤ s + h, Nt = 1) : proba
butions

Conditional * 0 event in (0, s] e −λs

Distribution
of the * 1 event in (s, s + h] ∼ λhe −λh
Arrival
Times
* 0 event in (s + h, t], e −λ(t−(s+h))
Further
Properties Therefore,
of Poisson
Process
P(s < T1 ≤ s + h, Nt = 1) ∼ λhe −λt ,
Compound
Poisson
Process and
Non-
λhe −λt h
homogeneous
P(s < T1 ≤ s + h | Nt = 1) ∼ = .
Poisson
process
(λt)e −λt t
That is
1
g (s) = 1{0<s<t} .
t
 The general case

STAT3603

Jeff Yao Theorem 5 Given Nt = n, the joint distribution of the n arrival (waiting)
The Poisson
times S1 , S2 , . . . , Sn is the same as that of the order statistics from a sample of
Process n i.i.d. random variables uniformly distributed over the interval (0, t), i.e. the
Interarrival density function is
and Waiting
Time Distri-
butions n!
f (s1 , . . . , sn |Nt = n) = 1{0<s1 <···<sn <t} .
Conditional tn
Distribution
of the
Arrival
Times I Recall that the order statistics (X(1) , . . . , X(n) ) from a sample X1 , . . . , Xn
Further where Xj ’s have a common density function f (x), has the joint density
Properties
of Poisson function
Process

Compound n!f (x1 ) · · · f (xn )1{x1 <···<xn } .

Poisson
Process

Non-
homogeneous
Poisson
process
I Example: the joint distribution of (min(X , Y ), max(X , Y )) of two i.i.d.
exponential E(λ) has the density function

2λ2 e −λ(x+y ) 1{0<x<y } .

 Proof of Theorem 5 (by the h-method)

STAT3603 I Looking for g (s1 , ..., sn ) such that for small positive h1 , . . . , hn ∼ 0
Jeff Yao
P(s1 < S1 ≤ s1 +h1 , . . . , sn < Sn ≤ sn +hn | Nt = n) ∼ g (s1 , s2 , ..., sn )×h1 h2 · · · hn .
The Poisson
Process I The joint event
Interarrival
and Waiting (s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn ) ∩ (Nt = n)
Time Distri-
butions = (s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn ) ∩ {0 event in (sn + hn , t]}
Conditional
Distribution
of the I Independent events and by Step 1 in the proof of Theorem 3:
Arrival
Times P (s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn , Nt = n)
Further = P (s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn ) P(0 event in (sn + hn , t])
Properties
of Poisson
Process
= e −λsn e −λhn λn × h1 · · · hn · e −λ(t−sn −hn )
Compound = e −λt λn × h1 · · · hn .
Poisson
Process
I Therefore,
Non-
homogeneous
Poisson P(s1 < S1 ≤ s1 + h1 , . . . , sn < Sn ≤ sn + hn | Nt = n)
process
e −λt λn × h1 · · · hn n!
∼ (λt)n
= × h1 · · · hn .
e −λt n! tn
n!
This implies that g (s1 , . . . , sn ) = , 0 < s1 < · · · < sn < t.
tn
 Example 3.21. Insurance claims

STAT3603
I Insurance claims arrive at P.P. times with rate λ;
Jeff Yao

The Poisson
I the successive claim amounts are independent r.v. having distribution G
Process
with mean µ, and are independent of the claim arrival times;
Interarrival
and Waiting
Time Distri- I Let Si and Ci denote, respectively, the time and the amount of the ith
butions
claim;
Conditional
Distribution
of the
Arrival
I the total discounted cost of all claims made up to time t, call it D(t), is
Times defined by
Further XNt
e −αSi Ci ,
Properties
of Poisson Dt =
Process
i=1
Compound
Poisson where α is the discount rate and Nt is the number of claims made by
Process
time t.
Non-
homogeneous
Poisson
process
Question: Find E (Dt ).
Solution:
λµ
(1 − e −αt ).
α
STAT3603
Claim Amount C1 C2 Cn-1 Cn
Jeff Yao
t
The Poisson
Process
S1
Interarrival
S2
and Waiting
Time Distri-
butions
Sn-1
Conditional
Distribution Sn
of the
Arrival
Times

Further
Main step:
Properties
n
!
of Poisson X
Process E [Dt |Nt = n] = µE e α X(i) |Nt = n
Compound i=1
Poisson
µn
Process
= µnE [e −αX1 ] = (1 − e −αt ) ,
Non- αt
homogeneous
Poisson where (X(i) ) are order statistics of uniform dist. on (0, t].
process
 One more conditional result

STAT3603

Jeff Yao
Theorem 6 Given that Sn = t, the n − 1 arrival times S1 , . . . , Sn−1 have the
The Poisson
Process
same distribution as the order statistics from a sample of n − 1 i.i.d. r.v. of
Interarrival
uniform distribution on (0, t).
and Waiting
Time Distri-
butions Proof: (Similar to the proof of Theorem 5)
Conditional
Distribution
of the
Looking for g (s1 , ..., sn−1 ) such that for small positive h1 , . . . , hn−1 ∼ 0, and
Arrival 0 < s1 < · · · < sn−1 < t = sn ,
Times

Further
Properties
P(s1 < S1 ≤ s1 + h1 , . . . , sn−1 < Sn−1 ≤ sn−1 + hn−1 | Sn = t)
of Poisson
Process ∼ g (s1 , s2 , ..., sn−1 ) × h1 h2 · · · hn−1 .
Compound
Poisson
Process

Non-
This leads to
homogeneous
Poisson (n − 1)!
process g (s1 , . . . , sn−1 ) = , 0 < s1 < · · · < sn−1 < t = sn .
t n−1
STAT3603

Jeff Yao Part IV

Normal
Distribution

BM Brownian Motion and Applications

Hitting
Time &
Maximum
Variable
1 Normal Distribution
Variations
on BM

Pricing
2 Brownian Motion
Stock
Options
3 Hitting Time, Maximum Variable, and the Gambler’s Ruin Problem

4 Variations on Brownian Motion

(i) Brownian Motion with Drift

(ii) Geometric Brownian Motion

5 Pricing Stock Options∗

STAT3603 Review
Jeff Yao

Review