2018 Fall MMAT 5340 1

THE CHINESE UNIVERSITY OF HONG KONG

Department of Mathematics
MMAT 5340 Probability and Stochastic Analysis
Suggested Solution of Homework 2
 
11.29 Let x(n) = P (Xn = 1) P (Xn = 2) . Then

 0.4 0.6 2 
 
2
 
x(2) = x(0)A = 0.4 0.6 = 0.334 0.666 .
0.3 0.7

Hence P (X2 = 1) = 0.334.

 
11.30 The stationary distribution p = p1 p2 satisfies

p1 = 0.4p1 + 0.3p2
p = pA ⇐⇒ ⇐⇒ 2p1 = p2 .
p2 = 0.6p1 + 0.7p2

Since p1 + p2 = 1, we have p = 31 23 .
 

11.31 Solving  
 0.4 − λ 0.6 
0 = |A − λI| =   = λ2 − 1.1λ + 0.1,
0.3 0.7 − λ 
the eigenvalues of A are given by
1
λ1 = 1, λ2 = .
10
1
Hence the rate of convergence is |λ2 |n = .
10n
11.32 Let px be the probability that, starting from x, the process hits 2 before 3. Then

p2 = 1, p3 = 0,

and

p1 = 0.4p1 + 0.4p2 + 0.2p3

= 0.4p1 + 0.4 + 0,

2
so that p1 = .
3
 2018 Fall MMAT 5340 2

11.36 The graph of the Markov chain is given below:

0.4 0.5

0.2

1 2
0.2

0.3

0.1 0.3

1
3 4
0.5

0.5

From the graph, we can see that 1 and 2 are transient states, while 3 and 4 are
recurrent states.
11.40 Let X = Z − 1. Then
P (X = 0) = P (Z = 1) = 0.5, P (X = 1) = P (Z = 2) = 0.52 = 0.25,
P (X ≥ 2) = 1 − P (X = 0) − P (X = 2) = 0.25
Hence, the transition matrix is
 
0.5 0.5 0
P = 0.5 0.25 0.25
0 0.5 0.5
 
We can find the stationary distribution p = p0 p1 p2 by solving pP = p: By
performing column operations,
   
−0.5 0.5 0 −0.5 0 0
P − I = 0.5 −0.75 0.25 ∼ 0.5
   −0.25 0.25 
0 0.5 −0.5 0 0.5 −0.5
   
−0.5 0 0 1 0 0
∼ 0.5 −0.25 0 ∼ −1 1
   0 .
0 0.5 0 0 −2 0
 2018 Fall MMAT 5340 3

 
That is p0 = p1 = 2p2 . Hence p = 0.4 0.4 0.2 .
The long-term average premium is
p0 r0 + p1 r1 + p2 r2 = (0.4)(0.5 · 0 + 1) + (0.4)(0.5 · 1 + 1) + (0.2)(0.5 · 2 + 1)
= 1.4 thousands of dollars.

12.12 Let n and m be the quantity of up and down steps. Then

(
n + m = 13 − 2 = 11
n − m = 3 − 2 = 1.
Hence n = 6, m = 5. Thus
 
11
P (S13 = 3|S2 = 2) = (0.7)6 (0.3)5 ≈ 0.1321.
6
12.13 By the reflection principle, every path from (2, 2) to (13, 3) which hits y = 1 corre-
sponds to a path from (2, 2) to (13, −1). Solving
(
n + m = 13 − 2 = 11
n − m = −1 − 2 = −3,
we have n = 4, m = 7. The number of such path is 11


4
. Thus
   
11 11
P (S13 = 3, Sn > 1, n = 2, 3, . . . , 13|P2 = 2) = − (0.7)6 (0.3)5 ≈ 0.0377.
6 4
12.14 Solving ( (
n + m = 10 − 0 = 10 n=6
=⇒
n − m = 2 − 0 = 2, m = 4.
So there are 10


6
paths from (0, 0) to (10, 2). By the reflection principle, every
path from (0, 0) to (10, 2) which hits y = −2 corresponds to a path from (0, 0) to
(10, −2 − (2 − (−2))) = (10, −6). Solving
( (
n + m = 10 − 0 = 10 n=2
=⇒
n − m = −6 − 0 = −6, m = 8.
So there are 10


2
paths from (0, 0) to (10, −6). Thus,
   
10 10
P (S10 = 2, Sn > −2, n = 0, . . . , 10) = − (0.5)10 ≈ 0.1611.
6 2
12.27 The risk-neutral probability p0 , q0 can be found by E0 P1 = P0 :
(
p0 · 2 + q0 · 0.3 = 1 7 10
=⇒ p0 = , q0 = .
p0 + q 0 = 1 17 17
Hence, the fair price is given by
v = E0 (P3 − K)+
= p30 (23 − 1) + 3p20 q0 (22 · 0.3 − 1) + 3p0 q02 (0) + q03 (0)
≈ 0.5485.
 2018 Fall MMAT 5340 4

13.20 Since Xn is the number of Heads during the first n tosses, we have
1 1 1
E(Xn+1 | Xn ) = (Xn + 1) + Xn = Xn + .
2 2 2
If Yn := 3Xn − cn is a martingale, then

Yn = E(Yn+1 | Y0 , . . . , Yn ) = E(Yn+1 |X0 , . . . , Xn )

= E(3Xn+1 − c(n + 1) | Xn )
= 3E(Xn+1 | Xn ) − c(n + 1)
3
= 3Xn + − c(n + 1)
2
3
= Yn + − c.
2
3
Hence c = .
2
13.30 Need to find c such that Mn = eSn −cn is a martingale:

Ee−1+X1 −c = e−1 =⇒ e4/2 = FN (0,4) (1) = ec =⇒ c = 2.

For x > 0, f (x) = x3 is convex, since f 00 (x) = 6x > 0. Note that Mn > 0. By
Doob’s martingale inequality, for λ > 0, we have,
3
 
EM100
P max Mn ≥ λ ≤ .
0≤n≤100 λ3

Now
3
M100 = e−3 e3S100 e−300c = e−3 e−300c e3X1 · · · e3X100 ,
so that
 100
(3)2 4
3 −3 −300c 3X1 100 −3 −300(2)
EM100 =e e (Ee ) =e e e 2 = e1197 .

Hence
e1197
 
P max Mn ≥ λ ≤ .
0≤n≤100 λ3

14.12 Note that τ4 − τ3 and τ3 − τ2 are i.i.d. Exp(λ) random variables, with λ = 3. Hence

τ4 − τ2 = (τ4 − τ3 ) + (τ3 − τ2 ) ∼ Γ(2, λ) = Γ(2, 3).

Therefore,
2
E(τ4 − τ2 ) = 2λ−1 = ,
3
2
Var(τ4 − τ2 ) = 2λ−2 = .
9
 2018 Fall MMAT 5340 5

14.13 Since N (t) − N (s) is independent of N (u), u ≤ s, and N (t) − N (s) ∼ Poi(λ(t − s)),
we have

P(N (5/2) = 3 | N (1) = 1) = P(N (5/2) − N (1) = 2 | N (1) = 1)

= P(N (5/2) − N (1) = 2)
(λ(3/2))2 −λ(3/2)
= e
2!
81
= e−9/2 .
8

14.18 Note that EN (t) = VarN (t) = λt = t. By (35) and (36),

EX(t) = EN (t) · EZk = t · 2 = 2t,

VarX(t) = EN (t) · VarZk + VarN (t) · (EZk )2 = t · 3 + t · 22 = 7t.

 
15.24 The stationary distribution ρ = ρ1 ρ2 ρ3 satisfies ρA = 0. Since ρ1 +ρ2 +ρ3 = 1,
we have
1
1 5 2 = 81 58 41 .
   
ρ=
1+5+2
15.25 Since 0 > λ2 > λ3 , the rate of convergence is eλ2 t = e−2t .

15.26 The transition matrix for the corresponding discrete-time Markov chain is
 1 1
0 2 2
P = 0 0 1 .

1 2
3 3
0

15.27 Let λ0i , 1 ≤ i ≤ 3, be the intensity of exit from state i, that is, the negative of
the diagonal entries of A. Then the stationary distribution π for the corresponding
discrete-time Markov chain is given by
1  0
λ1 ρ1 λ02 ρ2 λ03 ρ3

π=
λ01 ρ1
+ λ02 ρ2 + λ03 ρ3
8 1 5 3

= 4 8 4
132 5 6

= 13 13 13
.