Math 597/697: Homework 6
1. (a) Let X be normal random variable with mean µ and σ 2 . Let
Y = aX + b and a > 0
t−b
P {Y ≤ t} = P {aX + b ≤ t} = P {X ≤ }
a
(x−µ)2 (y−b−µ)2
e− e− 2a2 σ2
t−b
Z
a 2σ 2
Z t
= √ dx = √ dy (1)
−∞ 2πσ 2 −∞ 2πa2 σ 2
where we have used the change of variables y = ax + b in the last
equality. So the mean of Y is µ + b and the variance a2 σ 2 . The
case a < 0 is similar.
(b) Let X1 and X2 be independent normal random variable with
mean 0 and variance σi2 , i = 1, 2. Since X2 has mean zero and
the distribution of a normal RV is symmetric with respect to
x 7→ −x we have X2 = −X2 , meaning that they have the same
density. If X1 and X2 are independent so are X1 and −X2 and
so X1 + X2 = X1 − X2 . This is again a normal R.V with mean 0
and variance σ12 + σ22 (this is seen most easily by considering the
moment generating function).
2. Suppose Xt is a standard Brownian motion and let Yt = a−1/2 Xat with
a > 0. We have
(a) Y0 = X0 = 0
(b) Yt − Ys = a−1/2 (Xat − Xas ) so if Xt has independent increments
so does Yt .
(c) Xat − Xas is a normal RV with mean 0 and variance a(t − s). By
problem (1) Yt −Ys = a−1/2 (Xat −Xas ) has variance a−1 a(t−s) =
(t − s).
(d) If t 7→ Xt is continuous, so is t 7→ Yt .
Therefore Yt is a standard brownian motion.
3. Let Xt and Yt be independent standard one-dimensional Brownian
motion. Let Zt = Xt − Yt
(a) We have
i. Z0 = X0 − Y0 = 0
1
ii. Zt − Zs = (Xt − Xs ) − (Yt − Ys ) and so Zt has independent
increments since Xt and Yt do.
iii. By (1) (b) Zt − Zs is a normal random variable with param-
eter 2(t − s)
iv. The map t 7→ Yt is continuous.
So Zt is a Brownian motion with variance parameter 2.
(b) Brownian motion is recurrent so that P {Zt = 0 i.o.} = 1. There-
fore P {Xt = Yt i.o.} = 1.
4. Let Xt be a Brownian motion with variance parameter σ 2 . Let t ≥ s
then by the independent increments property
E[Xs Xt ] = E[Xs (Xt − Xs )] + E[(Xs )2 ]
= E[Xs ]E[(Xt − Xs )] + E[(Xs )2 ]
= E[(Xs )2 ] = s. (2)
Similarly if s ≥ t one finds E[Xs Xt ] = t. Therefore
E[Xs Xt ] = min(s, t) . (3)
5. Let Xt be a standard Brownian motion and let us define, for α > 0,
Vt = e−αt/2 X(eαt ) . (4)
The process Vt is called an Ornstein-Uhlenbeck process. We have,
using (5)
E[Vt ] = e−αt/2 E[X(eαt )] = 0 . (5)
For s ≤ t
E[Vs Vt ] = e−αs/2 e−αt/2 E[X(eαs )X(eαt )]
= e−αs/2 e−αt/2 eαs = e−α(t−s)/2 (6)
For t ≤ s one finds
E[Vs Vt ] = = e+α(t−s)/2 (7)
and so
E[Vs Vt ] = = e−α|t−s|/2 . (8)
2
6. The motion of an oil spill (assume it is pointwise) on the surface of the
ocean is a 2-dimensional Brownian motion with variance parameter
σ 2 = 1/2. Let C be a square with side lengths 2 and assume that
the oil spill, at time 0, is at the center of the square C. We assume
that 0 is the center of the square and so the motion of the oil spill is
(1) (2)
a two dimensional Brownian Xt = (Xt , Xt ) motion with variance
parameter 1/2. The desired probability PS is
(1) (2)
PS = P {|Xt | ≥ 1 or |Xt | ≥ 1 for some 0 ≤ t ≤ 1} (9)
Using the Reflection principle this can be estimated by
(1)
PS ≤ P {|Xt | ≥ 1 for some 0 ≤ t ≤ 1}
(2)
+ P {|Xt | ≥ 1 for some 0 ≤ t ≤ 1}
2
∞ e−x
Z
(1)
≤ 4P {X1 ≥ 1} = 8 √ (10)
1 π
7. Let Xt be a standard Brownian motion. Let M be the random variable
given by
M = sup Xt (11)
0≤t≤1
It is the maximum of Xt on the time interval [0, 1]. We have, by the
reflection principle,
( )
P {M ≥ a} = P sup Xt ≥ a = P {Xt ≥ a for some 0 ≤ t ≤ 1}
0≤t≤1
x2
∞ e− 2
Z
= 2P {X1 ≥ a} = 2 √ (12)
a 2π
Hence the cumulative distribution function of M is
x2
a e− 2
Z
P {M < a} = 2 √ (13)
0 2π
and its density is
x2
e− 2
f (a) = 2 √ a ≥ 0. (14)
2π
8. Let Xt be a Brownian motion with drift parameter µ and variance
parameter σ 2 .
3
(a) Since Xt = Yt +µt where Yt is a Brownian motion with parameter
σ2
E[Xs Xt ] = µ2 ts + σ 2 min(s, t) (15)
(b) We expand g(x) in a taylor series around Xt
1 1
= lim E g 0 (Xt )(Xt+h − Xt ) + g”(Xt )(Xt+h − Xt )2 +
h→0 h 2
i
+o((Xt+h − Xt )2 ) | X0 = x (16)
By the independent increments property, for any function h(x),
h(Xt ) is independent of Xt+h − Xt . We have
E[Xt+h − Xt | X0 = x] = hµ . (17)
and, using that, Xt = µt + Yt
h i
E (Xt+h − Xt )2 | X0 = x
h i
= E (µh + (Yt+h − Yt ))2 | Y0 = x
= µ2 h2 + σ 2 h (18)
and we have h i
E o((Xt+h − Xt )2 ) = o(h) . (19)
Therefore we have
d 1
E g 0 (Xt | X0 = x µh
g(t, x) = lim
dt h→0 h
1
+ E g 00 (Xt | X0 = x (µ2 h2 + h) + o(h)
2
∂g σ2 ∂ 2g
= µ (t, x) + (t, x) (20)
∂x 2 ∂x2
