0% found this document useful (0 votes)
162 views4 pages

2 Brownian Motion: T T T T T

1. The process Ut = Z√t, where Z is a standard normal random variable, has the same distribution as Brownian motion Bt at any time t. However, Ut is not a Brownian motion because the increments of Ut are not independent. 2. Brownian motion can be simulated in Excel by generating independent standard normal random variables and taking cumulative sums of their values scaled by time increments. 3. For the process Vt = 2Bt + 3t, where Bt is Brownian motion, the mean function is E(Vt) = 3t and the covariance function is Cov(Vs,Vt) = 4 min(s,t).

Uploaded by

will
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
162 views4 pages

2 Brownian Motion: T T T T T

1. The process Ut = Z√t, where Z is a standard normal random variable, has the same distribution as Brownian motion Bt at any time t. However, Ut is not a Brownian motion because the increments of Ut are not independent. 2. Brownian motion can be simulated in Excel by generating independent standard normal random variables and taking cumulative sums of their values scaled by time increments. 3. For the process Vt = 2Bt + 3t, where Bt is Brownian motion, the mean function is E(Vt) = 3t and the covariance function is Cov(Vs,Vt) = 4 min(s,t).

Uploaded by

will
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

2 Brownian motion

1. Let Z denote a standard


Normal random variable. Give the distribution
of the process Ut = Z t. Show that for any time t, the distribution of
Ut is the same as that of the Brownian motion Bt , but the process Ut is
not a Brownian motion. Sketch a graph of Ut as a function of t.
Solution
Using that aZ N (0, a2 ), Ut N (0, t). This is the same distribution
as B(t) for any fixed t. But increments of the process U (t)are not
independent: for t1< t2< t3 < t4 , U (t2 ) U (t1 ) = Z( t2 t1 ) and
U (t4 )U (t3 ) = Z( t4 t3 ), are clearly dependent on the same random
variable Z.
Alternatively,

Cov(U (t4 ) U (t3 ), U (t2 ) U (t1 )) = Cov(Z( t4 t3 ), Z( t2 t1 ))

= ( t4 t3 )( t2 t1 )V ar(Z) 6= 0,
thus the increments are correlated and therefore not independent.

2. Using Excel, generate the values of Brownian motion at points ti = i/100,


i = 0, . . . 100 and plot them. Hint: use the fact that increments of Brow-
nian motion are independent and Normally distributed, B(ti+1 )B(ti ) =

Zi ti+1 ti , where Zi is a sequence of standard Normal variables.
Solution
In the first cell of A column write =rand(). This gives U (0, 1) r.v. In
the first cell of B column write =Normsinv(A1), where A1 is the first
cell of A to get N (0, 1) r.v. In C1 use =B1*0.1 to have N (0, 0.01).

Note ti+1 ti = 0.1. Simulation of the Brownian motion is given by
the cumulative sums of the C column. Enter 0 in D1, then in D2 write
=D1+C1. Dragging down 100 places gives the path.

3. Define Vt = 2Bt + 3t, where Bt is a Brownian motion. Find the mean


and covariance functions of the process Vt .
Solution
The mean function:

E(Vt ) = E(2Bt + 3t) = 2E(Bt ) + 3t = 2 0 + 3t = 3t.

The covariance function:

(s, t) = Cov(Vs , Vt )
= Cov(2Bs + 3s, 2Bt + 3t)
= 4Cov(Bs , Bt ) + 6Cov(Bs , t) + 6Cov(s, Bt ) + 9Cov(s, t)
= 4 min(s, t) + 6 0 + 6 0 + 9 0
= 4 min(s, t)

where Cov(Bs , Bt ) = min(s, t) by Theorem 8 on p.14, and the other


terms are zero as covariance of a random variable with a non-random
number.

4. Bt is a Brownian motion.

(a) Show that the process Wt = Bt is also a Brownian motion.


(b) Show that the process Ut = 1 B2t is also a Brownian motion.
2
(c) Shwo that the process Vt = Bt+1 Bt is a Brownian Motion

Solution
(a)
Independence of increments: for s < t

Wt Ws = Bt (Bs ) = Bt + Bs = (Bt Bs ).

It is independent of Bu , u s. Hence also of Bu = Wu .


Since Bt Bs N (0, t s), and N (0, 2 ) = N (0, 2 ), Wt Ws
N (0, t s) = N (0, t s).
As Bt is continuous, so is Bt = Wt .
(b)
We want to show that Ut Us is independent of Uu with 0 u s t.
We can express Ut Us and Uu in terms of Bt Bs and Bu :
1 1
Ut Us = B2t B2s
2 2
1
= (B2t B2s )
2
B2t B2s is the increment of Brownian motion over time interval [2s, 2t]
therefore it is independent of any value of Brownian in at the time preced-
ing 2s. Since for u < s, 2u < 2s, it follows that B2t B2s is independent
of B2u .
For the Gaussian increments, Ut Us = 12 (B2t B2s ). We know that
Bt Bs N (0, t s), so B2t B2s N (0, 2t 2s). But we also have if
X N (, 2 ) then aX N (a, a2 2 ), so with a = 12 and 2 = 2t 2s,
we obtain Ut Us N (0, t s) as desired.
For the continuity, note that B2t is the composition of two continuous
functions: the function Bt and the function 2t. Because the composition
of continuous functions is again continuous and multiplying a continuous
function by a constant gives a continuous function, we have Ut = 12 B2t
is continuous.
(c)
Vt Vs = Bt+1 Bs+1 is independent of Vs = Bs+1 B1 . From the
properties of Brownian Motion, Bt+1 Bs+1 is independent of Bu for
u s + 1. So Bt+1 Bs+1 is independent of B1 and independent of Bs+1 .
Hence it is independent of Bs+1 B1 .
For the Gaussian increments, Vt Vs = Bt+1 Bs+1 N (0, (t+1)(s+1))
so Vt Vs N (0, t s).
Finally, for a continuous function f (x) and constants a and b, f (x+a)+b
is also continuous. So because Bt is continuous and B1 a constant (in
the sense that for a given path once B1 is observed its value does not
change) we have that Vt = Bt+1 B1 is continuous too.

5. Let B1 (t) and B2 (t) be independent


Brownian motions, and define
W (t) = (B1 (t) + B2 (t))/ 2.

(a) Show that W (t) is also a Brownian motion.


(b) Find the covariance between W (t) and B1 (t).

Solution (a)
First independence of increments. We can express W (t) W (s) in terms
of the independent Brownian motions:
1 1
W (t) W (s) = (B1 (t) + B2 (t)) (B1 (s) + B2 (s))
2 2
1 1
= (B1 (t) B1 (s)) + (B2 (t) B2 (s))
2 2
and W (u) = 12 B1 (u) + 12 B2 (u). By independence of increments of
each Brownian motion, B1 (t) B1 (s) is independent of B1 (u) for u s.
Since Brownian motions are independent of each other B1 (t) B1 (s) is
independent of B2 (u) (for any u). Hence B1 (t) B1 (s) is independent
of W (u). Similarly B2 (t) B2 (s) is independent of W (u). Therefore
W (t) W (s) is independent of W (u).
For the Gaussian increments, we have that B1 (t)B1 (s) and B2 (t)B2 (s)
have N (0, t s) distribution. Because they are independent their sum
has N (0, 2(t s)) distribution. Hence W (t) W (s) has 12 N (0, 2(t s))
distribution, which is the same as N (0, t s) as desired.
Finally, we note that the sum of continuous functions is continuous and
that a continous function multiplied by a constant is continuous. So
W (t) is continuous.
(b)
We use property of bilinearity of covariance
1
Cov(B1 (t), W (t)) = Cov(B1 (t), (B1 (t) + B2 (t)))
2
1 1
= Cov(B1 (t), B1 (t)) + Cov(B1 (t), B2 (t))
2 2
1
= V ar(B1 (t)) + 0 (due to independence of B1 , B2 )
2
1
= t + 0 (because B1 (t) N (0, t)).
2

6. Show that the process Bt3 is not a Brownian Motion


Solution
There are a number of ways of showing this, here we look at showing
that the independence of increments does not hold. Alternatively one
can show that the power 3 of a Normal distribution is not a Normal
distribution.
We have algebraic decomposition:

a3 b3 = (a b)(a2 + ab + b2 )

therefore the increment of the process B 3 between times s and t is

Bt3 Bs3 = (Bt Bs )(Bt2 + Bt Bs + Bs2 ).

Although Bt Bs is independent of Bs (and hence Bs3 ), Bt2 , Bt Bs and


Bs2 are not independent of Bs3 . Hence Bt3 Bs3 is not independent of Bs3 .

You might also like