Characteristic function
The characteristic function is defined for all random variables
Univariate distributions:
Let X be a random variable with p.d.f f(x). The characteristic function (c.f) of X is
(t ) E e f ( x) e
itX itx 2
t , i 1
x
f ( x) eitx dx t , i 2 1
x
d (t ) d itx d itx
(t ) f ( x) e e f ( x) ixe f ( x)
itx
dt dt x x dt x
d d itx
f ( x) e dx e f ( x) dx ixe f ( x) dx
itx itx
dt x x dt x
Substituting t=0
(0) i xf ( x) iE[ X ]
x
i x f ( x) dx iE[ X ]
x
d 2 (t ) d itx
dt 2
(t )
dt x
i x f ( x ) e
x
i 2 2 tx
x e f ( x)
d
i xf ( x) e dx i x e f ( x ) dx
itx 2 2 tx
dt x x
Substituting t=0
(0) i 2 x 2 f ( x) i 2 E[ X 2 ]
x
i x f ( x ) dx i E[ X ]
2 2 2 2
x
(0)
E[ X ]
i
2 (0)
E[ X ] 2
i
2
(0) (0)
Var[ X ] 2 i
i
Example 1:
Find the characteristic function of a random variable X whose p.d.f is
Pr( X x) p(1 p) x x 0,1, 2,...
Hence find the mean and variance of X.
(t ) eitx p (1 p ) x p [eit (1 p )]x
x 0 x 0
p
1 (1 p )eit
d p ip (1 p )eit
(t ) it
dt 1 (1 p )e (1 (1 p)eit ) 2
ip (1 p )ei 0 p (1 p )
(0) i0 2
(1 (1 p)e ) (1 (1 p)) 2
p(1 p) i (1 p)
2
p p
(0) (1 p)
E[ X ]
i p
d 2
p d ip(1 p)e
it
(t ) 2 it
it 2
dt 1 (1 p)e dt (1 (1 p)e )
(1 (1 p)eit ) 2 ieit eit 2i (1 p)(1 (1 p)eit )
ip(1 p) it 4
(1 (1 p)e )
p 2 i 2i (1 p ) p
(0) ip (1 p ) 4
p
i 2 (1 p ) 2i 2 (1 p) 2
p p2
(0) (1 p ) 2(1 p ) 2
E[ X 2 ] 2
i p p2
2 2
(1 p ) 2(1 p) (1 p)
Var[ X ] 2
p p p
(1 p )
p2
Example 2
Find the variance of a random variable Y whose characteristic function is
1
(t )
it
1
d it
(t ) 1
dt
1
it i
1
i
(0)
(0)
E[Y ]
i
1
d2 it d i it
(t ) 2 1 1
dt dt
2
i it i
( 1) 1
( 1)i 2
(0)
2
2 (0) ( 1)
E[Y ] 2
i 2
2
( 1)
Var[Y ] 2
2
Example 3
Find the mean and variance of a random variable X whose characteristic
function is
(t ) e (1 eit )
exp [1 e ]
it
d d (1 eit ) (1 e it )
(t ) e e ieit
dt dt
2
d d it (1 eit )
2
(t ) ie e
dt dt
it (1 eit ) it (1 e it )
ie e ie ie ieit
(1 ei 0 )
(0) e iei 0 i
(1 ei 0 ) (1 ei 0 )
(0) ie e i0 i0
ie ie ie i 0
2 i 2 i 2
(0) i
E[ X ]
i i
2
(0) (0) 2 2
i i 2
2
Var[ X ] 2 2
i i i
Bivariate distributions:
Let X and Y be two jointly distributed random variables with
j.p.d.f f(x , y). The joint characteristic function of X and Y is
(t1 , t2 ) E[eit1 X eit2Y ] f ( x, y )eit1 x it2 y (t1 , t2 ) 2
y x
eit1 x it2 y f ( x, y ) dxdy (t1 , t2 ) 2
y x
Using the j.c.f ,
(i) we can define marginal c.f of each variable
marginal c.f of X= X (t1 ) (t1 , 0)
marginal c.f of Y=Y (t2 ) (0, t2 )
(ii) We can obtain the mean, variance and covariance values of the two
variables
d d
(t1 , t2 ) (t1 , t2 )
dt1 t 0, t
dt2
2 0 t 0, t 2 0
E[ X ] 1
E[Y ] 1
i i
d2 d2
2
(t1 , t2 ) 2
(t1 , t2 )
dt1 dt 2
2 t1 0, t2 0 2 t1 0, t2 0
E[ X ] 2
E[Y ]
i i2
d d d d
(t1 , t2 ) (t1 , t2 )
dt1 dt2 t1 0, t2 0
dt2 dt1 t1 0, t2 0
E[ XY ] 2
i i2
Example 4
Two continuous random variables X and Y have joint characteristic
function
t1 , t2 exp 0.78it1 3.93it2 1.1t12 1.05t1 t2 2.78t22
Find the correlation coefficient between the two random variables
d d
t1 , t2 exp 0.78it1 3.93it2 1.1t12 1.05t1 t2 2.78t22
dt1 dt1
exp 0.78it1 3.93it2 1.1t12 1.05t1t2 2.78t22 (0.78i 2.2t1 1.05t2 )
d
t1 , t2 0.78i
dt1 t 0, t
1 2 0
d
t1 , t2
dt1 t 0, t 2 0
E[ X ] 1
0.78
I
d d
t1 , t2 exp 0.78it1 3.93it2 1.1t12 1.05t1 t2 2.78t22
dt2 dt2
exp 0.78it1 3.93it2 1.1t12 1.05t1t2 2.78t22 (3.93i 5.56t2 1.05t1 )
d
t1 , t2 3.93i
dt2 t 0, t
1 2 0
d
t1 , t2
dt2 t 0, t 2 0
E[Y ] 1
3.93
i
d2 d
2
dt1
t ,
1 2t
dt1
(0.78i 2.2t1 1.05t 2 ) exp 0.78it1 3.93it 2 1.1t1
2
1.05t t
1 2 2.78t 2
2
exp 0.78it1 3.93it2 1.1t12 1.05t1t 2 2.78t 22 (0.78i 2.2t1 1.05t 2 ) 2
2.2 exp 0.78it1 3.93it2 1.1t12 1.05t1t 2 2.78t 22
d2
2
t1 , t2 0.782 i 2 2.2 0.782 i 2 2.2i 2
dt1
t1 0, t2 0
d2
2
t1 , t2
dt1
t1 0, t2 0
E[ X 2 ] 2
2.8084
i
d2 d
2
dt 2
t ,
1 2t
dt2
(3.93i 5.56t 2 1.05t1 ) exp 0.78it1 3.93it 2 1.1t1
2
1.05t t
1 2 2.78t 2
2
exp 0.78it1 3.93it2 1.1t12 1.05t1t2 2.78t 22 (3.93i 5.56t 2 1.05t1 ) 2
5.56 exp 0.78it1 3.93it2 1.1t12 1.05t1t2 2.78t 22
d2
2
t1 , t2 3.932 i 2 5.56 3.932 i 2 5.56i 2
dt 2
t1 0, t2 0
d2
2
t1 , t2
dt 2
t1 0, t2 0
E[Y 2 ] 2
21.0049
i
d d d
dt1
t , t
1 2 (3.93i 5.56t 2 1.05t1 ) exp 0.78it1 3.93it 2 1.1t 2
1 1.05t t
1 2 2.78t 2
2
dt2 dt1
(3.93i 5.56t2 1.05t1 ) exp 0.78it1 3.93it 2 1.1t12 1.05t1t 2 2.78t 22 (0.78i 2.2t1 1.05t 2 )
exp 0.78it1 3.93it 2 1.1t12 1.05t1t2 2.78t 22 1.05
d d
t1 , t2 (3.93i 0.78i) 1.05 (3.93i 0.78i) 1.05i 2 4.1154i 2
dt1 dt2 t1 0, t2 0
d d
t1 , t2
dt1 dt2 t1 0, t2 0
E[ XY ] 2
4.1154
i
Var[ X ] 2.8084 0.782 2.2
Var[Y ] 21.0049 3.932 5.56
Cov( X , Y ) 4.1154 (0.78 3.93) 1.05
Cov( X , Y ) 1.05
Corr ( X , Y ) 0.3
Var[ X ]Var[Y ] 2.2 5.56
Multivariate
distribution:
Let X X 1 X 2 ..... X k be a random vector with j.p.d.f.
the j.c.f of Xis f ( x1 , x2 ,..., xk )
(t1 , t2 ,.., tk ) E[ei (t1 X1 t2 X 2 ..tk X k ) ]
...... f ( x1 , x2 ,.., xk )ei (t1 x1 t2 x2 ...tk xk ) 1 2 k
t , t ,.., t k
xk x2 x1
....... f ( x1 , x2 ,.., xk )ei (t1 x1 t2 x2 ...tk xk ) dx1dx 2 ..dx k t1, t 2 ,.., t k k
xk x2 x1
Using the joint c.f ,
(i) we can define marginal c.f of each variable
marginal c.f of X j = X j (t j ) X j (0,.., 0, t j , 0,., 0)
marginal c.f of (X1 ,X 2 ,.., X m )=(X1 ,X2 ,..,X m ) (t1 , t2 ,.., tm , 0, 0,.., 0)
(ii) We can obtain the mean, variance of each variable as well as and
covariance values of each pair of the variables
d d2
(t1 , t2 ,.., tk ) 2
(t1 , t2 ,.., tk )
dt j dt j
t1 0,t2 0,..,tk 0 2 t1 0,t2 0,..., tk 0
E[ X j ] E[ X ] j
i i2
d d d d
(t1 , t2 ,.., tk ) (t1 , t2 ,..., tk )
dt j dti t1 0,t2 0,..,tk 0 dti dt j t 0,t 0,..,t 0
E[ X i X j ] 2
1 2 k
i i2
Characteristic function for independent
random variables
For n independent random variables Z1 , Z 2 ,..., Z n their
joint characteristic function is
n
t1 , t2 ,..., tn Zi (ti )
i 1
where Zi (ti ) is the characteristic function of Zi
n
The c.f for a random variable W; W Zisi
i 1
n
t E[eitW ] Zi (t )
i 1
Example 5
Let X 1 , X 2 ,..., X n be independent random variables , each with p.d.f
e i xi i
Pr( X i xi ) xi 0,1, 2,...
xi !
n
Find the c.f of Z X i
i 1
n
j (1 eit )
Z (t ) e
j 1
it n it
e ( 1 2 .. n )(1 e )
exp j 1 e
j 1
