Вы находитесь на странице: 1из 45

Chapter 5.

Random Vectors
Doug Young Suh
Media Lab.
suh@khu.ac.kr
5.1 Probability Models
Multivariate Joint CDF (3 tosses of dice)


Multivariate Joint PMF


Multivariate Joint CDF and PMF (continuous)


MediaLab , Kyunghee University 2
]. ,...., [ ) ,...., (
1 1 1 1 ,....,
1
x X x X P x x F
n n X X
n
s s =
]. ,...., [ ) ,...., (
1 1 1 1 ,....,
1
x X x X P x x P
n n X X
n
= = =
.
) ,......, (
) ,...., (
1
1 ,....,
1 ,....,
1
1
n
n X X
n
n X X
x x
x x F
x x f
n
n
c c
c
=

5.1 Probability Models: properties


Discrete

Continuous



Probability of event(subset)

MediaLab , Kyunghee University 3
, 0 ) ,...., (
1 ,....,
1
>
n X X
x x P
n
. 1 ) ,...., (
1
1
1
1 ,....,
=

e e
X
n
X n
n
S x S x
n X X
x x P
, 0 ) ,...., (
1 ,....,
1
>
n X X
x x f
n
} }

=
1
1 1
, ) ,...., ( ) ,...., (
1 1 ,...., 1 ,....,
x x
n n X X n X X
n
n n
du du u u f x x F
} }

=
1
1
. 1 ) ,...., (
1 1 ,....,
x x
n n X X
n
n
du du u u f
), , ,......... ( ] [
, ,.........
1 ,....
1
1

e
=
A x x
n X X
n
n
x x P A P
. ) ,...., ( ] [
1 1 ,....,
1 } }
=
n n X X A
du du u u f A P
n

Ex. 5.1
n independent trials
r possible outcomes {s
1
, , s
r
}, each p
i
n
i
times for outcome s
i





Roll a dice 10 times, and probability to
have three 2, two 4, four 5, one 6.
Roll yut 10 times and probability to have
three do, two gae, four girl, one mo.
MediaLab , Kyunghee University 4
r
r
n
r
n n
r
r N N
p p p
n n
n
n n P
2 1
1
2 1
1
1 ,....
,.......
) ,....., (
|
|
.
|

\
|
=
Ex. 5.3
A continuous joint PMF



Event A =

Sol

MediaLab , Kyunghee University 5

= s s
=
otherwise
n i x
x x f
i
n X X
n
0
, ,....., 1 , 1 0 1
) ,...., (
1 ,....,
1
{ | max 1/ 2 1,..., }
i i i
X X for i n s =
] 2 / 1 ,........, 2 / 1 [ ] 2 / 1 max [ ] [
1
s s = s =
n i
i
X X P X P A P
.
2
1
1
2 / 1
0
1
2 / 1
0
n
n
dx dx = =
} }

5.2 Vector Notation
Random vector
i
Vector sample value
i
CDF
PMF
PDF
Pair of random vectors
Joint CDF
Joint PMF
Joint PDF
MediaLab , Kyunghee University 6
]'. [
1 n
X X X =
]'. [
1 n
x x X =
). ,...., ( ) (
1 ,....,
1
n X X X
x x F X F
n
=
). ,...., ( ) (
1 ,....,
1
n X X X
x x P X P
n
=
). ,...., ( ) (
1 ,....,
1
n X X X
x x f X f
n
=
); ,...., , ,...., ( ) , (
1 1 ,..... , ,...., ,
1 1
n n Y Y X X Y X
y y x x F y x F
n n
=
); ,...., , ,...., ( ) , (
1 1 ,..... , ,...., ,
1 1
n n Y Y X X Y X
y y x x P y x P
n n
=
). ,...., , ,...., ( ) , (
1 1 ,..... , ,...., . ,
1 1
n n Y Y X X Y X
y y x x f y x f
n n
=
Ex. 5.4
For = [1 2 3] Find CDF


Sol)
MediaLab , Kyunghee University 7

>
=

otherwise
x e
x f
x a
X
0
0 6
) (
'

>
=

otherwise
x e
x f
i
x x x
X
0
0 6
) (
3 2 1
3 2

>
=

otherwise
x e e e
x F
i
x x x
X
0
0 ) 1 )( 1 )( 1 (
) (
3 2 1
3 2
5.3 Marginal Probability Function
For a joint PMFs

Marginal PMFs are



Marginal PDFs
MediaLab , Kyunghee University 8
) , , , (
, , ,
z y x w P
Z Y X W

e
=
W
S w
Z Y X W Z Y X
z y x w P z y x P ), , , , ( ) , , (
, , , , ,

e e
=
X Y
S x S y
Z Y X W Z W
z y x w P z w P ), , , , ( ) , (
, , , ,

e e e
=
W X Y
S w S x S y
Z Y X W X
z y x w P x P . ) , , , ( ) (
, , ,
( )
X
P x
Y
X
P
X
(3)
Z
1
2
3
4
5
6
}


= , ) , , , ( ) , , (
, , , , ,
dw z y x w f z y x f
Z Y X W Z Y X
, , ,
( ) ( , , , ) .
Z W X Y Z
f z f w x y z dwdxdy


=
} } }
Ex. 5.5


? ?
Sol)
MediaLab , Kyunghee University 9

s s s s s s
=
otherwise
y y y y
y y f
Y Y
0
, 1 0 , 1 0 4
) ,..., (
4 3 2 1
4 1 ,..,
4 1
. ) 1 ( 4 4 ) , (
4 1 3 2
0
4 1 ,
1
4
4 1
y y dy dy y y f
y
y
Y Y
= =
} }

s s s s
=
otherwise
y y y y
y y f
Y Y
0
, 1 0 , 1 0 ) 1 ( 4
) , (
4 1 4 1
4 1 ,
4 1
). 1 ( 2 ) 1 ( 4 ) , ( ) (
3 2
1
0
3 2 2 3 2 , 3
3 2 3
y dy y y dy y y f y f
Y Y Y
= = =
} }

s s
=
otherwise
y y
y f
Y
0
, 1 0 ) 1 ( 2
) (
3 3
3
3
) , (
4 1 ,
4 1
y y f
Y Y
) (
3
3
y f
Y
5.4 Independence
If independent,


Ex. 5.6 independent?
Sol)




dependent!!
MediaLab , Kyunghee University 10
). ( ) ( ) ( ) ,...., (
2 1 1 ,....,
2 1 1
n X X X n X X
x P x P x P x x P
N n
=
). ( ) ( ) ( ) ,...., (
2 1 1 ,....,
2 1 1
n X X X n X X
x f x f x f x x f
n n
=

s s s s s s
=
otherwise
y y y y
y y f
Y Y
0
, 1 0 , 1 0 4
) ,..., (
4 3 2 1
4 1 ,..,
4 1
). ,..., (
, 0
, 1 0 , 1 0 ) 1 ( ) 1 ( 16

) ( ) ( ) ( ) (
4 1 ,..,
4 3 2 1 4 3 2 1
4 3 2 1
4 1
4 3 2 1
y y f
otherwise
y y y y y y y y
y f y f y f y f
Y Y
Y Y Y Y
=

s s s s s s
=
, 1 0 ), 1 ( 2 ) , ( ) (
1 1 4 4 1 , 1
4 1 1
s s = =
}


y y dy y y f y f
Y Y Y
. 1 0 , 2 ) , ( ) (
4 4 1 4 1 , 4
4 1 4
s s = =
}


y y dy y y f y f
Y Y Y
. 2 ) 1 ( 4 ) , ( ) (
2 3
1
0
3 2 3 3 2 , 2
3 2 2
y dy y y dy y y f y f
Y Y Y
= = =
} }


5.5 Functions
For


Ex. 5.8) [0,1) n trials of spinning wheel
Pr{max<Y
n
}? CDF and PDF of Y
n
?

Sol) CDF PDF

MediaLab , Kyunghee University 11
) (X g W =
: ( )
( ) [ ] ( ),
W X
X g X w
P w P W w P x
=
= = =

}
}
= = =
s
. ) (
1
] [ ) (
) (
n X
w W F w F
dx dx x f
w X g
W

>
s s
<
=
, 1 1
, 1 0
, 0 0
) (
y
y y
y
y F
n
Y
n
1
0 1,
( )
0 .
n
n
Y
ny y
f y
otherwise

s s
=

Theorem 5.10
Theorem 5.10 Y
k
=aX
k
+b, a>0


Proof)
MediaLab , Kyunghee University 12
, ,...., ) (
1
|
.
|

\
|

=
a
b y
a
b y
F y F
n
X Y
, ,....,
1
) (
1
|
.
|

\
|

=
a
b y
a
b y
f
a
y f
n
X
n
Y
| |
n n Y
y b aX y b aX P y F s + s + = ,......, ) (
1 1
. ,...., ,......, ) (
1 1
1
|
.
|

\
|

=
(

s =
a
b y
a
b y
F
a
b y
X
a
b y
X P y F
n
X
n
n Y
. ,....,
1
) ,....., (
) (
1
1
1 ,....
1
|
.
|

\
|

=
c c
c
=
a
b y
a
b y
f
a y y
y y F
y f
n
X
n
n
n Y Y
Y
n
n

Matrix multiplication
(kXm)X(mXn) = (kXn)
(2X3)X(3X2) = (2X2)
1 2 3
1 2 0
1 2
2 1
1 3
=
(1X2)X(2X1) = (1X1)

1 2
2
5
=12
(2X1)X(1X2) = (2X2)

2
5
1 2
=
Matrix transpose
1 2 3

=
1
2
3

MediaLab , Kyunghee University 13
5.6 Expected value
Definition 5.11 Expected Value Vector
- The expected value od a random vector X is a column Vector



Example 5.9 if X = [
1

2

3
] what are the components of XX?
MediaLab , Kyunghee University 14
. ] ] [ . . . ] [ ] [ [ ] [
3 2 1
'
X E X E X E = = X E
X

Example 5.9 Solution


| |
(
(
(

X X X X X
X X X X X
X X X X X
= X X X
(
(
(

X
X
X
= X
'
X
2
3 2 3 1 3
3 2
2
2 1 2
3 1 2 1
2
1
3 2 1
3
2
1
Note that (3X1)X(1X3) = (3X3).
Definition 5.12 E[A]
Definition 5.12 Expected Value of a Random
Matrix
- For a random matrix A with the random variable A
ij
as
its i, j-th element, E[A] is a matrix with i, j-th element
E[A
ij
].





MediaLab , Kyunghee University 15
Definition 5.13 Vector Correlation
The correlation of a random vector X is an n x n matrix Rx with i, j th
element R
X
(i,j) = E[X
i
X
j
] in vector notation,



Example 5.10
if X = [
1

2

3
], the correlation matrix of X is

MediaLab , Kyunghee University 16
] [ X
'
X E =
X
R
(
(
(

X E
X E
X E
=
(
(
(

X E X X E X X E
X X E X E X X E
X X E X X E X E
=
X X X X
X X X X
X X X X
X
] [
] [
] [
] [ ] [ ] [
] [ ] [ ] [
] [ ] [ ] [
2
3
2
2
2
1
2
3 2 3 1 3
3 2
2
2 1 2
3 1 2 1
2
1
2 , 3 1 , 3
2 , 1 1 , 2
3 , 1 2 , 1
r r
r r
r r
R
Definition 5.14 Vector Covariance
The Covariance of a random vector X is an n x n
matrix C
x
with components C
X
(i, j)=Cov[X
i
,X
j
].



Example 5.11
If X = [
1

2

3
], the covariance matrix of X is



MediaLab , Kyunghee University 17
] ) )( [(
'
X X E =
X
x x C
(
(
(

X X X X X
X X X X X
X X X X X
=
X
] [ ] , [ ] , [
] , [ ] [ ] , [
] , [ ] , [ ] [
3 2 3 1 3
3 2 2 1 2
3 1 2 1 1
Var Cov Cov
Cov Var Cov
Cov Cov Var
C
Theorem 5.12
Theorem 5.12
- For a random vector X with correlation matrix
, covariance matrix , and vector expected
value ,

Proof :


Since is a constant vector,
MediaLab , Kyunghee University 18
X

X
R
X
C
X X X X
'
= R C
] [
X X X X X
'
+ X
'

'
X X
'
X E = C
] [ ] [ ] [ ] [
X X X X
'
E + X
'
E
'
X E X
'
X E =
X
= X E ] [
X X X X X X X X X
'
=
'
+ X
'
E
'
X E = R R C ] [ ] [
Example 5.12
- Find the expected value , the correlation matrix , and the
covariance matrix of the 2-dimensional random vector with PDF

MediaLab , Kyunghee University 19
X
C
] [X E
X
R
X

s s s
= X
X
. 0
, 1 0 2
) (
2 1
othewise
f
_ _
Example 5.12 Solution
-The elements of the expected value vector are


-The integrals are and so that
The elements of the correlation matrix are
MediaLab , Kyunghee University 20
2
1
1 2 1 2
0 0
[ ] ( ) 2 , 1, 2.
x
i i i
f dx x dx dx i _ _

X

E X = X = =
} } } }
3 / 1 ] [
1
= X E
3 / 2 ] [
2
= X E
= X E =
X
] [
| |. 3 / 2 3 / 1
'
, 2 ) ( ] [
2 1
1
0 0
2
1 2 1
2
1
2
1
2
dx dx dx dx f
x
} } } }
= X = X E
X


_ _
, 2 ) ( ] [
2 1
1
0 0
2
2 2 1
2
2
2
2
2
dx dx dx dx f
x
} } } }
= X = X E
X


_ _
, 2 ) ( ] [
2 1
1
0 0
2 1 2 1 2 1 2 1
2
dx dx dx dx f
x
} } } }
= X = X X E
X


_ _ _ _
Example 5.12 Solution
-These integrals are
Therefore,



-We use Theorem 5.12 to find the elements of the covariance matrix
MediaLab , Kyunghee University 21
1/ 6 1/ 4
1/ 4 1/ 2
R
X
(
=
(

(

=
(

=
'
=
X X X X
18 / 1 36 / 1
36 / 1 18 / 1
9 / 4 9 / 2
9 / 2 9 / 1
2 / 1 4 / 1
4 / 1 6 / 1
R C
2 2
1 2 1 2
[ ] 1/ 6, [ ] 1/ 2, [ ] 1/ 4 and E X = E X = E X X =
Definition 5.15 Vector Cross-Correlation

- The cross-correlation of random vectors, X with n
components and Y with m components, is an n x m
matrix R
XY
with i, j-th element

MediaLab , Kyunghee University 22
] [ Y
'
X E =
XY
R
], [ ) , (
j i
j i R Y X E =
XY
Definition 5.16 Vector Cross-Covariance
Definition 5.16 Vector Cross-Covariance

- The cross-covariance of a pair of random vector
with n components and with m components is an
n x m matrix with i , j th element

MediaLab , Kyunghee University 23
] ) )( [(
'
Y X E =
Y X XY
C
X
Y
XY
C
], [ ) , (
j i
Cov j i C Y X =
XY
Theorem 5.13
Theorem 5.13
- is an n-dimensional random vector with expected
value , correlation , and covariance .,
The m-dimensional random vector where
A is an m x n matrix and b is an m-dimensional
vector, has expected value , correlation matrix
and covariance matrix given by
MediaLab , Kyunghee University 24
.
, ) ( ) (
,
A
'
A =
'
+
'
A +
'
A + A
'
A =
+ A =
X Y
X X X Y
X Y
C C
b b b b R R
b


X
X
R
b + AX = Y
X

X
C
Y

Y
R
Y
C
Proof: Theorem 5.13
Proof: Theorem 5.13
- We derive the formulas for the expected value
covariance of . The derivation for the correlation is
similar. First, the expected value of is


It follows that This implies

MediaLab , Kyunghee University 25
Y
] ) ) ( ))( ( [(
'
X A X A E =
X X Y
C
b b b + A = E + X AE = + AX E =
X Y
] [ ] [ ] [
). (
X Y
X A = Y
Y
. ] ) )( [( ] ) )( ( [ A
'
A = A
' '
X X AE = A
' '
X X A E =
X X X X X
C
Example 5.13
Example 5.13
- Given random vector X defined in Example 5.12,
let Y=AX+b
where




Find the expected value , the correlation , and
the Covariance
MediaLab , Kyunghee University 26
Y

1 0 0
6 3 2
3 6 2
and b
( (
( (
A = =
( (
( (

Y
R
Y
C
Example 5.13 solution
Example 5.13 solution
- From the matrix operations of Theorem 5.13, we
obtain and
MediaLab , Kyunghee University 27
| |
'
=
Y
3 2 3 / 1
;
5 . 12 25 . 9 3 / 4
25 . 9 5 . 7 12 / 13
3 / 4 12 / 13 6 / 1
(
(
(

=
Y
R
(
(
(

=
Y
5 . 3 25 . 3 3 / 1
25 . 3 5 . 3 12 / 5
3 / 1 12 / 5 18 / 1
C
Theorem 5.14
Theorem 5.14
- The vector and have cross-correlation
and cross-covariance given by
MediaLab , Kyunghee University 28
X
XY
R b + AX = Y
XY
C
b R R
'
+ A
'
=
X X XY
A
'
=
X XY
C C
Example 5.14
Example 5.14
- Continuing Example 5.13 for random vectors and
calculate

(a) The cross-correlation matrix and the cross-
covariance matrix
(b) The correlation coefficients




MediaLab , Kyunghee University 29
, b + AX = Y
3 1
,Y Y

X
XY
C
XY
R
1 2
Y X

and
Example 5.14 solution
(a) Direct matrix calculation using Theorem 5.14



(b) Referring to Definition 4.8 and recognizing that
, we have


Similarly,
MediaLab , Kyunghee University 30
;
12 / 29 3 / 5 4 / 1
3 / 4 12 / 13 6 / 1
(

=
XY
R
(

=
XY
12 / 5 3 / 1 36 / 1
3 / 1 12 / 5 18 / 1
C
) , ( ] [ i i C Var
i Y
= Y
1 3
1 3
1 3
[ , ] (1, 3)
0.756
[ ] [ ] (1,1) (3, 3)
Cov C
Var Y Var C C

Y
Y Y
Y Y
Y Y
= = =
Y
2 / 1
) 1 , 1 ( ) 2 , 2 (
) 1 , 2 (
] [ ] [
] , [
1 2
1 2
1 2
= =
Y X
Y X
=
Y X
XY
Y X
C C
C
Var Var
Cov

Quiz 5.6
Quiz 5.6
- In Quiz 5.3, the 3-dimensional random vector
has PDF



Find the expected value , and the correlation and
covariance matrices and
MediaLab , Kyunghee University 31
| |
'
X X X = X
3 2 1
] [X E
X
R

s s s s
= X
X
otherwise
f
0
, 1 0 6
) (
3 2 1
_ _ _
X
C
Quiz 5.6 Solution
Quiz 5.6 Solution
- We start by finding the components
of . To do so, we use the marginal PDFs found
in Quiz 5.3:


To find the correlation matrix , we need to find
For all i , and j. we start with the second moments:
MediaLab , Kyunghee University 32
X

X
R
}
}
}
= = X E
= = X E
= = X E
1
0
3
3
1
0
2
2
1
0
2
1
. 4 / 3 3 ] [
2 / 1 ) 1 ( 6 ] [
, 4 / 1 ) 1 ( 3 ] [
dx x
dx x x
dx x x
dx x xf
i
x i
) ( ] [
}


= X E
) (x f
i
x
] , [
j i
X X E
}
}
}
= = X E
= = X E
= = X E
1
0
4 2
3
1
0
3 2
2
1
0
2 2 2
1
. 5 / 3 3 ] [
, 10 / 3 ) 1 ( 6 ] [
, 10 / 1 ) 1 ( 3 ] [
dx x
dx x x
dx x x
Quiz 5.6 Solution
Quiz 5.6 Solution
- Using marginal PDFs from Quiz5.3z the cross terms are







Summarizing the results, X has correlation matrix
MediaLab , Kyunghee University 33
. 5 / 1 ] 3 2 [
1
| ) 3 2 (
1
4
1
2
1
1
0
1 1
1
0
1 3
3 2
3
2
1
3
3 1
= + =
|
|
.
|

\
|
=
=
=
} }
dx x x x dx
x x
x
x x x x
}


= X X E
2 1 2 1 2 1 2 1 2 1
), , ( ] [ x dx x x x fx x x
. 20 / 3 ] 2 3 [ ) 1 ( 6
1
4
1
1
0
3
1 1 1
1
0
1
2 2 2 1
1
= + =
|
.
|

\
|
=
} } }
dx x x x dx dx x x x
x
5 / 2 ] 3 3 [ 6 ] [
2
4
2
1
0
2
2 2 3 3
1
0
1
2
2 3 2
2
= = = X X E
} } }
dx x x dx dx x x
x
. ) ( 6 ] [
1 3 1 3 3
1
0
1
1 3 2
1
dx dx x x x x
x
= X X E
} }
(
(
(

=
X
5 / 3 5 / 2 5 / 1
5 / 2 10 / 3 20 / 3
5 / 1 20 / 3 10 / 1
R
Quiz 5.6 Solution
Quiz 5.6 Solution
- Vector X has covariance matrix






This problem shows that even for fairly simple joint
PDFs, computing the covariance matrix by calculus
can be a time consuming task.
MediaLab , Kyunghee University 34
] [ ] [
'
X E X E =
X X
R C
| | 4 / 3 2 / 1 4 / 1
4 / 3
2 / 1
4 / 1
5 / 3 5 / 2 5 / 1
5 / 2 10 / 3 20 / 3
5 / 1 20 / 3 10 / 1
(
(
(

(
(
(

=
.
3 2 1
2 4 2
1 2 3
80
1
16 / 9 8 / 3 16 / 3
8 / 3 4 / 1 8 / 1
16 / 3 8 / 1 16 / 1
5 / 3 5 / 2 5 / 1
5 / 2 10 / 3 20 / 3
5 / 1 20 / 3 10 / 1
(
(
(

=
(
(
(

(
(
(

=
5.7 Gaussian Random Vectors

=
1
(2)

2
[det (

)]
1/2
exp
1
2
(

1
(

)
MediaLab , Kyunghee University 35
Theorem 5.15
Theorem 5.15
- A Gaussian random vector X has independent components if and
only if is a diagonal matrix

Proof: Theorem 5.15
- First, if the components of X are independent, then for and are
independent. By Theorem 4.27(c), . Hence the off-diagonal
terms of are all zero. If is diagonal, then






It follows that has determinant det and that
MediaLab , Kyunghee University 36
X
C
i
X j i , =
j
X
0 ] , [ = X X
j i
Cov
X
C
X
C
.
/ 1
/ 1
2
2
1
1
2
2
1
(
(
(

=
(
(
(

=

X X
n n
C and C
o
o
o
o

X
C
2
1
) (
i
n
i
C o
= X
[ =
Proof: Theorem 5.15



From Definition 5.17, we see that






Thus , implying are independent.

MediaLab , Kyunghee University 37
.
) (
) ( ) (
1
2
2
1

X
=
'

n
i
i
i i
x x x
x C x
o


|
|
.
|

\
|

[
=

=
=
n
i
i i i
i
n
i
n
X
x X f
1
2
2
1
2 /
2 / ) ( exp
) 2 (
1
) ( o
o t
). 2 / ) ( exp(
2
1
2 2
2
1
i i i
i
n
i
x o
to
=
[
=
[
=
=
n
i
i i
x fx X fx
1
) ( ) (
n
X X ,...,
1
Example 5.15
Example 5.15
-Consider the outdoor temperature at a certain weather station. On May 5, the temperature
measurements in units of degrees Fahrenheit taken at 6AM are all Gaussian random variables,
with variance 16degrees . The expected values are 50degrees, 62 degrees, and
58degrees respectively. The covariance matrix of the three measurements is





(a) Write the joint PDFs of using the algebraic notation of Definition 4.17.
(b) Write the joint PDFs of using vector notation.
(c) Write the joint PDFs of using vector notation.



MediaLab , Kyunghee University 38
3 2 1
, , X X X
2
(
(
(

=
0 . 16 8 . 12 2 . 11
8 . 12 0 . 16 8 . 12
2 . 11 8 . 12 0 . 16
Cx
2 1
,X X
2 1
,X X
| |
'
=
3 2 1
X X X X
Example 5.15 Solution
Example 5.15 Solution
(a) First we note that and have expected values and , variance
and, covariance . It follows from Definition 4.8 that
the correlation coefficient is



From Definition 4.17, the joint PDF is





(b) Let denote a vector representation for random variables and .
From the covariance matrix , we observe that the 2 x 2 submatrix in the upper left
corner is the covariance matrix of the random vector W. Thus
MediaLab , Kyunghee University 39
1
X
. 8 . 0
16
8 . 12 ] , [
2 1
2 1
,
2 1
= = =
o o

X X Cov
X X
| |
'
=
2 1
X X W
2
X
50
1
= 62
2
=
16
2
2
2
1
= = o o
8 . 12 ] , [
2 1
= X X Cov
.
3 . 60
2 . 19
) 62 ( ) 62 )( 50 ( 6 . 1 ) 50 (
exp
) , (
2
2 2 1
2
1
2 1 ,
2 1
|
|
.
|

\
|
+

=
x x x x
x x f
X X
1
X
2
X
X
C
Example 5.15 Solution
Example 5.15 Solution


We observe that det and From Definition 5.17,
the joint PDF of W is



(c) For the joint PDF of X, We note X has expected value and that
det Thus



MediaLab , Kyunghee University 40
(

=
, 62
50
W

| |
'
= 58 62 50
X

.
0 . 16 8 . 12
8 . 12 0 . 16
(

=
W
C
16 . 92 ) ( =
W
C . 6 . 9 ) (
2 / 1
=
W
C
|
|
.
|

\
|
=

w w C w w w f
W
T
W
( ) (
2
1
exp
3 . 60
1
) (
1
. 717 . 22 ) (
2 / 1
=
X
C
|
|
.
|

\
|
=

x x C x x X f
x
T
X
( ) (
2
1
exp
8 . 357
1
) (
1
Theorem 5.16
Given an n-dimensional Gaussian random vector X with expected value
and covariance , and an m x n matrix A with rank(A) = m,


Is an m-dimensional Gaussian random vector with expected value
and covariance

Proof: Theorem 5.16
- The proof of Theorem 5.13 contains the derivations of and . Our
Proof that Y has a Gaussian PDF is confined to the special case when m=n
and A is an invertible matrix. The case of m < n is addressed in Problem
5.7.9. When m = n, we use Theorem 5.11 to write



MediaLab , Kyunghee University 41
X
C
b AX Y + =
b A
X
+ . A AC C
X Y
'
=
Y

)) ( (
| ) det( |
1
) (
1
b y A fx
A
y f
Y
=

X

=
Y

Y
C
2 / 1 2 /
1 1 1
| ) det( || ) det( | ) 2 (
] ) ( [ ] ) ( [
2
1
exp
X
n
X X X
C A
b y A C b y A
t

|
|
.
|

\
|
'
=

Proof of Theorem 5.16
In the exponent of , we observe that


Since
Applying (5.79)to (5.78) yields




Using the identities
We can write



Since , we see from Equation (5.81) that Y is a
Gaussian vector with expected value and covariance matrix



MediaLab , Kyunghee University 42
), ( )] ( [ ) (
1 1 1
Y X X
y A b A y A b y A = + =

b A
X Y
+ =
1 1 1 1
) ' ( ) ' (

= A AC A C A
X X
, ) ( ) ( | ) det( | | ) det( || ) det( |
1 1 2 / 1 2 / 1
'
=
' '
= A A and A AC C A
X X
Y

) (y f
Y
2 / 1 2 /
1 1 1
| ) det( || ) det( | ) 2 (
)] ( [ ] ) ( [
2
1
exp
) (
X
n
Y X Y
Y
C A
y A C y A
y f
t

|
|
.
|

\
|

'

=

2 / 1 2 /
1 1 1
| ) det( | ) 2 (
) ( ) ( ) (
2
1
exp
) (
A AC
y A C A y
y f
X
n
Y X Y
Y
'
|
|
.
|

\
|

' '

=

t

'. A AC C
X Y
=
Example 5.16
Continuing Example 5.15, use the formula to convert the three
temperature measurements to degrees Celsius.
(a) What is , the expected value of random vector Y ?
(b) What is , the covariance of random vector Y ?
(c) Write the joint PDF of using vector notation.

Example 5.16 Solution
(a) In terms of matrices, we observe that where



(b) Since from Theorem 5.16,


(c) The covariance of Y is . We note that where I is the 3 x 3
identity matrix. Thus . The PDF of Y is

MediaLab , Kyunghee University 43
) 32 )( 9 / 5 ( =
i i
X Y
Y

| |
'
=
3 2 1
Y Y Y Y
Y
C
b AX Y + =
.
1
1
1
9
160
,
9 / 5 0 0
0 9 / 5 0
0 0 9 / 5
(
(
(

=
(
(
(

= b A
| |, 58 62 50
'
=
Y

.
9 / 130
3 / 50
10
(
(
(

= + = b A
X Y

A AC C
X Y
' = I = ' = ) 9 / 5 ( A A
1 2 1
) 5 / 9 ( ) 9 / 5 (

= =
X Y X Y
C andC C C
. ) 9 ) (
50
81
exp
47 . 24
1
) (
1
|
|
.
|

\
|
=

Y X
T
Y Y
y C y y f
Definition 5.18
Definition 5.18 Standard Normal Random Vector
- The n-dimensional standard normal random vector Z is the n-dimensional Gaussian
random vector with

Theorem 5.17
- For a Gaussian random vector, let A be an n x n matrix with the property
The random vector is a standard normal random vector.

Proof : Theorem 5.17
- Applying Theorem 5.16 with A replaced by , and , we have that Z is a
Gaussian random vector with expected value


and covariance


MediaLab , Kyunghee University 44
( ) 0
Z
E Z and C = = I
) , (
X X
C
. '
X
C AA =
) (
1
X
X A Z =

| | | | , 0 ) ( ) (
1 1
= = =

X X
X E A X A E Z E
1
A
. ) ' '( ) (
1 1 1 1
I = = ' =

A AA A A C A C
X Z
X
A b
1
=
Theorem 5.18
Given the n-dimensional standard normal random vector Z, an invertible
n x n matrix A, and an n dimensional vector b, X=AZ+b is an n-
dimensional Gaussian random vector with expected value
X
and
covariance matrix C
X
=AA.

Proof : Theorem 5.18
- By Theorem 5.16, X is a Gaussian random vector with expected value


The covariance of X is

MediaLab , Kyunghee University 45
. ] [ ] [ ] [ b b Z AE AZ E X E
X X
= + = + = =
'. ' ' AA A A A AC c
Z X
= I = =

Вам также может понравиться