Scribd
Upload
29 views

Wχ f F F

The document contains solutions to 4 problems from a problem set. Problem 1 involves finding the density, distribution function, and inverse distribution function for a random variable. Problem 2 examines properties of the minimum and maximum of iid exponential random variables. Problem 3 explores the law of total variance. Problem 4 provides an example of random variables where the expected value approaches 0 but the expected squared value approaches 1.

Uploaded by

Karl Stessy Premier
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
29 views

Wχ f F F

The document contains solutions to 4 problems from a problem set. Problem 1 involves finding the density, distribution function, and inverse distribution function for a random variable. Problem 2 examines properties of the minimum and maximum of iid exponential random variables. Problem 3 explores the law of total variance. Problem 4 provides an example of random variables where the expected value approaches 0 but the expected squared value approaches 1.

Uploaded by

Karl Stessy Premier
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

Problem Set #2 9/20/2015

Problem #1
2
W 2= Gamma (1, 1/2). Find the density f n , function distribution Fn , and the

(a) If

inverse distribution function

F1
W
explicitly.

Solution:
If

W 22= Gamma (1, 1/2), then


f W ( w )=

( 12 ) exp( w2 ) for w 0.
W Exponential (1/2). So we can compute the distribution function as

We can clearly see that

FW ( w )=1exp

( w2 )

for w 0 .

Finally, the inverse distribution function is given as follows, let

Then,

exp

( 2y )=1w

y=F1
W ( w)

. Hence,

1
FW ( w )=2 log( 1w)

(b) Suppose that ( X , Y ) N ( 0, I ) . Show that

=tan1

( YX )

R and defined by R2= X 2+Y 2 and

are independent random variables with

R2 22 and Uniform ( 0, 2 )

.
Solution:
Since ( X , Y ) N ( 0, I ) , then the joint distribution is given by

f X ,Y ( x , y )=

Now for

1
x 2 + y 2
exp
for ( x , y )R2
2
2

x=rcos ( ) and

y=rsin() with

+
r R and (0, 2 )

x +y

1 y
)= ( r , )
The inverse function (, tan
x

1
f X ,Y ( x , y ) : ( x , y )

()

rcos ( + ) ,rsin ( + ) ( rcos ( ) + )


rcos ( ) +f X ,Y
Then
rcos ( ) ,rsin()
f R , ( r , ) =f X ,Y
1
r 2
1
r 2

exp
r+
exp
r
2
2
2
2

( )

( )

1
r 2
exp
r

( )

r 2
r exp
on( 0, ) ( 0,2 )
2

( )

So,

R and are independent variables.


(c) Use (a) and (b) to show how to use two independent random Uniform (0, 1) variables U V to
generate two standard normal distribution variables.

Solution:
If
get

UV

are independent Uniform (0, 1) random variables, we can use the inverse transformation to

( U )=2 log (1U ) 22


R2 F1

2
2

and

2 V

Thus, ( X , Y ) ( Rcos ( ) , Rsin ( ) ) N 2 ( 0, I ) .

Problem #2
Suppose that
(a) Show

X 1 , X 2 , , are iid Exponential ( ). Let


n M n d exponential( )

M n min X i
1 i n

and

T n max X i
1 i n

Solution:
Consider

X 1 , X 2 , , are iid Exponential ( ) then,

P ( n M n > x ) =P min X j >


1 j n

x
n

x
x
x
P X 1 > , X 2 > , , X n>
n
n
n

x
x
x
P X 2> P X n >
n
n
n

x n
x
= exp
n
n

P X1>

P X1>

) (

) (

) ( ( ))

( x)
exp
Therefore,

n M n d exponential( ) as required.

(b) Show that

( 1 )T

T n

where

has the double exponential extreme value distribution

function given by P (T x )=exp ( exp (x ) ) .


Solution:

T n then,

We know that maximum are at

( ()

P T n

) (

() )

1
1
log n x =P max X j x +
log n

1 j n

( 1 ) log n , X x+( 1 ) log n , , X x +( 1 ) log n)

( 1 ) log n ) P ( X x +( 1 ) log n) P ( X x+( 1 ) log n )

P X1 x +

P X1 x +

1
P X1 x +
log n

() )

exp (x )
n

( 1exp (xlogn ) ) = 1

1
(
( ) log n x)=P (T x )=exp (exp(x ) )

P T n

Hence,

(c) Now suppose that

as n approaches infinity and =1 .

X 1 , X 2 , , X n are iid with distribution function

0< F ' ( 0 ) < ; here

F '(0) is the right-derivative of

F satisfying

F at 0 :

F ( x )F ( 0 )
'
=F ( 0 ) . Show that n M n d exponential ( F ' ( 0 )) .
x

lim
x 0

Solution:

P ( n M n > x ) =P min X j >

Consider,

x
x
x
P X 1 > , X 2 > , , X n>
n
n
n

x
x
x
P X 2> P X n >
n
n
n

x
n

P X1>

nF

Because

) (

) (

x
n

()
n

P X1>

1 j n

x
n

F has a derivative

F ' (0) the right derivative of F ( 0 ) =0 ; Then we can follow the

formula:

nF ( x /n ) =

( nx ) =F ( 0)

F ( x /n ) F ( 0 )
x=
x /n
1/n

'

Hence,

nF
1

x
n

()
n

n M n d exponential ( F ' ( 0 )) .
exp (F ' ( 0 ) x ) , as required

Problem #3
Suppose that

(a) Show that

is a random variable with E ( Y ) < .

Var ( Y )=E { Var (Y | X ) } +Var { E ( Y |X ) }

Solution:

E (Y | X )E ( Y )
[]
{ 2 } .
We want to show that,
2
E [ ( Y E (Y | X ) ) |X } + E
2

E ( Y E ( Y ) ) =E
By direct computation:
2

Var ( Y )=E [ Y E ( Y ) ] =E [ Y E (Y | X ) + E ( Y |X ) E (Y ) ]

( Y E ( Y | X ) ][ E ( Y | X )E ( Y ) ] + E [ E ( Y| X )E ( Y ) ]
2
E [ Y E ( Y |X ) ] +2 E

Now, because we have conditional expected value, we should expect 0 for the middle term. That is,

( Y E ( Y | X ) ][ E ( Y | X )E ( Y ) ]=E ( E ( [ Y E ( Y | X ) ][ E ( Y |X )E ( Y ) ]|X ) )
E
E ( [ E ( Y |X )E ( Y ) ] E ( [ Y E ( Y |X ) ]|X ) )
E ( [ E ( Y |X )E ( Y ) ] ( [ E ( Y | X )E ( Y X ) ] ) )
E ( [ E ( Y |X ) E (Y ) ] 0 ) =0
Hence,

Var ( Y )=E E ( [ Y E ( Y | X ) ] | X ) +Var ( E ( Y |X ) )

E ( Var ( Y | X ) ) ) +Var ( E ( Y |X ) )
(b) Interpret (a) geometrically.
Solution:
From the textbook definition we can say that

Y E ( Y | X ) is orthogonal to

E ( Y |X ) E(Y ) in

L2 ( P ) probability space. The identity defined in problem (a) is interpreted as a Pythagorean Theorem
equality.
2

(c) Suppose that Y n () . Compute

E(Y ) and Var (Y ) . Use

E ( Y )=E ( E ( Y |X ) ) and

(a).
Solution:
2
Suppose that ( Y |X ) 2 X +n

where

E ( Y )=E ( E ( Y |K ) ) =E ( 2 K + n )=n+ 2

Then using part (a)

Poisson ( / 2) . Then,

( 2 )=n+

Var ( Y )=E ( Var (Y | X ) ) +Var ( E ( Y |K ) )

E ( 2 )=2(2 K +n)+Var (2 K + n)

( 2 )+ 2n+ 4 ( 2 )=4 +2 n
2

(d) Show that

n ( )( n+ )
d N (0,1) as either n or .
2n+ 4

Solution:
Suppose

Y d ( Z 1+ ) + Z22 ++ Z 2n and Z i N (0,1) independent. Then,


2

1
2
2n ( )( n+ ) ( ( Z 1+ ) ( 1+ ) + ( Z 21 ) + + ( Z n1 ) )

( 2n+ 4 )
2n+ 4 d
2

( Z 1+ ) ++(Z 2n 1) 2 Z 1
d
+
( 2n+ 4 )
2n+ 4

T 1 ++T n
2 Z1
+
( 2 n+ 4 ) 2n+ 4

Thus,

2 Z1
p 0 and
2 n+ 4

T 1 ++T n
d N ( 0,1) as n .
( 2 n)

n ( )( n+ )
d N (0,1) as n .
2n+ 4

Hence,

Problem # 4
Give an example of random variables such that

E|X n| 0 and E|X n|2 1 .

Solution:
Consider the random variable

Then,

Clearly,

E X n n Pn=n

X n such that X n = n
0, else , with probability

1
1
=
2
n
n

( )

and

E|X n| 0 as n gets larger.

E|X n| =n2 p n=n2

pn=1/n2

1
=1 for all n .
n2

( )

You might also like