Submit Assignment For Help

Go To Answer Directly
info@statisticsassignmenthelper.com

14.30 PROBLEM SET 4

TA: Tonja Bowen Bishop

Due: Tuesday, March 21, by 4:30 p.m.

PROBLEM SET 4
Note: The …rst three problems are required, and the remaining two are
practice problems. If you choose to do the practice problems now, you will
receive feedback from the grader. Alternatively, you may use them later in
the course to study for exams. Credit may not be awarded for solutions
that do not use methods discussed in class.

Problem 1
Suppose that X, Y , and " are random variables with Y = + X + ".
Assume E (X) = 0, V ar (X) = 2 , " has a uniform distribution over 21 ; 21 ,
and Cov (X; ) = 0.
a. What is the distribution (pdf) of Y jX?

b. What is E (Y jX)? What is V ar (Y jX)?

c. What is E (Y )? What is V ar (Y )?

Problem 2
a. Assume that X is distributed uniformly over the interval [0; 4].
Calculate the moment generating function of X. Use the MGF to …nd the
mean and variance of X.

b. Use the Chebyshev inequality to calculate an upper bound on the

probability that X is outside the interval [0:5; 3:5].

c. Now Use
use the fact that X is distributed uniformly over the interval
[0; 4] to calculate the probability that X is outside the interval [0:5; 3:5]. Is
it higher or lower than the answer you got in part b?

d. Based on your answers in parts b and c, comment on the usefulness

of the Chebyshev inequality.

Problem 3
Let X be distributed uniformly on the interval [0; 1].
a. Let Y = X 2 . Use the 1-step and 2-step methods to …nd the pdf of
Y.

b. Let Z = log(X), for > 0. Calculate the pdf of Z.

1

https://www.statisticsassignmenthelper.com/
2 14.30 PROBLEM SET 4

Problem 4
Let X have the following distribution.
1 2 x
f (x) =
3 3
x 2 f0; 1; 2; :::g
X
Find the pdf of Y = X+1 .

Problem 5
Let X have the pdf f (x) = 12 (x + 1) for 1 x 1.
a. Let Y = X 2 . Use the 1-step and 2-step methods to …nd the pdf of
Y.

b. Find the MGF of Y .

c. What is E (Y )? What is V ar (Y )?
1
d. What is E Y jX 2 ?
 14.30 PROBLEM SET 4 - SUGGESTED ANSWERS

TA: Tonja Bowen Bishop

Problem 1
a. The distribution of Y jX is the distribution of + X + ", where
" is the only random variable and + X is …xed, call it C. Then, the
distribution of C + e is that of e , shifted by the value C; or U [C 21 ; C + 12 ]:
Thus Y jX U + X 21 ; + X + 21 :

b. For a random variable Z with a uniform distribution U [a; b]; fZ (z) =

1=(b a);
Rb h 2 ib
b2 a2
E[Z] = a b z a dz = 2(bz a) = 2(b b+a
a) = 2 ;
h 3 ib a
Rb 2 b3 a3 b2 +a2 +2ab
E[Z 2 ] = a bz a dz = 3(bz a) = 3(b a) = 3 ;
a
b2 +a2 +2ab (b+a)2 (b a)2
V ar[Z] = E[Z 2 ] E[Z]2 = 3 4 = 12 :
Therefore, for U [ + X 0:5; + X + 0:5];
+ X 0:5+ + X+0:5
E[Y jX] = b+a
2 = 2 = + X;
(b a)2 [ + X+0:5 ( + X 0:5)]2 1
V ar[Y jX] = 12 = 12 = 12 :

c. We can calcualte the expected value directly:

E[Y ] = E[ + X + e]
= + E[X] + E[e]
= + 0+0
=
Or, we could use the law of iterated expectations:
E[Y ] = EX [EY [Y jX]]
= EX [ + X]
= + E [X]
=
For the variance, we can again calculate either directly or using our con-
ditional variance identity.
1
2 14.30 PROBLEM SET 4 - SUGGESTED ANSWERS

V ar[Y ] = V ar[ + X + e]
2
= V ar[X] + V ar[e] + 2 Cov(X; e)
2 2 [0 + 0:5 (0 0:5)]2
= + +2 0
12
2 2 1
= +
12
or
V ar[Y ] = E(V ar(Y jX)) + V ar(E(Y jX))
1
= E( ) + V ar( + X)
12
1
= + 2 V ar[X]
12
1
= + 2 2
12

Problem 2
a. The moment generating function is de…ned as
MX (t) = E etX
Since X is distributed uniformly over [0; 4], we have
MX (t) = E etX
Z 4
1 tx
= e dx
0 4
1 4t
= e 1
4t
To …nd the mean, we need to take the …rst derivative of the MGF and
evaluate it at t = 0, using l’Hopital’s rule in the fourth line.
@ @ 1 4t
MX (t) j = e 1 j
@t t=0 @t 4t t=0
e4t e4t 1
= j
t 4t2 t=0
4te4t e4t + 1
= j
4t2 t=0
16te4t + 4e4t 4e4t
= j
8t t=0
= 2 = E (X)
Then, to …nd the variance, we will take the second derivative of the MGF
at t = 0,and
(again
thenmaking usesquare
subtract the of l’Hopital’s rule), and then subtract the square
 14.30 PROBLEM SET 4 - SUGGESTED ANSWERS 3

of the expected value.

@2 4te4t e4t 16t2 e4t 8t e4t 1
MX (t) j = j
@t2 t=0 t2 16t4 t=0
4te4t e4t 2te4t e4t + 1
= j
t2 2t3 t=0
8t2 e4t 4te4t + e4t 1
= j
2t3 t=0
32t2 e4t + 16te4t 16te4t 4e4t + 4e4t
= j
6t2 t=0
16e4t
= j
3 t=0
16
= = E X2
3
16 4
V (X) = 4=
3 3

b. The Chebyshev inequality states that Pr (jX E [X]j t) V ar[X] t2

for t > 0. Here: E [X] = 2; V ar [X] = 43 ; t = 32 so: Pr (X 2
= (0:5; 3:5)) =
(4=3)
Pr jX 2j 23 (3=2)2
= 16
27 :

R 0:5 1
R4 1 1 16
c. Pr (X 2
= (0:5; 3:5)) = 0 4 dx + 3:5 4 dx = 4 < 27 so we get a lower

result than in part a.

d. The Chebyshev inequality is very useful for evaluating distributions

for which you only know the mean and the variance, but not the actual
distribution. If you know the actual distribution you can get a more precise
answer. But this is only because you are using additional information.

Problem 3
a. As instructed, we will …nd the pdf of Y using both the 1-step and
2-step methods. Note that Y will take on values in [0; 1]. We begin with
the 2-step method by calculating the CDF:
FY (y) = Pr (Y y) = Pr X 2 y
p p
= Pr ( y X y)
p
= Pr (0 X y)
p p
= FX ( y) = y
1 1
fY (y) = y 2 for y 2 [0; 1]
2
= 0 elsewhere
4 14.30 PROBLEM SET 4 - SUGGESTED ANSWERS

Then we use the 1-step method. Because the transformation is already

monotonic on the relevant range, we do not have to worry about dividing
the range into monotonic pieces.

@r 1 (y)
1
fY (y) = fX r (y) for y 2 [0; 1]
@y
p 1 1
fY (y) = fX ( y) y 2 for y 2 [0; 1]
2
1 1
= y 2 for y 2 [0; 1]
2
= 0 elsewhere

b. Now we have Z = ln X. So the range of Z is [0; 1). We again

have a monotonic transformation, so we will use the 1-step method. Note
z
that r 1 (z) = e .

@r 1 (z)
1
fZ (z) = fX r (z) for z 2 [0; 1)
@z
z
z e
= fX e for z 2 [0; 1)
z
e
= for z 2 [0; 1)
= 0 elsewhere

Note that this

This is the exponential distribution. It turns out that most
distributions can be constructed as a transformation of a U [0; 1] random
variable.

Problem 4
X
Because Y = X+1 is a one-to-one transformation, we know that there will
be only one x value that corresponds to each valid y value.

X
fY (y) = Pr (Y = y) = Pr =y
X +1
y y
= Pr X = = fX
1 y 1 y
y
1 2 1 y
=
3 3
 14.30 PROBLEM SET 4 - SUGGESTED ANSWERS 5

for valid values of y. What is the support of Y ? We can substitute in

for the …rst few potential values of X to see the pattern:

0
= 0
0+1
1 1
=
1+1 2
2 2
=
2+1 3
3 3
=
3+1 4

so Y 2 f0; 21 ; 23 ; 34 ; :::g, or Y 2 fyjy = x

x+1 for whole numbers xg

Problem 5
a. We will use the 1-step method, and so we must …rst divide the range
of X into segments for which the transformation function is monotone,
( 1; 0) and (0; 1) (we can ignore the endpoints of these intervals because
the probability that X equals a particular point is zero). We will use the
1-step method on each segment and then sum our results.
For the segment ( 1; 0), our transformation function is r (x) = x2 , and
p
the inverse function is r 1 (y) = y. The range of Y is (0; 1). So we
have

p 1 1
fY (y) = fX ( y) y 2 for y 2 (0; 1)
2
1 p 1
= (1 y) y 2 for y 2 (0; 1)
4
1 1
= y 2 1 for y 2 (0; 1)
4
= 0 elsewhere

Then, for the segment (0; 1), our transformation function is r (x) = x2 ,
p
and the inverse function is r 1 (y) = y. The range of Y is again (0; 1).
So we have

p 1 1
fY (y) = fX ( y) y 2 for y 2 (0; 1)
2
1 p 1
= (1 + y) y 2 for y 2 (0; 1)
4
1 1
= y 2 + 1 for y 2 (0; 1)
4
= 0 elsewhere
6 14.30 PROBLEM SET 4 - SUGGESTED ANSWERS

So adding these two functions together we get

1 1 1 1
fY (y) = y 2 1 + y 2 +1 for y 2 (0; 1)
4 4
1 1
= y 2 for y 2 (0; 1)
2
= 0 elsewhere

b. To get the moment generating function of Y , we …nd E etY .

MY (t) = E etY
Z 1
1 1
= ety y 2 @y
0 2
which does not have a closed form.

c. We can still use the MGF to get E (Y ):

@
E (Y ) = MY (t) j
@t t=0
Z 1
@ 1 1
= ety y 2 j @y
0 @t 2 t=0
Z 1
1 1
= yety y 2 j @y
2 0 t=0
Z
1 1 1
= y 2 @y
2 0
1 3 1 1
= y2 j =
3 y=0 3

And similarly, we can get E Y 2 :

@2
E Y2 = MY (t) j
@t2 t=0
Z 1 2
@ 1 1
= 2
ety y 2 j @y
0 @t 2 t=0
Z
1 1 ty 3
= e y 2 j @y
2 0 t=0
Z 1
1 3
= y 2 @y
2 0
1 5 1 1
= y2 j =
5 y=0 5
1 1 4
So V ar (Y ) = 5 9 = 45 .
 14.30 PROBLEM SET 4 - SUGGESTED ANSWERS 7

1
d. We start by …nding the pdf of Y jX 2 . This will be the similar
to part a., except that our transformation function is monotonic over the
1
relevant range, and we will use the pdf of XjX 2 , which is just the pdf of
X over the relevant range scaled up by one over the probability that X is
in this range:
1
2 (x + 1) 1
fXjX 1 (x) = R1 1
for x 1 (and 0 elsewhere)
2
1 (x + 1) dx 2
2
2
x+1
= 1
x2
2 +x j
x= 21
8
= (x + 1)
7
Thus we can calculate
p 1 1 1
fY jX 1 (y) = fXjX 1 ( y) y 2 for y 2 ; 1 (and 0 elsewhere)
2 2 2 4
4 1
= 1+y 2
7
1
It is then straightforward to calculate E Y jX 2 :
Z 1
1 4y 1
E Y jX = 1 + y 2 dy
2 1 7
4
Z 1
4 1
= y + y 2 dy
7 1
4

4 y2 2 3 1
= + y2 j
7 2 3 y= 41
101
= 0:601
168