Probability Theory – Final Exam 2023 – Solutions

dr. A. Zocca

1st June 2023, 15:30 – 17:30 (+30 min for extra timers)

The exam consists of six exercises worth a total of 95 points = 10 + 15 + 25 + 15 + 10 + 20. The midterm
exam grade is calculated as 1 + (your points)/10 (the maximum is thus 10.5). The exam lasts for
2h (+ 30 min for extra timers). Do not forget to add the name of the formula you are using and/or
explanations next to your algebraic/numerical solutions. No book and no notes, you are allowed to use
only a simple calculator.

Exercise 1 [10 points = 2 + 3 + 5]

Let X be an exponentially distributed random variable with parameter 15 .
(a) [2 points] What does it mean that the random variable X is memoryless?
(b) [3 points] Calculate P(1 ≤ X ≤ 2).
x
(c) [5 points] Derive the density function of the random variable Y = g(X) with g(x) := e− 5 .

Solution:
(a) It means that P(X > s + t | X > t) = P(X > s) for every s, t ≥ 0.
1
(b) The density of X is given by fX (x) = 15 e− 5 x 1{x≥0} .
The required probability can be calculated as
Z 2
1 2 1
P(1 ≤ X ≤ 2) = fX (x)dx = [−e− 5 x ]21 = −e− 5 + e− 5 ≈ 0.1484.
1

(c) Solution using densities:

The function g maps [0, +∞) onto the interval (0, 1], is differentiable, one-to-one, and its derivative is
always nonzero.
x
The inverse of the function g(x) = e− 5 is thus defined only for y ∈ (0, 1]. and its algebraic expression is
g −1 (y) = −5 ln y.
We are thus applying the following formula (cf. fact 5.24)
(
fX (g −1 (y)) · |g′ (g−1
1
(y))|
if g −1 (y) exists and g ′ (g −1 (y)) ̸= 0,
fY (y) =
0 otherwise.

Applied to the functions above, this gives

(
1
fX (−5 ln y) · |− y5 |
if 0 < y ≤ 1,
fY (y) =
0 otherwise,

which further simplifies into (

1 if 0 < y ≤ 1,
fY (y) =
0 otherwise.

Alternative solution using CDFs:

In view of the fact that Y can take values only (0, 1], we can conclude that FY (y) = 1 for y ≥ 1 and
FY (y) = 0 for y ≤ 0. In the case y ∈ (0, 1], we have

FY (y) = P(Y ≤ y) = P(e−X/5 ≤ y) = P(−X/5 ≤ log y) = P(X ≥ −5 log y) = 1 − FX (−5 log y).
 and, hence,
d d −5
fY (y) = FY (y) = (1 − FX (−5 log y)) = −fX (−5 log y) · = · · · = 1.
dy dy y

Exercise 2 [15 points = 2 + 4 + 3 + 6]

Let X1 , . . . , X100 be independent standard normal random variables.
(a) [2 points] For a generic index i, what is the value of P(Xi > 0)?
(b) [4 points] Using the table provided, calculate the probability
 X1 + X2 + X3 + X4 
P −1≤ ≤1 .
4

(c) [3 points] Let Y be equal to the number of indices i ∈ {1, 2, . . . , 100} for which Xi > 0. Using (a),
conclude which is the distribution of Y .
(d) [6 points] Using the Central Limit Theorem and the continuity correction, give an approximation of
the probability P(45 < Y ≤ 52).

Solution:
(a) Since Xi is a standard normal r.v., whose density is symmetric around the origin, we can immediately
conclude that P(Xi > 0) = 12 .
(b) The sum of four independent normal r.v.’s is itself normal , and specifically X1 + X2 + X3 + X4 ∼
N (0, 4).
 X1 + X2 + X3 + X4   
P −1≤ ≤ 1 = P − 4 ≤ X1 + X2 + X3 + X4 ≤ 4
4
 X1 + X2 + X3 + X4 
=P −2≤ √ ≤2
4
= Φ(2) − Φ(−2)
= Φ(2) − (1 − Φ(2))
= 2Φ(2) − 1
≈ 2 · 0.9772 − 1 = 0.9544

(c) Thanks to the independence assumption about the Xi ’s , Y is a binomial distribution with n = 100
and, in view of (a), p = 12 . In symbols, Y ∼ Bin(100, 21 ) .
(d) Using the continuity correction, P(45 < Y ≤ 52) = P(45.5 ≤ Y ≤ 52.5). Y is a r.v. with mean
np = 50 and variance np(1 − p) = 25 . Therefore,
 45.5 − 50 Y − 50 52.5 − 50 
P(45.5 ≤ Y ≤ 52.5) = P √ ≤ √ ≤ √
25 25 25
≈ Φ(0.5) − Φ(−0.9)
= Φ(0.5) − (1 − Φ(0.9))
≈ 0.6915 − (1 − 0.8159) = 0.5074
Exercise 3 [25 points = 5 + 5 + 5 + 5 + 5]
Let (X, Y ) be the random vectors, whose components describe the two coordinates of a point distributed
uniformly at random over the set D ⊂ R2 defined as

D := {(x, y) ∈ R2 | 0 < y < ex , 0 < x < 2}.

(a) [5 points] Draw the set D, calculate its area, and derive an explicit expression for the joint probability
density fX,Y (x, y).
X
(b) [5 points] Calculate P(Y > e 2 ).
(c) [5 points] Calculate E[XY ]. Are X and Y independent?
(d) [5 points] Calculate the density of the random variable Z = X + Y .
(e) [5 points] Calculate Cov(2X − 1, −Y + 2) and Cov(Z, X).
Hint: You do not need the answer to question (d) to make these calculations!

Solution:
(a) The set D is drawn in the figure below.

0.5 1.0 1.5 2.0

Its area is given by Z Z 2 Z ex

Area(D) = 1 dxdy = 1 dxdy = [ex ]20 = e2 − 1.
D 0 0

1
The joint density corresponding to the uniform distribution over D is equal to fX,Y (x, y) = Area(D) =
1
e2 −1
.
X
(b) We can think of the event {Y > e 2 } as a subset of points in the plane, say B = {(x, y) ∈ D : ex/2 <
y < ex , 0 < x < 2} ⊂ D ⊂ R2 . Then,
Z
X
P(Y > e 2 ) = fX,Y (x, y) dxdy
B
Z 2 Z ex
1 1 h x x/2
i2
= 2
dydx = e − 2e
0 ex/2 e − 1 e2 − 1 0
1 − 2e + e 2 e−1
= = ≈ 0.4621.
e2 − 1 e+1
(c) We first calculate
Z
E[XY ] = x · y · fX,Y (x, y) dxdy
R2
Z 2 Z ex Z 2 ex
xy xy 2
= dydx = dx
0 0 e2 − 1 0 2 (e2 − 1) 0
"
#2
2
e2x x e2x x2 − 41
Z
= dx =
0 2 (e2 − 1) 2 (e2 − 1)
0
1 + 3e4
= ≈ 3.224
8 (e2 − 1)

X and Y are not independent.

To prove that X and Y are not independent, we first calculate
2 Z ex 2 e x 2
ex x
Z Z Z Z
x xy
EX = x · fX,Y (x, y) dxdy = dydx = dx = dx
R2 0 0 e2 − 1 0 e2 − 1 0 0 e2−1
2
ex (x e2

− 1) 1+
= = ≈ 1.313,
e2 − 1 0 e2 − 1

and
2 Z ex 2 ex 2
y2 e2x
Z Z Z Z
y
EY = y · fX,Y (x, y) dxdy = 2
dydx = dx = dx
R2 0 0 e −1 0 2 (e2 − 1) 0 0 2 (e2 − 1)
2
e2x

1
1 + e2 ≈ 2.097.


= =
4 (e2 − 1) 0 4
2
(1+e2 )·(1+e2 ) (1+e2 )
Therefore, EX · EY = (e2 −1)4
= 4(e2 −1)
≈ 2.754, so in particular E[XY ] ̸= EX · EY , so X and Y
are not independent.
(d) In view of the geometry of the subset D and the fact that Z = X + Y , we must have Z ∈ [0, 2 + e2 ].
The density fZ (z) must then be equal to zero for z < 0 and for z > 2 + e2 .
Consider
Z +∞ now the case z ∈ (0, 2 + e2 ). We calculate the required density using the formula fZ (z) =
f (x, z − x) dx which yields
−∞
Z 2
1
fZ (z) = 1 1 x dx.
0 e2 − 1 {z−x>0} {z−x<e }

Let x∗ (z) ∈ R be the unique real number that solves the equation z − x = ex . The fact that this point
exists and is unique can be proved graphically by checking for the intersection of the line y = z − x and
of the function y = ex . The integral above then becomes
Z 2
1
fZ (z) = 2
1{x<z} 1{x>x∗ (z)} dx
0 e −1
Z min{2,z}
1
= 2
dx
max{0,x∗ (z)} e − 1

z
 e2 −1
 if z ≤ 1,
1 ∗ 1 ∗
= 2 (min{2, z} − max{0, x (z)}) = e2 −1 (z − x (z)) if 1 < z ≤ 2,
e −1 
(2 − x∗ (z)) if 2 < z ≤ 2 + e2 .
 1
e2 −1

The plot of fZ (z) looks as follows:

 0.25

0.20

0.15

0.10

0.05

-2 2 4 6 8 10

(e) Using the bilinearity of the covariance operator , we can conclude that

Cov(2X − 1, −Y + 2) = −2 · Cov(X, Y ) = −2(E[XY ] − EX · EY ) ≈ −2 · (3.224 − 2.754) [= −0.94].

Using again the bilinearity of the covariance operator, we can conclude that

Cov(Z, X) = Cov(X + Y, X) = Cov(X, X) + Cov(X, Y ) = Var(X) + Cov(X, Y )

To obtain the final answer, we need to calculate the variance of X, for which we need to obtain the
second moment of X (since we already calculated EX). Thus,
Z 2 Z ex Z 2  2 ex Z 2 x 2
x2
Z
2 2 x y e x
E(X ) = x · fX,Y (x, y) dxdy = 2
dydx = 2
dx = 2
dx
R2 0 0 e −1 0 e −1 0 0 e −1
"
#2
ex x2 − 2x + 2
= =2
e2 − 1
0

The variance is then Var(X) = E(X 2 ) − (EX)2 ≈ 2 − 1.313 = 0.687 and Cov(Z, X) ≈ 0.687 + 0.47 = 1.157.

Exercise 4 [15 points = 4 + 5 + 6]

A continuous random variable X has probability density function f (x) given by:

5 5x
3e
 if x ≤ 0,
f (x) := c x 2 if 0 < x ≤ 21 ,

 13
64x4
if x > 12 .

(a) [4 points] Find the value of c such that f is a well-defined probability density.
(b) [5 points] Calculate the expected value of X and its variance.
X2
(c) [6 points] Derive the probability density function fY (y) of the random variable Y = 3 .

Solution:
R +∞
(a) We must have −∞ f (x) dx = 1, therefore
1
Z +∞ Z 0 Z Z +∞
5 5x 2 13 1 c 13
1= f (x) dx = e dx + cx2 dx + 4
dx = + + .
−∞ −∞ 3 0 1 64x 3 24 24
2
Solving for c, gives c = 3. This is a valid value because the resulting density is always non-negative.
(b)
R +∞
Using the formula EX = −∞ x · f (x) dx , we get
1
Z 0 Z Z +∞
5 2 13 1 3 13 371
EX = x e5x dx + 2
x · 3x dx + x 4
dx = − + + = ≈ 0.3865.
−∞ 3 0 1
2
64x 15 64 32 960

R +∞
We first calculate the second moment of X. Since E(X 2 ) = −∞ x2 · f (x) dx , we get
1
Z 0 Z Z +∞
5 2 13 2 3 13 271
E(X 2 ) = x2 e5x dx + x2 · 3x2 dx + x2 4
dx = + + = ≈ 0.4516.
−∞ 3 0 1 64x 75 160 32 600
2

271 371 2 55723

Therefore, Var(X) = E(X 2 ) − (EX)2 =


600 − 960 = 184320 ≈ 0.302.
(c)
Solution using CDF:
We first derive the CDF FY (y). For y ≤ 0, clearly FY (y) = P(Y ≤ y) = P(X 2 ≤ 3y) = 0 , and we obtain
d
the pdf fY (y) as fY (y) = dy FY (y), so that

fY (y) = 0, ∀ y ≤ 0.

For y > 0, we can argue as follows:

FY (y) = P(Y ≤ y) = P(X 2 ≤ 3y)

p p
= P(− 3y ≤ X ≤ 0) + P(0 ≤ X ≤ 3y)
Z 0 Z √3y
= √ fX (x) dx + fX (x) dx
− 3y 0
√ √
Z 0 Z min{ 3y, 12 } Z 3y
5 5x 2 13
= √
e dx + 3x dx + √
dx
− 3y 3 0 min{ 3y, 12 } 64x4
 5x 0
 imin{√3y, 1 }  √3y
e h 13 2
= + x3 + −
3 − 3y√ 0 192x3 min{√3y, 1 }
2
(
1 −5
√
3y
√ 3/2 1
(1 − e ) + 3 3y if 0 ≤ y ≤ 12 ,
= 31 −5
√
3y 1 13 13 1
3 (1 − e ) + 8 + 24 − 576√3y3/2 if y > 12 .

We can then obtain the expressions of the pdf fY (y) for y > 0 by taking the first derivative of FY (y), i.e.,
 √
5e−5 3y 9p 1
if 0 < y ≤ 12


 2 √ + 3y ,
d 
3 3y
√
2
fY (y) = FY (y) =
dy  5e−5 3y 13 1
 2√ + if y > 12 .

 p
3 3y 384 3y 5

Solution using densities:

2
The function g(x) = x3 is differentiable everywhere and its derivative g ′ (x) = 23 x is always nonzero except
for the point x = 0.
We are thus applying the following formula (cf. fact 5.27)
1
 X

 fX (x) · ′ if at least one such x exists,
fY (y) = x:g(x)=y, g′ (x)̸=0 |g (x)|

0 otherwise.

 x2
For the specific function g(x) = 3 , this formula gives

1
 X

 fX (x) · 2 if y > 0,
fY (y) = x:x2 =3y 3x

0 if y ≤ 0.

 p
fX (− 3y) · 1 p 1
2√ + fX ( 3y) · 2√ if y > 0,
= − 3 3y 3 3y
0 if y ≤ 0.


Focusing only on the y > 0 case, we need to distinguish two cases, depending on whether the argument
√ √
3y of fX (·) is above or below 12 , since the density fX is defined differently. Solving 3y = 12 yields the
1
threshold value y = 12 . Then,
√ √
fX (− 3y) fX ( 3y)

1
 √ + 2√ if 0 < y ≤ 12 ,
− 23 √3y

3y

fY (y) = 3 √
fX (− 3y) fX ( 3y) 1
√ + 2√ if y > 12 ,


− 23 3y

3 3y
 5 −5√3y √
 3e√ 3( 3y)2 1
+ 2√ if 0 < y ≤ 12 ,

 2 3y

3y
3
= 5 −5√3y 3
13
√
3e 64( 3y)4
 1
 2√ + 2√ if y > 12 ,


3 3y 3 3y
 √
5e−5 3y 9p 1
if 0 < y ≤ 12

 2√ + 3y ,


3 3y
√
2
=
 5e−5 3y 13 1
 2√ + if y > 12 .

 p
3 3y 384 3y 5

Exercise 5 [10 points = 5 + 5]

To celebrate the end of the course, 10 Business Analytics students and 6 Mathematics students go out for
dinner to an Italian restaurant, where they randomly seat at a round table with 16 seats. Let X be the
number of Business Analytics students with exactly one neighbor who is a Business Analytics student
and one neighbor who is a Mathematics student.
(a) [5 points] Express the random variable X as a sum of suitable binary random variables.
(b) [5 points] Use the representation in (a) to calculate the expected value of X.

Solution:
First solution method
(a) The random variable X rewrites as X = X1 + · · · + X16 where Xi is the binary random variable
defined as
(
1 if BA student is seated at chair i with one Math and one BA neighbor,
Xi =
0 otherwise.

These 16 random variables are identically distributed, by symmetry of the seats. .

(b) Using the linearity of expectation, EX = E(X1 + · · · + X16 ) = EX1 + · · · + EX16 = 16 · EXi . Since
all Xi are identically distributed (cf. (a)), we can calculate
10 9 6 9
EXi = P(Xi = 1) = · · ·2= ,
16 15 14 28
since on chair i there is a BA student w.p. 10 9 6
16 , and with probability 15 · 14 his/her two neighbors are one
Math and one BA, and the factor 2 accounts for the two possible permutations of these two neighbors.
9 36
Therefore EX = 16 · EXi = 16 · 28 = 7 ≈ 5.143.
Second solution method
Number the BA students from 1 to 10. The random variable X rewrites as X = Y1 + · · · + Y10 where Yi
is the binary random variable defined as
(
1 if the i-th BA student has one Math and one BA neighbor,
Yi =
0 otherwise.
9 6 18
In this case, EYi = P(Yi = 1) = 15 · 14 ·2= 35 , which then leads to the same answer EX = 10 · EYi =
· · · = 36
7 ≈ 5.143.

Exercise 6 [20 points = 5 + 5 + 5 + 5]

Suppose the joint probability density of X and Y is given by
  
 −2 xy +y −1
4e y if x, y > 0,
fX,Y (x, y) :=
0 otherwise.

(a) [5 points] Derive the marginal density fY (y). What is the distribution of Y ?
(b) [5 points] Derive the conditional density fX|Y (x|y).
(c) [5 points] Calculate P(X > 1 | Y = y).
(d) [5 points] Calculate E(X | Y = y).
Hint: You can use (without proving it) that for every a, b ̸= 0 the antiderivative of the function
a
ax · ebx is the function 2 ebx (bx − 1).
b

Solution:
R +∞
(a) Since fY (y) = −∞ fX,Y (x, y) dx, we get that for y ≤ 0 fY (y) = 0 , while for y > 0
Z +∞      +∞
x
−2 +y −1 −2 x +y
fY (y) = 4e y
y dx = −2e y
= 2e−2y .
−∞ −∞

Therefore Y is an exponential distribution with parameter λ = 2.

(b) The conditional density can be calculated only for y > 0 (since otherwise fY (y) = 0). For such values
of y,
fX,Y (x, y)
fX|Y (x|y) =
fY (y)
−2 x
(
2e y y −1 if x > 0,
=
0 otherwise.

(c) We can calculate

Z +∞ Z +∞
−2 x − 2x +∞
h i
P(X > 1 | Y = y) = fX|Y (x|y) dx = 2e y −1
y = −e y = e−2/y .
1 1 1

Alternative solution:
The same answer could also be obtained by noticing from (b) that r.v. X|Y = y is exponentially distributed
with parameter y2 and using the well-known CDF of the exponential distribution.
(d) We can calculate
Z +∞ Z +∞  +∞
−2 x
y −1
1 − 2x y
E(X | Y = y) = x · fX|Y (x|y) dx = x · 2e y = − e y (2x + y) = .
−∞ 0 2 0 2

Alternative solution:
The same answer could also be obtained by noticing from (b) that r.v. X|Y = y is exponentially
distributed with parameter λ = y2 and using the well-known formula for the expected value of the
exponential distribution, we get E(X|Y = y) = λ1 = y2 .