Chapter 1

Probabilities

1.1 Conditional probability

In a group of 50 people,

• 30 persons speak english

• 20 persons speak german

• and 5 persons speak both english and german.

What is the probability that a person picked at random speaks english, given that the same
person also speaks german ?

1.2 Total probability theorem

A group of people contains

• 40% of persons having the french nationality

• 60% of persons having the german nationality.

We assume that no-one has the double nationality. We also know that

• 70% of the french are chess players

• 20% of the germans are chess players.

If a person picked at random can play chess, what is the probability that this person has the
german nationality ?

1.3 Independence of two events

A factory produces products in two independent phases. During phase 1, a defect of type A
occurs with a probability of 2%. During phase 2, a defect of type B occurs with a probability
of 8%. What is the probability that a product picked at random presents:
a) both defects
b) one and only one defect among both defects
c) no defect among both defects
d) at least one defect among both defects.

1
1.4 Expectation and variance of discrete r.v.s
A player has the choice between two strategies:

1. two independent dice rolls, each with equally likely outcomes in {1, 2, 3, 4, 5, 6}

2. a single dice roll, with equally likely outcomes in {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}.

The player having the highest number of points wins.

a) The player chooses the strategy with the highest expected number of points. Which
strategy is it ?
b) Calculate the variance of the number of points obtained with each strategy.

1.5 Binomial and Poisson p.m.f.s

You go to a party with 500 guests. What is the probability that exactly one other guest has
the same birthday as you ? Calculate this probability exactly and also approximately using
the Poisson p.m.f.
For simplicity, exclude birthdays on February 29.

1.6 Joint and marginal p.m.f.s

Consider a r.v. X (resp. Y ) whose range is {0, 1, 2} (resp. {0, 1, 2, 3}). The joint p.m.f. of X
1
and Y has the form pX,Y (x, y) = 42 (2x + y), 0 ≤ x ≤ 2, 0 ≤ y ≤ 3.
a) Find the marginal p.m.f.s of X and Y .
b) Find the probability of the event {X = 2, Y = 1}
c) Find the probability of the event {X ≥ 1, Y ≤ 2}.

1.7 Gaussian r.v.

The height of the students in your class is Gaussian distributed with mean 1.7 m and standard
deviation 5 cm. Indication: Φ(1) ≈ 0.8413 and Φ(2) ≈ 0.9772.
a) What is the probability that the height of a student is larger than 1.8 m
b) What is the probability that the height of a student is strictly smaller than 1.6 m
c) What is the probability that the height of a student is in the interval [1.65, 1.75] m.

1.8 Joint p.d.f., c.d.f. and change of variable

Two continuous r.v.s X and Y have the following joint p.d.f.
 1
 xy, 0 < x < 4, 1 < y < 5
fX,Y (x, y) = 96
0, otherwise.


a) Show that X and Y are independent.

b) Calculate the joint c.d.f. of X and Y .
c) Using the change of variable U = X + Y, V = X, calculate P (X + Y < 3).

2
1.9 Central limit theorem
In the game defined in Exercise 1.4,

1. Player 1 uses strategy 1

2. Player 2 uses strategy 2.

Using an approximation based on the central limit theorem, show that after n > 30 independent
games, the probability that the number of points collected by player 1 is larger than the number
of points collected by player 2, exceeds 50%.

1.10 Polling
At an election

• p is the actual proportion of people in favor of candidate A.

• 1 − p is the actual proportion of people in favor of candidate B.

We choose a sample of size n from the total population. Define the r.v.
(
1, if the i-th person in the sample is in favor of candidate A
Xi =
0, otherwise.

Assuming that the sample is chosen uniformly at random in the total population, the r.v.
Sn = X1 + · · · + Xn ∼ B(n, p). Let us define the empirical estimator of p as the r.v. Pn = Snn .
a) Assuming n > 30, np > 5 and n(1−p) > 5, use a Gaussian approximation of the binomial
distribution to show that
r !
p(1 − p)
P |Pn − p| ≤ t = 2Φ(t) − 1.
n

b) Three days before the election, the pollster obtains for n = 2500 an observed value for Pn
equal to p2500 = 0.48. Give a 95% confidence interval for the unknown value of p. Indication:
Φ(2) ≈ 0.9772. Can candidate A still reasonably hope to get elected ?

3
4
Chapter 2

Random processes

2.1 Poisson process

Arrival times of costumers in a store are modeled by a Poisson process with rate λ = 10
costumers per minute.
a) Let M be the number of costumers arriving between 9:00 and 9:10. What is the proba-
bility distribution of M ?
b) Let N be the number of costumers arriving between 9:10 and 9:15. What is the proba-
bility distribution of N ?
c) Calculate the expectation and the variance of the number of arrivals between 9:00 and
9:15.

2.2 Poisson interarrival times

Each morning, you decide to pull out of your driveway and make an illegal U-turn instead of
driving around the block. Police cars drive by according to a Poisson process of rate λ. You
decide to make a U-turn once you see that the road has been clear of police cars for τ time
units. Let N be the number of police cars you see, before you actually make the U-turn. Find
the expected value of N , as a function of λ and τ .

2.3 Markov process

Consider a discrete-time Markov process {Xn , n ∈ N}, taking values in the set of states {s1 , s2 }
and having the following state transition diagram:

1-b s1 s2 1-r

r
whith parameters b = 0.1 and r = 0.4.
a) Build the transition probability matrix P and show that the Markov chain is regular.
b) Find the steady-state probability vector, π.

5
2.4 Queueing
Packets arrive at a telecommunication network, where they are stored in a buffer and then
transmitted. The storage capacity of the buffer is m packets:

• if m packets are already present, any newly arriving packets are discarded

• otherwise, a newly arriving packet is stored in the buffer.

A each discrete time instant, exactly one of the following occurs:

• a new packet arrives, with probability b > 0

• one packet in the buffer completes transmission, with probability 0 < d 6= b

• none of the two previous events occurs.

a) Show that the number of packets in the buffer can be modeled by a birth-death process,
having the following state transition diagram. Is the corresponding Markov process regular ?

1-b 1-b-d 1-b-d 1-d

b b b b

0 1 … m-1 m

d d d d
b) Find the steady-state probabilities of the number of packets in the buffer, π = [π0 , π1 , . . . , πm ],
as a function of ρ = db .

6
Chapter 3

Estimation theory

3.1 Estimation of the mean and variance of a random

variable
Consider a random process formed by a collection of i.i.d. random variables X = (X1 , X2 , . . . , XN )T ,
such that ∀i, (
E[Xi ] = m
V [Xi ] = σ 2 .

and the corresponding observation set is x = (x1 , x2 , . . . , xN )T . We are interested in the natural
estimators of the mean and variance

 N
X
M̂ = Xi /N





i=1
 N
X
(Xi − M̂ )2 /N.

V̂ =




i=1

a) Calculate the expectation and the variance of the mean estimator.

b) Show that the proposed variance estimator is biased and find a remedy.

3.2 Estimation of the parameter of a Bernoulli r.v.

We consider the outcomes of N independent coin tosses, where p is the probability of a head
and 1 − p the probability of a tail. Define the r.v. Xi associated to the i-th coin toss such that
(
1, for a head
Xi =
0, for a tail.

Consider the following estimator of p

N
X
P̂ = Xi /N.
i=1

Show that P̂ is unbiased and that its variance decreases with N .

7
3.3 MVUE for the mean of Gaussian r.v.s
Consider a random process formed by a collection of i.i.d. random variables X = (X1 , X2 , . . . , XN )T ,
such that Xi ∼ N (m, σ 2 ), ∀i, and the corresponding observation set is x = (x1 , x2 , . . . , xN )T .
a) Assuming m is an unknown but deterministic constant, find the joint p.d.f. of X param-
eterized by m, denoted by fX (x; m). Show that fX (x; m) satisfies the regularity condition and
calculate the CRLB.
b) Use the second part of the CRLB theorem to find the MVUE for m.
c) Show that the maximum-likelihood estimate of m, m̂ML , coincides with the estimate
provided by the MVUE.

3.4 BLUE for the mean of Gaussian r.v.s

Consider a random process formed by a collection of i.i.d. random variables X = (X1 , X2 , . . . , XN )T ,
such that Xi ∼ N (m, σ 2 ), ∀i, and the corresponding observation set is x = (x1 , x2 , . . . , xN )T .
Asuming a linear observation model of the form

X = ms + W,

where s = (1, . . . , 1)T and W = (W1 , W2 , . . . , WN )T is a collection of i.i.d. random variables

such that Wi ∼ N (0, σ 2 ), ∀i. Find the BLUE for m, M̂BLUE . Show that it coincides with
the MVUE obtained in Exercice 3.3.

3.5 Bayesian estimation of the parameter of an exponen-

tial p.d.f.
Consider a random process formed by a collection of i.i.d. random variables X = (X1 , X2 , . . . , XN )T ,
such that Xi ∼ E(θ), ∀i, and the corresponding observation set is x = (x1 , x2 , . . . , xN )T . We
assign the following prior distribution for θ, considered as a random variable Θ:
(
λe−λθ , if θ ≥ 0
fΘ (θ) =
0, otherwise

for some λ ≥ 0.
a) Find the MAP estimate of θ, θ̂MAP .
b) Let θ̂ML be the maximum-likelihood estimate of θ. Show that as λ → 0, θ̂MAP → θ̂ML .
Give a theoretical interpretation.

3.6 Bayesian estimation of the mean of Gaussian r.v.s

Consider a random process formed by a collection of i.i.d. random variables X = (X1 , X2 , . . . , XN )T ,
such that Xi ∼ N (m, σ 2 ), ∀i, and the corresponding observation set is x = (x1 , x2 , . . . , xN )T .
We assign the following prior distribution for m, considered as a random variable M :

fM (m) = N (m0 , σ02 ).

a) Find the MAP estimate of m, m̂MAP .

8
b) Find the MMSE estimate of m, m̂MMSE .
c) What is the value of m̂MAP , when σ02 → 0. Give an interpretation.
d) What is the value of m̂MAP , when σ02 → +∞. Give a theoretical interpretation.