Chap0 Discrete

Table of contents
Discrete Random Variable
anis.rezguii@gmail.com
INSAT
Mathematics Department
September 20, 2022
anis.rezguii@gmail.com Discrete Random Variable

Expected learning outcomes
Definitions
Usual probability distributions
Table of contents
1 Expected learning outcomes
2 Definitions
The Mass Distribution Function (MDF)
The cumulative distribution function CDF
The mathematical expectation
The variance and the standard deviation
3 Usual probability distributions
The uniform distribution
The Binomial distribution
The Hypergeometric distribution
The Geometric distribution
The Poisson distribution

Definitions
After this lecture the student should:
Understand the concept of discrete random variable (d.r.v).
Understand the intrinsic Mass Distribution Function (MDF)

associated to a d.r.v
Know usual examples of MDF.
Chose the right model for a given random situation.

Definitions
Definition
Definition
Let (Ω, P) be a probability space describing a random experiment
and X a mapping
X : Ω −→ R
ω 7−→ X (ω).
X is said to be a discrete random variable (d.r.v) if its range,

X (Ω) is a discrete subset of R.

Definitions
Discrete subset of R
A discrete subset of R is any subset which contains only isolated

points.
the subset {0, 1, 2, 3} is a discrete subset of R.

any finite subset is a discrete set.
the subset of all integers N is a discrete set.
the set of rational numbers Q is not a discrete set.
Any interval of R is not a discrete subset of R.

Definitions
Example of d.r.v
Consider the experiment of tossing two perfect coins. The

associated sample space is Ω = {H, T }2 = {HH, TT , HT , TH}
where ”H” denotes the ”head” face and ”T” denotes the ”tail”
face. Let X be the number of heads obtained after each toss.
It is clear that its values set is X (Ω) = {0, 1, 2} so X is a d.r.v.
The d.r.v X can be represented as follows:
Ω HH TT HT TH
X 2 0 1 1

Definitions
Definition
We associate to a given d.r.v, X , its The Mass Distribution
Function (MDF) denoted by PX and defined by the mapping:
PX : X (Ω) −→ [0, 1]
xi 7−→ PX (xi ) = P{X = xi }
where {X = xi } = {ω ∈ Ω : X (ω) = xi }

Definitions
Table representation of the MDF
In general we represent the MDF of a d.r.v X , PX , by the following

table:
X (Ω) PX
x1 PX (xi )
.. ..
. .

Definitions
Graphical representation of the MDF
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0 1 2

Definitions
Example
We reconsider the same example of tossing two perfect coins and
where X was the number of heads. To get its MDF PX we need to
evaluate:
PX (0) = P{X = 0} = P{TT } = P{T } × P{T } = 21 × 12 = 14 ,
note that we have used the independence between the two
tosses.
PX (1) = P{X = 1} = P{TH, HT } = P{TH} + P{HT } =
1 1 1
4 + 4 = 2.
PX (2) = P{X = 2} = P{HH} = P{H} × P{H} = 21 × 12 = 14 .

Definitions
Example
We reconsider the same example of tossing two perfect coins and
where X was the number of heads. To get its MDF PX we need to
evaluate:
PX (0) = P{X = 0} = P{TT } = P{T } × P{T } = 21 × 12 = 14 ,
note that we have used the independence between the two
tosses.
PX (1) = P{X = 1} = P{TH, HT } = P{TH} + P{HT } =
1 1 1
4 + 4 = 2.
PX (2) = P{X = 2} = P{HH} = P{H} × P{H} = 21 × 12 = 14 .
X (Ω) PX
0 0.25
PX is given by ⇝
1 0.5
2 0.25
Definitions
Notes
X X
1 Note that PX (xi ) = P{X = xi } = P(Ω) = 1.
i i

Definitions
Notes
X X
i i
2 Note the similarity between a discrete random variable and a
discrete statistical variable:
X(Ω) PX xi fi
x1 PX (xi ) x1 f1
.. .. .. ..
. . . .
P P
1 100%

Definitions
Notes
X X
i i
2 Note the similarity between a discrete random variable and a
discrete statistical variable:
X(Ω) PX xi fi
x1 PX (xi ) x1 f1
.. .. .. ..
. . . .
P P
1 100%
3 The latter similarity is used to model a statistical phenomenon
by a theoretical d.r.v and then to generate artificial data by
computer simulation without any cost. This is very important
for prediction and making decision.

Definitions
Definition
To any discrete random variable X we could associate its
cumulative distribution function defined as follows:
FX : R −→ [0, 1]
x 7−→ FX (x) = P{X ≤ x}
where {X ≤ x} = {ω ∈ Ω : X (ω) ≤ x}.

Definitions
Example
Again, we reconsider the same example of tossing two perfect coins

where X was the number of heads, the associated cumulative
distribution function is as follows:
1 if x < 0 then P{X ≤ x} = P{∅} = 0

2 if 0 ≤ x < 1 then P{X ≤ x} = P{X = 0} = 0.25
3 if 1 ≤ x < 2 then
P{X ≤ x} = P{X = 0 or X = 1} = 0.25 + 0.5 = 0.75
4 if x ≥ 2 then P{X ≤ x} = P{X = 0 or X = 1 orX = 2} =
0.25 + 0.5 + 0.25 = 1

Definitions
The cumulative distribution function is summarized as follows:



 0 if x < 0




 0.25 if 0 ≤ x < 1


FX =
0.75 if 1 ≤ x < 2








if x ≥ 2

1

Definitions
Graphical representation of a CDF
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−1 −0.5 0 0.5 1 1.5 2 2.5 3

Definitions
Note 1
Let X a given d.r.v and denote by PX , FX its MDF and its CDF
respectively. The CDF FX satisfies:
i. FX is non-decreasing.
ii. FX is càdlag i.e continuous to the right and bounded
to the left.
iii. lim FX (x) = 0 and lim FX (x) = 1.
x→−∞ x→+∞

Definitions
Note 2
It is very important to know that PX and FX are equivalent in the

sense that the knowledge of one suffices to determine the other. If
we have PX it is clear that we can determine FX , just by definition.
Now suppose we have the cumulative distribution function, FX , we
use the following rule to determine PX :
P{X = a} = FX (a) − FX (a− ).
where FX (a− ) = lim FX (x).

x→a−

Definitions
The Mathematical expectation
Definition
Let X be a d.r.v and X (Ω) = {xi } its set of values.
The mathematical expectation of X , denoted by E(X ), is the
quantity: X
E(X ) = xi PX (xi ).
i

Definitions
Example
Reconsider the last example where X was the number of heads, we
get:
1 1 1
E(X ) = 0 × +1× +2×
4 2 4
1 1
= + =1
2 2

Definitions
Example
Reconsider the last example where X was the number of heads, we
get:
1 1 1
E(X ) = 0 × +1× +2×
4 2 4
1 1
= + =1
2 2
Note: If any random experiment is repeated many times we

expect, in the ”usual” cases, that most of X’s values are ”close” to
the mathematical expectation E(X ). Thus the mathematical
expectation of a d.r.v is considered as a parameter of central
tendency.

Definitions
Exercise
Let X be a d.r.v with three possible values {−1, 0, 1} such that its
MDF is as follow:
X (Ω) PX
−1 q
0 1/2
1 p
and its expectation E(X ) = 0.
1 Determine the MDF of X , PX .
2 Suppose now P{X = 0} is also unknown, could you determine
PX ?

Definitions
Solution
1 On one hand E(X ) = −1 × p + 0 × 21 + 1 × q = 0 which

means p = q, on the other hand p + 12 + q = 1, then we get
p = q = 14 .
2 We obtain two equations with three unknowns, so there is
infinitely many possibilities.

Definitions
Solution
1 On one hand E(X ) = −1 × p + 0 × 21 + 1 × q = 0 which

means p = q, on the other hand p + 12 + q = 1, then we get
p = q = 14 .
2 We obtain two equations with three unknowns, so there is
infinitely many possibilities.
Note:
In general the only knowledge of the expectation does not
determine the MDF of the random variable. However there are
many interesting particular cases, of random variables, where it is
the case, see section usual MDF.

Definitions
The Markov Inequality
The next theorem gives a rough estimation of some probabilities by

the only use of the expectation.
Theorem
Let (Ω, Σ, P) be a probability space, X a positive d.r.v and λ a
positive real number. Then
E(X )
P{X ≥ λ} ≤ .
λ

Definitions
The variance
Definition
Let X be a d.r.v and X (Ω) = {xi } its range. Its variance, denoted
by V(X ) or σX2 , is the non negative quantity:
X
V(X ) = σX2 = (xi − E(X ))2 PX (xi ).
i
Its standard deviation is

p
σX = V(X ).

Definitions
Notes
1 Suppose that a given random experiment is repeated many

times, we expect, in the usual cases, that most of X’s values
lies in the interval [E(X ) − σX , E(X ) + σX ].
2 The variance as well as the standard deviation are parameters

of dispersion. In fact they give idea about how expected data
are dispersed around the mathematical expectation.

Definitions
The Tchebychev Inequality

Theorem
Let (Ω, Σ, P) be a probability space, and X a d.r.v, then for any
λ > 0 we have
V(X )
P{|X − E(X )| ≥ λ} ≤ .
λ2

Definitions
The Tchebychev Inequality

Theorem
Let (Ω, Σ, P) be a probability space, and X a d.r.v, then for any
λ > 0 we have
V(X )
P{|X − E(X )| ≥ λ} ≤ .
λ2
Proof:
Let λ > 0
P{|X − E(X )| ≥ λ} = P{|X − E(X )|2 ≥ λ2 }
and so by using Markov inequality, we get
E{|X − E(X )|2 } V(X )
P{|X − E(X )| ≥ λ} ≤ = ,
λ2 λ2
which finishes the proof.
Definitions
Note
Tchebychev inequality is often stated as follow:
1
P{|X − E(X )| ≥ nσ} ≤
n2
which can be read
1
P {X ∈ [E(X ) − nσ, E(X ) + nσ]} > 1 −
n2
If for example we take n = 6 we obtain that for any random
variable X , which means for any phenomenon we have the
following estimation:
1
P {X ∈ [E(X ) − 6σ, E(X ) + 6σ]} > 1 − = 97.22%
62
.
Definitions
Note
We have a different, but, easy to prove and useful different

formulation of the variance: For any discrete random variable X :
X 2
V(X ) = xi − E(X ) P{X = xi }
i
X
= xi2 − 2E(X )xi + E(X )2 P{X = xi }
i
= E(X 2 ) − E(X )2 .

Expected learning outcomes The Binomial distribution
Definitions The Hypergeometric distribution
Usual probability distributions The Geometric distribution
Definition
Let X be a d.r.v with a finite range, X (Ω) = {x1 , · · · , xn }, for
some n ∈ N. We say that X follows the uniform distribution on
{x1 , · · · , xn } and we denote
X ,→ U{x1 , · · · , xn }
if
1
P{X = x1 } = · · · = P{X = xn } = .
n

Examples
1 If we toss a perfect coin and set X = 0 for ”Tail” and X = 1

for ”Head”. It is easy to check that
X ,→ U{0, 1}.
2 If we throw a perfect die and set X as the obtained digit. It is
easy to check that
X ,→ U{1, · · · , 6}.

Examples
1 If we toss a perfect coin and set X = 0 for ”Tail” and X = 1

for ”Head”. It is easy to check that
X ,→ U{0, 1}.
2 If we throw a perfect die and set X as the obtained digit. It is
easy to check that
X ,→ U{1, · · · , 6}.
Note:
In case of uniform distribution we have the following formula:
#A
P{X ∈ A} = ,
n
for any A ⊂ X (Ω).

Expectation and variance
Let X ,→ U{x1 , · · · , xn }, we can check easily that:

i.
x1 + · · · + xn
E(X ) = .
n
ii.
n n
1X 2 1 X 2
V(X ) = xi − 2 xi .
n n
i=1 i=1

Let X ,→ U{x1 , · · · , xn }, we can check easily that:

i.
x1 + · · · + xn
E(X ) = .
n
ii.
n n
1X 2 1 X 2
V(X ) = xi − 2 xi .
n n
i=1 i=1
Note: Note the similarity with the sample mean and the sample
variance of any set of data.

The Binomial distribution
Definition
Let X be a d.r.v, n ∈ N and p ∈]0, 1[. X is said to follow a
binomial distribution with n repetitions and a ”success” probability
p, or just with parameters (n, p), if:
i) X (Ω) = {0, 1, · · · , n},
ii) for any k = 0, 1, · · · , n, P{X = k} = Cnk p k (1 − p)n−k .
We denote
X ,→ β(n, p).

Typical Example
Consider the experiment of throwing a perfect die 10 times and X

the number of occurrences of the digit ”6”. We can check easily
that
1
X ,→ β(10, ).
6

Checking
− It is clear that X (Ω) = {0, 1, · · · , 10}.

− The eventn o n o
{X = 0} = ”6” doesn’t appear = 10 times {1, · · · , 5} ,
this implies that:
5 5 5 10
P{X = 0} = P{1, · · · , 5} × · · · × P{1, · · · , 5} = × · · · × = .
n o6 6 6
− The event {X = 1} = ”6” appears only once , then
1 5 9
P{X = 1} =10×P{X = 6} × (P{1, · · · , 5})9 = 10 × × ,
6 6
where 10 in the formula denotes the 10 possible positions of
occurrence of the digit ”6”.

When we use the binomial model?
In general we use the binomial distribution when we repeat a given

random experiment n times, independently, and when X denotes
the number of occurrences of a particular event called ”success or
failure” with a given probability p ∈]0, 1[. In that case we have
X ,→ β(n, p).

Let X be a d.r.v following a binomial distribution β(n, p) for

n ∈ N∗ and p ∈]0, 1[:
i. its mathematical expectation is:
E(X ) = n × p,
ii. its variance is:
V(X ) = n × p × (1 − p).

The Hypergeometric distribution
Definition
Let N1 , N2 and n three integers, such that N1 + N2 > n, and X a
given discrete random variable. It is said to follow a
hypergeometric distribution of parameters (N1 , N2 , n), and denoted
X ,→ HG(N1 , N2 , n) if
i) X (Ω) = {0, 1, · · · , n}.
ii) for any k = 0, · · · , n
CNk 1 CNn−k
2
P{X = k} = .
CNn1 +N2

Typical example
X is the number of red balls drawn after n successive draws of one

ball, without replacement, from a box that contains N1 red balls
and N2 white balls. Check that X follows a hypergeometric
distribution, X ,→ HG(N1 , N2 , n).

Expectation and Variance
On can check that

E(X ) = np
and
N −n
V(X ) = np(1 − p)
N −1
N1
where p = and N = N1 + N2 .
N

Definition
Let p ∈]0, 1[ and X a discrete random variable. X is said to follow
a geometric distribution of parameter p, and is denoted
X ,→ G(p), if
i) X (Ω) = N∗ .
ii) For any k ≥ 1,
P{X = k} = p(1 − p)k−1 .

Typical example
Let X be the position of the first occurrence of the face ”head”

when we toss a balanced coin infinitely many times. Check that X
1
follows a geometric distribution of parameter .
2

The expectation and the variance
It is easy to show that

1
E(X ) =
p
and
1−p
V(X ) = .
p2

Definition
Let λ ∈ R∗+ and X a d.r.v, we say that X follows a Poisson
distribution with parameter λ, and we denote X ,→ P(λ), if
i) X (Ω) = N.
ii) for any k ≥ 0,
λk
P{X = k} = e −λ .
k!

Typical example
In general, the Poisson distribution is used to model the

number of clients (customers) which arrive during a time
period T , with a frequency f , to ask for a given service.
In that case their random number can be modeled by a

Poisson distribution of parameter the average number of
clients λ = T × f .

Let λ ∈ R∗+ and X ,→ P(λ). We prove that:

i.
E(X ) = λ.
ii.
V(X ) = λ.

Exercise
Consider the records of highway accidents between Makkah and AL

Madinah during 50 days.
accidents number number of days
0 21
1 18
2 7
3 3
4 1
1. Compute the average number of accident, m.

Still the exercise
2. Suppose that the theoretical probability distribution of the

accidents number is of Poisson type with parameter λ.
Determine an approximation of λ and compute the probability
that no accident happen, only one accident happen, 2
accident ...
3. Compute the theoretical number of days ti such that we
observe i accidents for i = 0, · · · , 4. Compare theoretical and
empirical results.
4. How many days without accidents we will observe during a
year.

Solution
1. The average m is just the total number of accidents divided
by the number of days.
1 × 18 + 2 × 7 + 3 × 3 + 4 × 1
m= = 0, 9.
50
2. Let X ,→ P(λ), if the latter is a model for the accidents

number we must have E(X ) = m. Thus we have
λ = m = 0, 9. We obtain the probabilities that happen i
accidents during 50 days for i = 0, · · · , 4 by using the formula
(0, 9)i
P{X = i} = e −0,9 . And so
i!
X 0 1 2 3 4
PX 40, 6% 36, 5% 16, 4% 4, 9% 1, 1%
Still solution
3. The theoretical number of days where there was exactly i
accidents is ti = PX (i) × 50.
i 0 1 2 3 4
ti 20,328 18,295 8,233 2,469 0,555
It is clear that the theoretical frequency distribution is very
close to the experimental (statistical) one and so, we have a
good reason to trust the Poisson model.
4. Without waiting for a year, we can estimate the number of
days without any accidents
accidents number per year = 365 × PX (0) = 148, 397.

Chap0 Discrete

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chap0 Discrete

Uploaded by

Copyright:

Available Formats

Table of contents

Discrete Random Variable

September 20, 2022

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

Expected learning outcomes

After this lecture the student should:

Understand the concept of discrete random variable (d.r.v).

Understand the intrinsic Mass Distribution Function (MDF)

Know usual examples of MDF.

Chose the right model for a given random situation.

anis.rezguii@gmail.com Discrete Random Variable

X is said to be a discrete random variable (d.r.v) if its range,

anis.rezguii@gmail.com Discrete Random Variable

A discrete subset of R is any subset which contains only isolated

the subset {0, 1, 2, 3} is a discrete subset of R.

anis.rezguii@gmail.com Discrete Random Variable

Consider the experiment of tossing two perfect coins. The

It is clear that its values set is X (Ω) = {0, 1, 2} so X is a d.r.v.

The d.r.v X can be represented as follows:

anis.rezguii@gmail.com Discrete Random Variable

The Mass Distribution Function (MDF)

anis.rezguii@gmail.com Discrete Random Variable

Table representation of the MDF

In general we represent the MDF of a d.r.v X , PX , by the following

anis.rezguii@gmail.com Discrete Random Variable

Graphical representation of the MDF

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

The cumulative distribution function CDF

where {X ≤ x} = {ω ∈ Ω : X (ω) ≤ x}.

anis.rezguii@gmail.com Discrete Random Variable

Again, we reconsider the same example of tossing two perfect coins

1 if x < 0 then P{X ≤ x} = P{∅} = 0

anis.rezguii@gmail.com Discrete Random Variable

The cumulative distribution function is summarized as follows:

anis.rezguii@gmail.com Discrete Random Variable

Graphical representation of a CDF

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

It is very important to know that PX and FX are equivalent in the

P{X = a} = FX (a) − FX (a− ).

where FX (a− ) = lim FX (x).

anis.rezguii@gmail.com Discrete Random Variable

The Mathematical expectation

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

Note: If any random experiment is repeated many times we

anis.rezguii@gmail.com Discrete Random Variable

anis.rezguii@gmail.com Discrete Random Variable

1 On one hand E(X ) = −1 × p + 0 × 21 + 1 × q = 0 which

anis.rezguii@gmail.com Discrete Random Variable

1 On one hand E(X ) = −1 × p + 0 × 21 + 1 × q = 0 which

anis.rezguii@gmail.com Discrete Random Variable

The Markov Inequality

The next theorem gives a rough estimation of some probabilities by

anis.rezguii@gmail.com Discrete Random Variable

Its standard deviation is

anis.rezguii@gmail.com Discrete Random Variable

1 Suppose that a given random experiment is repeated many