Chapter 5: Usual discrete and continuous probability laws
By definition, discrete random variables take discontinuous integer values over a
given interval.
B.5.1 Uniform Law
A probability distribution follows a uniform law when all the values taken by the random
variable are equally likely. If is the number of different values taken by the va Then:
1
∀𝑖, ℙ(𝑋 = 𝑥𝑖 ) =
𝑛
Example
We roll a die, the law of probability is:
𝑥𝑖 1 2 3 4 5 6
ℙ(𝑋 = 𝑥𝑖 ) 1 1 1 1 1 1
6 6 6 6 6 6
1 4 6 4 1 1
𝐸(𝑥) = 1. + 2. + 3. + 3. + 5. + 6.
6 6 6 6 6 6
21
𝐸(𝑥) = = 3,5
6
1 4 6 1 4
𝑉𝑎𝑟(𝑥) = (1 − 3,5)2 6 + (2 − 3,5)2 6 + (3 − 3,5)2 6 + (4 − 3,5)2 6 + (5 − 3,5)2 6+(6 −
1
3,5)2 6
17.5
𝑉𝑎𝑟(𝑥) = =2,92
6
Cas particuliers
If the values of the r.v X corresponding to the rank 𝑖 = 𝑥𝑖
∀𝑖𝜖[1, 𝑛]
𝑛+1
𝐸 (𝑥 ) =
2
And
𝑛2 − 1
𝑉𝑎𝑟(𝑥) =
12
1
Example (previous)
𝑛=6
𝑛+1 6+1
𝐸(𝑥) = =
2 2
7
𝐸(𝑥) = 2=3.5
𝑛2 −1 36−1
𝑉𝑎𝑟(𝑥) = =
12 12
35
𝑉𝑎𝑟(𝑥) = 12=2.92
B.5.2 Bernoulli’s law
Let X be a random variable, and let a universe Ω made up of two eventualities, S for
success and E for failure:
𝛺 = {𝐸, 𝑆}
Such as during a test, if S is achieved, 𝑋 = [Link] E is realized 𝑋 = [Link] call
Bernoulli variable the random variable X such that:
𝑋: 𝛺 → ℝ
𝑋(𝛺) = {0,1}
The probability law associated with the Bernoulli variable X such that:
ℙ(𝑋 = 0) = 𝑞
ℙ(𝑋 = 1) = 𝑝
With : 𝑝 + 𝑞 = 1
Is called Bernoulli's law noted ℬ(1, 𝑝)The
expectation of the Bernoulli variable is:
𝐸(𝑥) = ∑2𝑖=1 𝑝𝑖 . 𝑥𝑖
𝐸(𝑥) = (0. 𝑞) + (1. 𝑝)
𝐸(𝑥) = 𝑝
The variance of the Bernoulli variable is:
2
2
𝑉𝑎𝑟(𝑋) = ∑[𝑥𝑖 − 𝐸(𝑋)]2 ℙ(𝑋 = 𝑥𝑖 )
𝑖=1
𝑉𝑎𝑟(𝑋) = (0 − 𝑝)2 𝑞 + (1 − 𝑝)2 𝑝
𝑉𝑎𝑟(𝑋) = 𝑝2 𝑞 + 𝑞 2 𝑝
𝑉𝑎𝑟(𝑋) = 𝑝𝑞(𝑝 + 𝑞)
𝑉𝑎𝑟(𝑋) = 𝑝𝑞
Method 2
𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − [𝐸(𝑋)]2
𝑉𝑎𝑟(𝑋) = (02 . 𝑞 + 12 . 𝑝) − 𝑝2
𝑉𝑎𝑟(𝑋) = 𝑝 − 𝑝2 = 𝑝(1 − 𝑝)
𝑉𝑎𝑟(𝑋) = 𝑝𝑞
B.5.3 Binomial Law
The binomial distribution is one of the most common probability distributions
encountered in statistics. Either the application
𝑆𝑛 : 𝛺 𝑛 → ℝ𝑛
With:
𝑆𝑛 = 𝑋1 + 𝑋1 + ⋯ + 𝑋𝑖 + ⋯ + 𝑋𝑛
𝑋𝑖 ::Bernoulli variable.
𝑆𝑛 ::Binomial variable, represents the number of successes obtained during the repetition of n identical
and independent tests.
The probability law followed by the sum of n Bernoulli variables is the binomial law with parameters
n and p.
𝑆𝑛 : 𝛺 𝑛 → ℝ𝑛
𝑛
𝑆𝑛 = ∑ 𝑋𝑖 → ℬ(𝑛, 𝑝)
𝑖=1
The probability that𝑆𝑛 = 𝑘 (obtaining k successes during n independent tests) is:
ℙ(𝑆𝑛 = 𝑘) = 𝐶𝑘𝑛 𝑝𝑘 𝑞 𝑛−𝑘 (24)
The expectation of a binomial variable 𝑆𝑛 is :
𝑛
𝐸(𝑆𝑛 ) = 𝐸(𝑋1 + 𝑋1 + ⋯ + 𝑋𝑖 + ⋯ + 𝑋𝑛 ) = ∑ 𝐸(𝑋𝑖 )
𝑖=1
3
With :
𝐸(𝑋𝑖 ) = 𝑝 (Bernoulli variable)
SO :
𝑛
𝐸(𝑆𝑛 ) = ∑ 𝑝
𝑖=1
𝐸(𝑆𝑛 ) = 𝑛𝑝
The variance of a binomial variable 𝑆𝑛 is :
𝑛
𝑉𝑎𝑟(𝑆𝑛 ) = 𝑉𝑎𝑟(𝑋1 + 𝑋1 + ⋯ + 𝑋𝑖 + ⋯ + 𝑋𝑛 ) = ∑ 𝑉𝑎𝑟(𝑋𝑖 )
𝑖=1
With :
𝑉𝑎𝑟(𝑋𝑖 ) = 𝑝𝑞 (Bernoulli variable)
So :
𝑛
𝑉𝑎𝑟(𝑆𝑛 ) = ∑ 𝑝𝑞
𝑖=1
𝑉𝑎𝑟(𝑆𝑛 ) = 𝑛𝑝𝑞
Example (binomial law)
A factory manufactures transistors each transistor has a 3% probability of being
defective. What is the law of the number of defective transistors in a batch of 100
transistors? What is his expectation? And its variance?
Solution
The law of the number of defective transistors in a batch of 100 transistors is a binomial law ℬ(𝑛, 𝑝). (It
is the sum of n v. a independent of Bernouill's law ℬ(1, 𝑝)).
With 𝑛 = 100 , 𝑝 = 0.03 𝑎𝑛𝑑 𝑞 = 1 − 𝑝 = 1 − 0.03 = 0.97
His expectation is:
𝐸(𝑋) = 𝑛𝑝
𝐸(𝑋) = 100.0.03 = 3
Var(X) = 𝑛𝑝𝑞
Var(X) = 100. (0.03)(0.97)
Var(X) = 2.91
Page 73
4
B.5.4 Poisson’s law
Poisson's law often applies to accidental phenomena where the probability is very weak (𝑝 <
0.05) .it is defined as the limit of a binomial law. Thus events which occur randomly such as
machine breakdowns, plane accidents, mistakes in a text, etc. can be considered as a Poissonian
process.
if : 𝑛 → ∞ 𝑒𝑡 𝑝 → 0
So 𝑋: ℬ(𝑛, 𝑝) → 𝑝(𝜆) 𝑎𝑣𝑒𝑐 𝑛𝑝 → 𝜆
This approximation is correct if 𝑛 ≥ 50 𝑒𝑡 𝑛𝑝 ≤ 5.
A random variable X has values in ℝ follows a Poisson law with parameter λ (λ>0) if the real
numbers 𝑝𝑘 are given by:
𝜆𝑘 𝑒 −𝜆
ℙ(𝑋 = 𝑘) = (25)
𝑘!
We notice :
𝑋~𝑃(𝜆)
If X and Y are independent random variables following Poisson laws respectively
𝑋~𝑃(𝜆)
And
𝑌~𝑃(𝜇)
So : 𝑋 + 𝑌~𝑃(𝜆 + 𝜇)
Remarks: a Poisson distribution is given by its probability law:
1) ∀ 𝑘, ℙ(𝑋 = 𝑘) > 0
𝑒 −𝜆 𝜆𝑘 𝜆𝑘
2) ∑𝑘≥0 ℙ(𝑋 = 𝑘) = ∑𝑘≥0 = 𝑒 −𝜆 ∑𝑘≥0 𝑘!
𝑘!
With :
𝜆𝑘
∑ = 𝑒𝜆
𝑘!
𝑘≥0
Where :
∑ ℙ(𝑋 = 𝑘) = 𝑒 −𝜆 𝑒 𝜆 = 1
𝑘≥0
To calculate the values of successive probabilities, it is possible to use a recurrence
formula:
5
𝜆
(𝑋 = 𝑘) = ℙ(𝑋 = 𝑘 − 1)
𝑘
The expectation of a Poisson variable is:
𝐸(𝑋) = ∑ 𝑘ℙ(𝑋 = 𝑘)
𝑘≥0
𝜆𝑘 𝑒 −𝜆
𝐸(𝑋) = ∑ 𝑘
𝑘!
𝑘≥0
With : 𝑘! = 𝑘(𝑘 − 1)!
So :
𝜆𝑘 𝑒 −𝜆
𝐸(𝑋) = ∑ 𝑘
𝑘(𝑘 − 1)!
𝑘≥0
𝜆𝑘 𝑒 −𝜆
𝐸(𝑋) = ∑
(𝑘 − 1)!
𝑘≥0
And : 𝜆𝑘 = 𝜆𝜆𝑘−1
𝜆𝜆𝑘−1 𝑒 −𝜆
𝐸(𝑋) = ∑
(𝑘 − 1)!
𝑘≥0
𝜆𝑘−1
𝐸(𝑋) = 𝜆𝑒 −𝜆 ∑
(𝑘 − 1)!
𝑘≥0
𝜆𝑘−1
With : ∑𝑘≥0 (𝑘−1)! = 𝑒 𝜆
𝐸(𝑋) = 𝜆𝑒 −𝜆 𝑒 𝜆
𝐸(𝑋) = 𝜆
The variance of a Poisson variable is:
𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − [𝐸(𝑋)]2
𝑉𝑎𝑟(𝑋) = ∑ 𝑘 2 𝑃(𝑋 = 𝑘) − [𝜆]2
𝑘≥0
𝜆𝑘 𝑒 −𝜆 2
𝑉𝑎𝑟(𝑋) = ∑ 𝑘 − [𝜆]2
𝑘!
𝑘≥0
With: 𝑘 2 = 𝑘(𝑘 − 1) + 𝑘
6
𝜆𝑘 𝑒 −𝜆
𝑉𝑎𝑟(𝑋) = ∑[𝑘(𝑘 − 1) + 𝑘] − [𝜆]2
𝑘!
𝑘≥0
𝜆𝑘
𝑉𝑎𝑟(𝑋) = 𝑒 −𝜆 ∑[𝑘(𝑘 − 1) + 𝑘] − [𝜆]2
𝑘!
𝑘≥0
𝜆𝑘 𝜆𝑘
𝑉𝑎𝑟(𝑋) = 𝑒 −𝜆 (∑ 𝑘(𝑘 − 1) + ∑ 𝑘 ) − [𝜆]2
𝑘! 𝑘!
𝑘≥0 𝑘≥0
With :
𝑘 1
=
𝑘! (𝑘 − 1)!
So :
𝜆𝑘 𝜆𝑘
𝑉𝑎𝑟(𝑋) = 𝑒 −𝜆 (∑(𝑘 − 1) +∑ ) − [𝜆]2
(𝑘 − 1)! (𝑘 − 1)!
𝑘≥0 𝑘≥0
And :
𝑘−1 1
=
(𝑘 − 1)! (𝑘 − 2)!
𝜆𝑘 𝜆𝑘
𝑉𝑎𝑟(𝑋) = 𝑒 −𝜆 (∑ +∑ ) − [𝜆]2
(𝑘 − 2)! (𝑘 − 1)!
𝑘≥0 𝑘≥0
With 𝜆𝑘 = 𝜆2 𝜆𝑘−2 et 𝜆𝑘 = 𝜆𝜆𝑘−1
−𝜆
𝜆2 𝜆𝑘−2 𝜆𝜆𝑘−1
𝑉𝑎𝑟(𝑋) = 𝑒 (∑ +∑ ) − [𝜆]2
(𝑘 − 2)! (𝑘 − 1)!
𝑘≥0 𝑘≥0
𝑉𝑎𝑟(𝑋) = 𝑒 −𝜆 (∑ 𝜆2 𝑒 𝜆 + ∑ 𝜆𝑒 𝜆 ) − [𝜆]2
𝑘≥0 𝑘≥0
𝑉𝑎𝑟(𝑋) = 𝜆2 𝑒 −𝜆 𝑒 𝜆 + 𝜆𝑒 −𝜆 𝑒 𝜆 − [𝜆]2
𝑉𝑎𝑟(𝑋) = 𝜆2 + 𝜆 − 𝜆2
𝑉𝑎𝑟(𝑋) = 𝜆
Example (Poisson's law)
An electronic component produces on average 1 error per 100000 hours. What is the
probability of one or more errors if the part operates 20000 hours?
7
Solution
Let's set X the value which gives the number of errors over 20000 hours of operation
𝑋~𝑃(𝜆)
Where: λ is the average number of errors in 20000 hours ( 𝐸(𝑋) = 𝜆)
1 𝑒𝑟𝑟𝑒𝑢𝑟 → 100000ℎ𝑜𝑢𝑟𝑠
𝑥 𝑒𝑟𝑟𝑒𝑢𝑟 → 20000ℎ𝑜𝑢𝑟𝑠
20000 2 1
𝜆 = 100000 = 10 = 5=0.2
On cherche : ℙ(X≥ 1) =?
ℙ(X≥ 1) = 1 − ℙ(𝑋 < 1)
ℙ(X≥ 1) = 1 − ℙ(𝑋 = 0)
𝜆0 𝑒 −𝜆
ℙ(X≥ 1) = 1 − ( )
0!
(0.2)0 𝑒 −(0.2)
ℙ(X≥ 1) = 1 − ( ) = 1 − 𝑒 −0.2
1
ℙ(X≥ 1) = 1 − 0.818 = 0.18127