Chapitre 7

Elements of Sampling Theory

Dr Khalil AMINE
[Link]@[Link]

EHTP 2024/2025
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 1 / 87
 Introduction

1. Introduction

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 2 / 87
 Introduction

Introduction

A sample is any representative subset of a given

population, referred to as the parent population.
The theory of sampling aims to study the
relationships between the parameters of a population
and those of the samples derived from it.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 3 / 87
 Introduction

Introduction

Inferential statistics allow us to:

evaluate the parameters of the samples of a
population whose parameters are known
extrapolate results obtained from samples to
characterise a parent population whose
parameters are unknown.

Two important techniques of inferential statistics are

thus referred to: sampling and estimation.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 4 / 87
 Introduction

Introduction

Sampling is the process of selecting individuals to

observe (study) from a parent population.

Avoid conducting observations on all individuals

in the population
=⇒ Reduce the cost of data collection in terms of
time and money

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 5 / 87
 Introduction

Introduction

Remark
In certain situations, a partial survey is the only
feasible solution, such as when exhaustive observation
is destructive or when the state of the population
(regarding the studied characteristic) changes
rapidly.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 6 / 87
 Introduction

Introduction

Definition: Population, Individual

A population is a set of elements subject to
statistical study, called individuals, statistical
units, or basic elements.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 7 / 87
 Introduction

Introduction

Definition: Sample
A sample is any representative subset of a parent
population.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 8 / 87
 Introduction

Introduction

Definition: Sampling
Sampling refers to the organised process of selecting
elements for the sample.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 9 / 87
 Introduction

Introduction

Definition: Census
A census refers to an exhaustive survey that collects
data on all individuals in the population.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 10 / 87
 Introduction

Introduction

Definition: Survey
A survey refers to a non-exhaustive investigation
where only a portion of the individuals in the
population are observed.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 11 / 87
 Introduction

Introduction

Definition: Sampling Frame

A sampling frame is the list that enumerates all the
individuals in a population to be studied.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 12 / 87
 Introduction

Introduction

Definition: Sampling Error

Sampling error refers to the difference between the
results obtained from a sample and what a
comparable census of the population would reveal.

Remark
The larger the sample size, the smaller the sampling
error.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 13 / 87
 Introduction

Introduction

Definition: Sampling Fraction

The sampling fraction, or sampling rate, is the
proportion of the population included in the sample.

It is the percentage ratio of the sample size n to the

population size N :
n
f =
N

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 14 / 87
 Sampling Methods

2. Sampling Methods

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 15 / 87
 Sampling Methods

Sampling Methods

Sampling methods refer to the different approaches

used to create a sample from the studied population.

These methods aim to produce a sample that is

representative of the population studied; otherwise, it
may not exhibit the same characteristics as the parent
population.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 16 / 87
 Sampling Methods

Sampling Methods

Two types of sampling methods:

Non-probabilistic / empirical sampling methods

Random / probabilistic sampling methods

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 17 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

Also known as empirical methods or

judgment-based sampling, these methods involve
selecting individuals without randomisation, based on
feasibility, resemblance to the target population, and
subjective criteria determined by the investigators.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 18 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

Convenience sampling: The selection of

individuals is based on their availability and the
cost of access. In this case, the sample is
obtained without any particular method. The
study is likely to be biased (subjective): a
convenience sample may not necessarily be
representative, as factors such as location and
time can introduce selection bias.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 19 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

Example: Each day, a quality controller in a factory

selects the first 100 items they see to check for
compliance.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 20 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

The advantages of convenience sampling are:

easy to implement;
minimal investigator influence.

The disadvantages of convenience sampling are:

biased: selection is governed by accessibility;
lack of representativeness: the individuals
selected may not represent the characteristics of
the population.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 21 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

Voluntary response sampling: This technique

relies on volunteers to form the sample.

Example: To evaluate customer satisfaction, a

restaurant places a short questionnaire on its tables.
Customers who wish to participate complete the
questionnaire.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 22 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

The advantages of voluntary response sampling are:

ethically appealing: only individuals interested in
participating are surveyed;
useful in exploratory phases.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 23 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

The disadvantages of voluntary response sampling are:

biased: selection is governed by willingness, not

the population’s characteristics;
lack of representativeness: respondents may not
reflect the population’s traits.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 24 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

Quota sampling: This involves constructing a

sample that serves as a scaled-down model of the
studied population.

This means replicating the characteristics of the

parent population as accurately as possible, using
statistical information about its structure.

This approach reduces the risk of obtaining an

anomalous sample (i.e., one exhibiting
exceptional characteristics of the population).
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 25 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

Quota sampling relies on the use of control

variables (such as gender, age, profession, etc.),
ensuring that their proportions in the parent
population are approximately reflected in the sample.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 26 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

Example: For a population comprising 48% men,

52% women, and 25% young adults aged 25 to 30, a
researcher tasked with surveying 100 individuals
would need to include 48 men, 52 women, and 25
people aged 25 to 30.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 27 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

The advantages of quota sampling are:

easy to use;
cost-effective: there is no obligation to survey
specific individuals;
does not require a sampling frame: only
knowledge of the population’s distribution is
needed.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 28 / 87
 Sampling Methods Non-Probabilistic Methods

Non-Probabilistic Methods

The disadvantages of quota sampling are:

biased: the selection of individuals cannot be
entirely objective;
lack of representativeness: control variables may
not comprehensively reflect the population;
unknown response rates: there is no explicit
guidance on sample size or sampling rates.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 29 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Random or probabilistic sampling methods are

approaches for constructing a sample based on rules
determined by chance.

The individuals in the population have a known,

non-zero probability of being selected as part of the
sample. Thus, only chance governs whether an
individual is included in the sample, without
researcher intervention.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 30 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Simple random sampling : All elements in the

population have the same probability of being
included in the sample.

A random number generator or methods such as

blind drawing can be used to select a given
number of individuals.

This method produces a sample that is

representative of the population because each
individual has an equal probability of being
included.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 31 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Example:
A teacher uses random number generation software to
select the roll numbers of students who will
participate in a tutorial session.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 32 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Stratified random sampling : This involves

subdividing the population into homogeneous
subgroups called strata, defined based on one or
more criteria (referred to as variables of interest).

From each stratum, individuals are randomly

selected to form simple random sub-samples.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 33 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

This method produces a representative sample

because all individuals within a group have the same
probability of being included in the sub-sample, and
the resulting sample is representative of the
population regarding the variable of interest.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 34 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Example:
A survey on training at EHTP was conducted among
90 first-year students. The sample was obtained by
randomly selecting 10 students from each of the 9
fields of study.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 35 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Cluster random sampling : This involves

subdividing the population into subgroups called
clusters, randomly selecting clusters, and
including their members to form the sample.

This method provides a representative sample of

the population if the clusters are similar to one
another and individuals within a cluster are
heterogeneous.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 36 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Example:
Selecting two first-year classes at EHTP to study the
integration of new students at the school.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 37 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Systematic random sampling : This method

involves listing all elements in the population
(the sampling frame) and randomly selecting
every n th element to construct the sample.

Systematic sampling is thus a variation of simple

random sampling, where the sample is selected at
predetermined intervals.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 38 / 87
 Sampling Methods Random Sampling Methods

Random Sampling Methods

Example:
Using a list of all students in a school, a guidance
counsellor randomly selects a starting name and then
selects every 20th name to form the sample.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 39 / 87
 Study of Sample Size

3. Study of Sample Size

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 40 / 87
 Study of Sample Size

Study of Sample Size

The study of sample size involves finding a

compromise between the precision of sample
representativeness and minimising the cost of the
survey.

In other words, the sample size must strike the best

balance between sampling errors and the cost of
sampling itself.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 41 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Bienaymé-Tchebycheff Inequality

The Bienaymé-Tchebycheff inequality is a result that

shows that a random variable is likely to take a value
relatively close to its expectation.

This inequality is used only if the distribution of the

random variable is completely unknown. It leads to
large sample sizes.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 42 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Proposition: Bienaymé-Tchebycheff Inequality

Given a random variable X with expectation
E[X ] = µ and finite variance var(X ) = σ 2 , for any
strictly positive real number ε, we have:

σ2
P (|X − µ| < ε) ≥ 1 −
ε2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 43 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

This inequality implies that for a variable whose

behaviour is unknown, but whose mean (expectation)
and dispersion (variance) are known (or can be
estimated), we can determine an interval, called the
fluctuation interval, such that there is a measurable
probability that the true value of the variable lies
within this interval:
σ2
µ−ε < X < µ+ε with a probability of at least 1− 2 .
ε

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 44 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

In cases where the population parameters are

unknown, this inequality can be used (within the
framework of estimation theory, rather than sampling)
to determine an interval, called the confidence
interval, for these parameters:

σ2
X −ε < µ < X +ε with a probability of at least 1− .
ε2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 45 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

In terminology:
ε is referred to as the margin of error.
σ2
1 − 2 , denoted as 1 − α, is the confidence
ε
level, where α is referred to as the risk level or
error probability.

In practice, confidence levels of 95% to 99% are used,

corresponding to error probabilities of 5% to 1%.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 46 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Remark
To easily recall the difference between a fluctuation
interval and a confidence interval:
If the theoretical proportion of a population is
known and we wish to decide on the actual
frequency (e.g., to assess whether a dice is fair),
then the fluctuation interval is used to evaluate
(by bounding) the observed frequency relative to
the theoretical proportion.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 47 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Remark
If the theoretical proportion of a population is
unknown, but an observed frequency is available,
the confidence interval is used to estimate (by
bounding) the theoretical proportion using the
observed frequency.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 48 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Example:
The probability of an event occurring is assumed to
be p. We consider samples of size n drawn, with
replacement, from a population of size N assumed to
be infinite or very large. For each extracted sample,
there is a random variable representing the frequency
Fn of the event’s occurrence such that:
E[Fn ] = p
p(1 − p)
var(Fn ) =
n
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 49 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

The confidence level in this case is expressed as:

σ2 p(1 − p)
1−α=1− 2 =1− .
ε nε2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 50 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Thus, a characterisation of the sample size is given by:

p(1 − p)
n= .
αε2
For example, if the theoretical proportion is 0.5, the
error risk α = 5%, and the margin of error is 3%, the
required sample size for the sampling to be
0.25
representative is n = 0.09×0.05 = 55.56, rounded to
n = 56.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 51 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Definition: Convergence in Probability
A sequence (Xn )n≥1 of random variables in R is said
to converge in probability to the random variable
X in R, if:

∀ε > 0 , lim P (|Xn − X | ≥ ε) = 0

n→+∞

P
This is denoted as: Xn −−−−−→ X
n→+∞

This is equivalent to:

∀ε > 0 , lim P (|Xn − X | < ε) = 1
n→+∞ . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 52 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Theorem: Weak Law of Large Numbers
Let (Xn )n≥1 be a sequence of independent and
identically distributed random variables in R with
1∑ n
expected value µ and variance σ 2 . Let X n = Xi .
n i=1
Then X n converges in probability to µ, meaning:
( )
∀ε > 0 , lim P |X n − µ| < ε = 1
n→+∞

Concretely, this theorem implies that as n grows

larger, the random variable X n becomes increasingly
close to the mathematical expectation µ. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 53 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Example: A dice is rolled. If a 6 is rolled, we win and

score 1 point; otherwise, we lose and score 0 points.

Let Xi be the random variable corresponding to the

points obtained on the i-th roll.
5 1
The probabilities are: P(X = 0) = , P(X = 1) = ,
6 6
1
and E[X ] = .
6

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 54 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Example: (continued)
The same experiment is repeated n times. The n
random variables X1 , X2 , . . . , Xn share the same
probability distribution.
To measure the number of successes, we examine the
random variable X n : Frequency of Successes
with
Number of Successes 1
Xn = = (X1 + X2 + · · · + Xn ) .
Number of Trials n

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 55 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Example: (continued)
1
Ideally, we expect X n = .
6
1
For n = 3, it is unlikely to find X 3 = .
6
1
For n = 30, the probability of finding X 30 =
6
increases but remains moderate.
For n = 1000, the value of X n approaches the
1
theoretical value of .
6
The theorem asserts that as n grows, X n gets closer to
1
the theoretical value .
6 .
.
.
.
.
. . . . .
. . . .
. . . .
. . . .
. . . .
. . . . .
.
.
.
.
.
.
.
.
.

Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 56 / 87

 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Theorem: Strong Law of Large Numbers

Let (Xn )n≥1 be a sequence of independent and
identically distributed random variables in R with
1∑ n
expected value µ and variance σ 2 . Let X n = Xi .
n i=1
Then X n converges almost surely to µ, meaning there
exists an event A of probability 1 such that:

∀ω ∈ A , lim X n (ω) = µ
n→+∞

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 57 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Interpretation:
Repeating the same experiment a large number of
times and observing the frequency of a result r, the
Strong Law of Large Numbers states that the
empirical frequency of r tends to the probability of r
as the number of trials approaches infinity.

For example, tossing a fair coin a large number of

times results in the relative frequency of heads
converging almost surely to 50%.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 58 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Theorem: Central Limit Theorem

Let X1 , X2 , . . . , Xn be n independent and identically
distributed random variables with expected value µ
2 1∑ n
and variance σ . Let X n = Xi . Then, for
n i=1
sufficiently large n, X n (approximately
) follows the
2
σ
normal distribution N µ, .
n

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 59 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Remark
In most cases, n is considered sufficiently large when
n ≥ 30. However, this depends on the nature of the
population and the study context.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 60 / 87
 Study of Sample Size Bienaymé-Tchebycheff Inequality

Study of Sample Size

Remark
The Central Limit Theorem relies on the concept of
convergence in distribution. Specifically,
( ) stating that
2
σ
X n approximately follows N µ, is equivalent to
n
saying that the random variable Xn converges in
distribution to a random variable that follows the
normal distribution.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 61 / 87
 Sampling Distributions

4. Sampling Distributions

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 62 / 87
 Sampling Distributions

Sampling Distributions

Let consider the case of simple random sampling.

Given a population of size N and a random variable X

with expectation E[X ] = µ and variance var(X ) = σ 2 ,
let consider n independent random variables
X1 , X2 , . . . , Xn that follow the same distribution as X ,
such that the i th observation xi can be considered as
an observation of the random variable Xi .

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 63 / 87
 Sampling Distributions

Sampling Distributions

The random variables (X1 , X2 , . . . , Xn ) are said to

form a sample ej , and any function defined on the
sample is referred to as a statistic.
The random variable mean (expectation) of the
1∑ n
sample ej : X nj = Xi
n i=1
The random variable variance of the sample ej :
1∑ n ( )2
Σ2nj = Xi − X n
n i=1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 64 / 87
 Sampling Distributions

Sampling Distributions

√ standard
The random variable
n (
deviation of the
1∑ )2
sample ej : Σnj = Xi − X n
n i=1
The random variable frequency of the sample ej :
1∑ n
Fnj = fi
n i=1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 65 / 87
 Sampling Distributions

Sampling Distributions

If, from a population, we take all possible samples of

size n: e1 , e2 , . . . , we can define new random variables
that represent statistical series of the random variables
(mean, variance, standard deviation, frequency, . . . )
for each sample ei : X ni , Σ2ni , Σni and Fni .

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 66 / 87
 Sampling Distributions

Sampling Distributions
We are dealing then with a sampling distribution
concerning the characteristic under study. Thus, we
define :
sampling
( distribution ) of the mean :
X n = X n1 , X n2 , . . .
sampling distribution of the variance :
Σ2n = (Σ2n1 , Σ2n2 , . . . )
sampling distribution of the standard deviation :
Σn = (Σn1 , Σn2 , . . . )
sampling distribution of the frequency :
Fn = (Fn1 , Fn2 , . . . )
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 67 / 87
 Sampling Distributions

Sampling Distributions

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 68 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

Proposition: Distribution of Means

Given a population and a random variable X with
expectation E[X ] = µ and variance var(X ) = σ 2 , for
any sampling distribution of means X n of sample size
n, we have :
1. If we consider an infinite population or a finite
population with replacement, then :
E[X n ] = µ
σ2
var(X n ) =
n

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 69 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

Proposition: Distribution of Means

2. If we consider a finite population without
replacement, then :
E[X n ] = µ √
σ N −n
σ(X n ) = √
n N −1
N −n
The value is called the exhaustiveness
N −1
coefficient or correction coefficient.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 70 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

Remark
The first case arises from the fact that
N −n
lim =1
N →+∞ N − 1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 71 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

Proposition: Convergence in Probability

If we are dealing with one of the following two cases :
1. The random variable X follows any probability
distribution on the population with E[X ] = µ and
var(X ) = σ 2 , and we take samples of size n such
that n ≥ 30.
2 The random variable X follows a normal
distribution with parameters E[X ] = µ and
var(X ) = σ 2 , i.e., X ∼ N (µ, σ 2 ) on the
population, and we take samples of any size n.
( ) ( )
Then : X n ∼ N (E X n , var X n ) .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 72 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

In other words : ) (
σ2
Infinite case : X n ∼ N µ,
n
( )
σ2 N − n
Finite case : X n ∼ N µ, ×
n N −1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 73 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

Example:
A machine produces parts in large quantities. To each
part randomly drawn, we assign its length in
millimetres; thus, we define a random variable X .

We assume that X follows the normal distribution

N (28.20, 0.0272 ).

Let X n be the random variable that associates the

average length of the n parts in a random sample of
size n.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 74 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

Example (continued):
The property then tells us that for sufficiently
( ) large n,
2
σ
X n follows the normal distribution N µ, or
( ) n
0.0272
N 28.20, .
n
Suppose the sample size is 100, then X 100 follows the
distribution N (28.20, 0.00272 ).

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 75 / 87
 Sampling Distributions Sampling Distribution of Means

Sampling Distribution of Means

Exercise:
A machine produces parts whose length X follows a
normal distribution with mean 60 mm and standard
deviation 4 mm. For quality control purposes, we take
samples of 25 parts.

Let X 25 be the random variable that associates the

mean length of the sample of 25 parts.
1. What is the probability distribution that the
random variable X 25 can follow?
2. Calculate the probability that the mean length in
the sample is greater than or equal to 61 mm. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 76 / 87
 Sampling Distributions Sampling Distribution of Variances

Sampling Distribution of Variances

Proposition: Variance Distribution

Given a population with a random variable X having
E[X ] = µ and var(X ) = σ 2 , the sampling distribution
of variances Σ2n from samples of size n satisfies:
n−1 2
E[Σ2n ] = σ
n
n − 1( )
var(Σ2n ) = (n − 1)µ 4
− (n − 3)σ 4
n3

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 77 / 87
 Sampling Distributions Sampling Distribution of Variances

Sampling Distribution of Variances

Remark
var(Σ2n ) −→ 0 as n → +∞.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 78 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

Proposition
Given a population with a random variable X having
E[X ] = µ and var(X ) = σ 2 , the sampling distribution
of frequencies Fn from samples of size n satisfies:
1. For an infinite population or a finite population
with replacement:
E[Fn ] = p
p(1 − p)
var(Fn ) =
n

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 79 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

Proposition
2. For a finite population without replacement:
E[Fn ] = p
p(1 − p) N − n
var(Fn ) = ×
n N −1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 80 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

Proposition: Convergence in Probability

If one of the following conditions holds:
1. The random variable X follows any probability
distribution on the population with E[X ] = µ,
var(X ) = σ 2 , and frequency p, and samples are of
size n ≥ 30.
2. The random variable X follows a normal
distribution N (µ, σ 2 ) on the population, and
samples are of any size n.
Then: Fn ∼ N (E[Fn ], var(Fn ))
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 81 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

In other words: ( )
p(1 − p)
Infinite population: Fn ∼ N p,
n
Finite population:
( )
p(1 − p) N − n
Fn ∼ N p, ×
n N −1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 82 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

Example:
An urn contains 100 balls numbered from 1 to 100,
indistinguishable by touch. During a random draw,
the probability of obtaining a number less than or
equal to 37 is p = 0.37. A success is defined as
drawing one of the balls numbered from 1 to 37.

A sample of size 50 is obtained through random

sampling with replacement of 50 balls. We are
interested in the frequency of successes in this sample
of 50 balls.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 83 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

Example (continued):
Let F50 be the random variable that assigns to each
sample of size 50 its success frequency.

Xi is the random variable that assigns 1 if the i th ball

results in a success, and 0 otherwise.

The Xi are independent random variables following

the same Bernoulli distribution with parameter
p = 0.37, with E(Xi ) = 0.37 and
var(Xi ) = p(1 − p) = 0.23.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 84 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

Example (continued):
X1 + X2 + · · · + X50
We have F50 = , with mean
50
0.37 × 0.63
p = 0.37 and variance = 0.004.
50
Note that here, the frequencies Fn are measured using
random variables Xi following the Bernoulli
distribution.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 85 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distribution of Frequencies

Exercise:
A machine produces parts. On average, it produces
3% defective parts. A customer receives a box of 500
parts directly from the machine.
1. What is the probability that fewer than 1% of the
parts in the box are defective?
2. What is the probability that more than 5% of the
parts in the box are defective?

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
Dr Khalil Amine (EHTP, Casablanca) [Link]@[Link] 86 / 87
 Sampling Distributions Sampling Distribution of Frequencies

Sampling Distributions
Exercise:
Consider a population consisting of a large number of individuals
who may possess a characteristic A. To study this population,
samples of n = 2500 individuals are drawn.
Let X denote the number of individuals with characteristic A in
a sample. The random variable X follows a distribution with
mean 5 and variance 1502 .
Let X n denote the random variable that assigns the mean of X
in the sample to each sample.
1. What probability distribution can the random variable X n
follow?
2. Calculate the probability that the mean number of
individuals X n in the sample is less than 8.
3. Calculate the probability that the mean number of . . . . . . . . . . . . . . . . . . . .

individuals X n in [Link]@[Link]
Dr Khalil Amine (EHTP, Casablanca)
the sample is greater than 2.
. . . . . . . . . . . . . . . . . .
87 / 87
. .