Scribd
Importer
7 vues3 pages

Corrélation et régression des variables

Transféré par

junromjiatsa
Copyright
© All Rights Reserved
Nous prenons très au sérieux les droits relatifs au contenu. Si vous pensez qu’il s’agit de votre contenu, signalez une atteinte au droit d’auteur ici.
Formats disponibles
Téléchargez aux formats ODT, PDF, TXT ou lisez en ligne sur Scribd
7 vues3 pages

Corrélation et régression des variables

Transféré par

junromjiatsa
Copyright
© All Rights Reserved
Nous prenons très au sérieux les droits relatifs au contenu. Si vous pensez qu’il s’agit de votre contenu, signalez une atteinte au droit d’auteur ici.
Formats disponibles
Téléchargez aux formats ODT, PDF, TXT ou lisez en ligne sur Scribd

Exercice 4 :

1) le nombre d’individu qui ont une anciennete au travail de moin d’un an est de :
n_individu= 10+8+5+8+9+4= 44
2) determinons les deux distribution marginales

anciennete(X) [0,6[ [6,12[ [12,18[ [18,24[


centre 3 9 15 21
effectif 23 21 35 11

age(Y) [20,25[ [25,30[ [30,35[


centre 22.5 27.5 32.5
effectif 36 34 20

3) Déterminons la distribution de X conditionnelle à Y = [25, 30], c'est-à-dire, X/Y = [25, 30].

ancienneté(X) [0,6[ [6,12[ [12,18[ [18,24[


effectif 8 9 11 6

4) Non elles ne sont pas indépendantes car l’effectif de chaque ancienneté au chômage dépend de
la tranche d’age prise de chaque individu pris

n
1
5) Calculons les moyennes arithmétiques X̄=
n
∑ x ×c
i i
i=1
23×3+21×9+35×15+ 11×21
x̄= =11.27
90
22.5×36+27.5×34+ 32.5×20
ȳ= =26.62
90
6) Calculons le coefficient de corrélation linéaire ;
n
1
• Calculons les écarts-type des variables X et Y ;σ = ∑ (c i− x̄)2×x i
2
n i=1
1
σ X 2= ×((3−11.27)2×23+(9−11.27)2 ×21+(15−11.27)2×35+(21−11.27)2×11)=35.7
90
donc σ x = √ σ x =√ 35.7=5.98
2

1
σ Y 2= ×((22.5−26.62)2×36+(27.5−26.62)2×34 +(32.5−26.62)2×20)=14.8
90
doncσ y =√ σ y =√ 14.8=3.84
2

n
1
• calculons la covariance entre X et Y;cov ( X ,Y )= ∑ (x × y − x̄× ȳ)×nij
n i=1 i i
Centre X/Centre Y 22.5 27.5 32.5
3 (22.5*3-11.27*26.62)*10 (27.5*3-11.27*26.62)*8 (32.5*3-11.27*26.62)*5
9 (22.5*9-11.27*26.62)*8 (27.5*9-11.27*26.62)*9 (32.5*9-11.27*26.62)*4
15 (22.5*15- (27.5*15- (32.5*15-11.27*26.62)*9
11.27*26.62)*15 11.27*26.62)*11
21 (22.5*21-11.27*26.62)*3 (27.5*21-11.27*26.62)*6 (32.5*21-11.27*26.62)*2
On obtient
-2325,074 -1740,0592 -1012,537
-780,0592 -472,5666 -30,02
562,389 1237,4186 1687,4334
517,4778 1664,9556 764,9852
Donc Cov(X,Y)= 74,34/90=0.826
• Calculons le coefficient de corrélation :

cov (X , Y ) 0.826
r xy =σ × σ = =0.036
x y 3.84×5.98

7) Donnons l’equation de la droite de regression


cov ( X ,Y ) 0.036
Y=aX+b on : a= = =10−3
Var (X ) 35.7
−3
b= ȳ −a x̄=26.62−10 ×11 ,27=26 ,6
-3
donc (D) :Y=10 X+26,6
8 ) Determinons l’age d’une personne d’anciennete 15 mois
Y=10^-3×15+26 ,6=26.61~27 ans
Exercice 5 :
n
1
1) Calculons les moyennes arithmétiques des variables X et Y : X̄=
n
∑ x
i
i=1
80+ 100+115+110 +70+125+105+ 90+110+ 95
x̄= =100
10
32+50+ 62+ 56+8+ 80+62+50+62+38
ȳ= =50
10
n
1
2) Calculons les écarts-type des variables X et Y ;σ 2= ∑ (c i− x̄)2
n i=1
2 1
σ X = ×((80−100) +(100−100) +(115−100) +(110−100)2+(70−100)2 +(125−100)2)
2 2 2
10
2 2 2 2
+(105−100) +( 90−100) +(110−100) +( 95−100) =250
donc σ x = √ σ x =√ 250=15.82
2

1
σ Y 2= 2 2 2 2 2 2
×((32−50) +(50−50) +(62−50) +(56−50) +(8−50) +(80−50) )
10
2 2 2 2
+(62−50) +(50−50) +(62−50) +(38−50) =360
donc σ y =√ σ y =√ 360=18.98
2

n
1
3) Calculons la covariance entre X et Y ;cov ( X ,Y )= ∑ (x i× y i − x̄× ȳ)
n i=1
1
AN: cov ( X ,Y )= ×((2560−500)+(500−500)+(7130−500)+(6160−500)+(560−500))
10
(10000−500)+(6510−5000)+(4500−5000)+(6820−5000)+(3610−5000)=285
4) Calculons le coefficient de corrélation
cov (X , Y ) 285
r xy =σ × σ = =0.94
x y 18.98×15.82
cov(X,Y) est supérieure à 0 donc la corrélation est forte , il existe donc une dépendance
entre x et y
5) Déterminons la droite de régression de Y en X
cov ( X ,Y )
Y=aX+b on : a= =1.14
Var (X )
b= ȳ −a x̄=−64
donc on aura : Y=1.14X-64

Common questions

Alimenté par l’IA

The conditional distribution of a variable is created by analyzing the distribution of values for that variable given a certain condition related to another variable. For example, to determine the distribution of seniority (X) conditional on age being within [25, 30], we examine the distribution of seniority levels only for individuals in this age range. The document provides such a conditional distribution by listing effectives for seniority levels specifically for ages within [25, 30].

To evaluate if there is a dependence between two sets of data points, one can calculate the correlation coefficient. A high positive or negative correlation coefficient indicates strong dependence, while a value near zero suggests weak or no dependence. In the document, a correlation coefficient of 0.94 suggests strong dependence between the variables, showing a significant relationship .

The correlation coefficient r_xy is calculated using the formula r_xy = cov(X,Y) / (σ_x × σ_y), where cov(X,Y) is the covariance between variables X and Y, and σ_x and σ_y are the standard deviations of X and Y, respectively. The correlation coefficient indicates the strength and direction of the linear relationship between the two variables. A value close to 1 or -1 signifies a strong positive or negative linear relationship, respectively, while a value close to 0 indicates a weak linear relationship .

To determine the independence of two variables, such as seniority and age, we need to check if the effectives (frequencies) of seniority levels are consistent across different age groups. In this context, seniority (X) and age (Y) are considered independent if the distribution of X does not change with Y. However, the document states they are not independent because the number of individuals in each seniority level is dependent on the age group .

Using fixed intervals for classifying data points in statistical analysis can simplify the data and make it easier to identify patterns, as it reduces the variability and allows for a structured comparison. However, it can also lead to the loss of detailed information and potential bias if the intervals do not align well with natural breaks in the data. The choice of interval size can greatly impact the analysis and interpretation of data, as seen in the choice of seniority and age intervals in the document .

Standard deviations are crucial in calculating correlation as they provide a measure of the dispersion or variability in the data sets. When computing the correlation coefficient, the standard deviations of both variables are used to normalize the covariance, enabling a dimensionless measure of correlation that represents the strength of a linear relationship on a standardized scale. Without the standard deviations, the raw covariance does not provide intuitive insights into the relationship strength .

Covariance is a measure of how two variables change together. It is calculated as cov(X,Y) = 1/n Σ(x_i - x̄)(y_i - ȳ). A positive covariance indicates that the variables increase together, while a negative covariance indicates an inverse relationship. Covariance contributes to understanding the relationship between variables by indicating the direction of their linear relationship, but unlike correlation, it does not provide standardized measures, making it difficult to gauge the strength. In the provided example, the calculated covariance is used to compute the correlation coefficient, which standardizes this measure for easier interpretation .

The arithmetic mean provides a central value or average of a dataset, which serves as a representative measure of the data. It is fundamental in summarizing the data's central tendency, allowing for comparisons across different datasets or populations. The mean is also used as a foundational component in calculating other statistical measures such as variance, standard deviation, and the regression line, as demonstrated in the document .

The equation of the regression line Y = aX + b is determined using the slope a = cov(X,Y) / Var(X) and the intercept b = ȳ - a*x̄, where ȳ and x̄ are the means of Y and X, respectively. This equation represents the line of best fit for predicting the dependent variable Y based on the independent variable X. In the source, this process is used to derive the line Y = 1.14X - 64 for one dataset, indicating a slight increase in Y for every unit increase in X .

Outliers can have a significant impact on the correlation coefficient, as they can artificially inflate or deflate the perceived strength of the linear association between two variables. This is because the correlation coefficient is sensitive to extreme values, which can skew the results by either exaggerating or minimizing the apparent relationship. The document doesn't explicitly discuss outliers, but in general, removing or accounting for outliers can provide a more accurate representation of the correlation [General knowledge].

Vous aimerez peut-être aussi