Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
009611 355046
) .(Uncertainty - Incertitude
.
.
) Dependent variable variable
(dpendante ).(Independent variable Variable indpendante
) (Ct
) (Yt :
C t = 0 + 1 Yt U t
) (Inexact relationship Relation exacte Ut
.
1000 .
) (Subpopulations Sous populations
. ) (
) (Fixed variable - variable fixe .
170 $10000 .
$12000
.
.
.
. . . .
) (
.
) (On the average - n moyennee
. ) Expected value Valeur
(espre
13
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
E ( C / Y ) = cy
E (C / Y ) = cy = + Y
cy ) (
.
)(Clustered - Rgroupes
.
:
C = + Y + U1
:
C = + Y + U 2
:
C = + Y + U k
Uk U2 U1
.
U1
U2 .
.
U U1
U2 U3 .Uk.....
) (Arithmetic Mean - Moyenne arithmtique .
U
(Assumptions underlying the error term Hypothses concernant le terme
(derreur
:
. )(
)( ) Random variable
(Variable alatoire .
14
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
Z
Z
) (10
) (11 ) (12 . . . . ):(1
)(1
Z
Y
X
)E(Y/X
32 34 30 32 31 33 10
34 36 33 34 36 31 11
36 42 38 31 34 35 12
38 45 33 37 36 39 13
40 40 39 51 36 34 14
. Z ) (10
Z :
34 30 32 31 33
Z .32
Z ) (10
32
) (11 34 .
) (1 :
15
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
16
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
VIEW / SHOW
17
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
18
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
)
Y Y ) E (Y / X
) (Deviation -Deviation The observed ) Y
(value La valeur observe ) E (Y / X
:
:
.
) . (observation - observation ) xy = E (Y / X
.
: (Independent random
(variable Variable alatoire indpendante
19
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
Y = + X + U
U = Y X
U = (Y X )
U = Y X
U = Y n X
U = Y n(Y X ) X
U = Y n Y + nX X
U = Y Y + X X
U = 0
E (U ) = 0
( Variance - Variance) :
(Constant variance Variance
. y ( constante
:
Y = E (Y / X ) + U
y2 = E [Y yx ]2
y2 = E [Y E (Y / X )]
2
y2 = E (U )2
y2 = u2 = yx
2
Aigner, p:25 1
. . = Y X
20
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
:
) (Normally distributed distribu normalement ) .E(Y/X
(F and t-tests
) .
.
2
U ~ N (0, )
U
.
: ) (covariance covariance
U X : 2
) XU = X (Y X
XU = XY X X 2
XU = XY X (Y X ) X 2
XU = XY Y X + X X X 2
XU = XY [ ) ( Y )( X
n
] [ X 2
) ( X ) ( X
n
]
xy
XU = xy x2
. x2
)
Y = b0 + b1 x + e Y e
e U . Y b0 b1
Cov( X , Y )= n
xy 2
. .
Aigner .25: x = X X , y = Y Y :
X Y
(X X )(Y Y ) = xy = XY n
:
) ( X 2
(X X ) = x 2 = X 2 n
2
= xxy2
:
,
)
Y e = (b0 + b1 X )e = b0 e + b1 X e = 0
21
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
XU = 0
: U t t
U t 1 . U t + 1
: .
: ) (Multicollinearity - Multicolinarit
. .
: (There is no specification .
.(error - Il ny a pas derreur de spcification
: ) Aggregation -
(Agrgation .
)(
) (Estimation - estimation ) E (Y / X
. :
Y = + X + U
(The
b0 and b1 (regression coefficient ) (estimates - estimations
) , (parameters - paramtres .
22
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
) X X1
Y1
Y1 (The regression line La
(droite de regression e1 ) .(Residual - Rsidu
X e n .....e 2 , e1
. (Y Y ) = e = 0
)
.
:
( )
)2
Y Y = e 2 = e12 + e 22 .....
. b0 , b1
)
Y = b0 + b1 X )The sum
(squares of residuals ) (Minimum - Minimum
. ) The normal equations Les
(quations normales ) Partial derivative Drive
(partielle b0 b1
:
Y = b0 + b1 X + e
e = Y b0 b1 X
) e 2 = (Y b0 b1 X
2
e2
b0 = 2 (Y b0 b1 X ) ( 1) = 0
Y = b0 n + b1 X )(1
23
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
e2
b1 = 2 (Y b0 b1 X ) ( X ) = 0
XY = b0 X + b1 X 2 )(2
) (1 ) (2 .
:
Y = b0 + b1 X
Y = b0 n + b1 X
X :
XY = b0 X + b1 X 2
b0 ) (1 :n
Y
= + b1 X
b0 n
n n n
Y = b 0 + b1 X
b0 = Y b1 X )(3
b1 ) (1) (2
) (Cramer's rule Rgle de Cramer :
n Y
X XY n XY ( X ) )( Y
= b1 n X
= n X 2
( X )2
X X2
= b1 XY
X2 )(4
XY X
.Y X 2 X .
24
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
b0 b1 :
.
.
.
)
Y = b 0 + b1 X
b0
Y ) . (The origin Lorigine b1 (The
(slope La pente
. b0 b1
. b1
(Simple regression coefficient Coefficient de la rgression
) simple
bi (Partial regression coefficient
(coefficient de la rgression partielle Y
X (keeping the
effect of other independent variables constant En gardant leffet des autres
.(variables indpendantes constant bi
) (Multiple regression Rgression multiple
) (
.
) (Bushels
) In deviation form
b0 . y = b1 x + e
(Dviation de la moyenne
b1 b1:
= 2 ( y bx ) ( x ) = 0
2
e
b1
xy = b1 x 2
xy
= b1 x 2
b1 , b2
X X
Z x = SX
ZY = Z X
:
Z X ZY Z X ZY
= Z X2
= n 1 =r
r
:
X Y
b= =r
SY SY
SX SX
, .
Z X = Z Y = 0 0 = Z Y 1 Z X = 0
Y e = 0 Y
. Y = Y SY 2: Y = (Y + e) = Y
= SY 2
1
n
( 2 1
n
)
) Y Y = (b0 + b1 X Y
2
2
]) = (Y b1 X + b1 X Y ) = [b1 ( X X
2 1 2 1
SY
n n
) (X X
2
2
SY = b 1
2
= b12 S X2
n
: S y2
2
S S2
r = Y2 = b12 X2
2
Sy Sy
Aigner p:29
26
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
= r = b SS X
Y
Y X
( . )
) .( SY = 4 . S X = 3
= 0.60
0.60 Y
0.60 4 = 2.40 3
.
)Standard
Partial regression coefficients Coefficients standardiss de la rgression
(partielle Y X
.
) Matrices -
(Matrices:
Y = Xb + e
) b = ( X X ) ( X Y
1
( X X )1 . 3 X
:
) (Best - Meilleur ) (Linear - Linaire ) Unbiased
b ( Non biaise , .
) (Linear combination Combinaison linaire
Y
=b XY
X2
Drapper & Smith ,. "Applied regression Analysis" john Wiley & Sons, inc. 3
1966. pp:44-52.
.111-109
b0 b1 . b0 b1
27
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
X
X2 X
b ) Weighted ) (Weight- Pondration
(sum Somme pondre . 4
xy
=b x2
b = y
) b = (Y Y
b = Y Y
b = Y Y xx2
x = 0
b = Y
Y ) Statistically independent statistiquement
Y2 (indpendante Y1
) (Repeated random sample chantillon alatoire rpt Y
Y
: X
X (X X ). X
= X = .X
X2 X 2
X 2
XX
= X X 2
=1
: b
) xy x( x + e
=b =
x 2 x 2
x 2 xe
=b + 2
x 2 x
E (b ) =
28
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
b
b
E (b )
E (b ) =
b b
b = b
b .
E (b ) = E (b ) =
b b
) (Variance - Variance
) Smaller confidence interval le plus petit intervalle de
(confiance
) (Statistically significant Statistiquement significatif
b0 + b1 X
b0 + b1 X
+ X c + dx
.
) Consistent estimator Estimateur
.(consistant b Asymptotic unbiased)
(estimator Estimateur asymptotiquement non biais
:
b
)(2
)(1
)(3
) (1 n
n ) (2 n
) (3 b .
):(2
30
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
)(2
) (
6
) ( 1978
)
Y = 718.07 + 0.137 X
r 2 = 0.86
" 6 " :
1981 190 208
31
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
32
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
86%
)(
.
) (cross-section data - donnes en coupes instantanes
) (time series data donnes chronologiques
= 0.137
0.137
X Y
X Y
137 .
= 718 .07
X = 0
b0 .7
) (2 b1
.
7 .
) (Time series data Y X
) (.
33
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
b
) (b0 + b1 X ) ( + X . 8
. :
b b
b b2 b2
. 9
) Coefficient of determination
)Coefficient de dtermination
(
( ) X ) Y
)( ) (variance - variance
) (variation - variation
. 10
8 ) (Regression ) .(Galton
) (Inheritance
.
. )(Regression toward mediocrity
.:
Fred N Kerlinger and Elazar j. pedhazur., Multiple Regression in Behavioral Research. Holt, Rinhert and
Winston, lnc., 1973,p:18
9 67:
10 ) (degree of freedom
:Y
) (Y Y
2
= y2
n 1
34
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
Y X
X . Y
)The
(mean La moyenne Y Y
. r = 0.0 r = 1.0
. r = 0.0
r = 1.0
Y X . 11
) (Residuals . SS Re sd :
Y = b0 + bX + e
)
Y = b0 + bX
)
Y = Y bX + bX
) Y = Y + b( X X
)
)(1
) (1 Y :
) Y Y = Y Y b( X X
)
)(2
) (2 :
( )
]) Y Y = [(Y Y ) b( X X
)2 2
( )2
)
) Y Y = (Y Y ) + b 2 ( X X ) 2b (Y Y )( X X
2 2 )(3
) (3 :
) 2b (Y Y )( X X ) = 2b. b ( X X
2
:
) b = (Y Y )( X 2 X
XX( )
) ( 3 :
( )2
)
) Y Y = (Y Y ) +b2 ( X X) 2b2 ( X X
2 2 2
( )2
)
) Y Y = (Y Y ) b 2 ( X X
2 2
)(4
) (1:
b 2 (X X ) = Y Y
2 )
( ) 2
) (4 :
( )
Y Y = (Y Y ) Y Y
)2 2 )
( )
2
)(5
) (Y Y
2 )
(
= Y Y )2
( )2
+ Y Y )
= +
) (5 :
36
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
(Y Y )2
)
(
Y Y )2 (
Y Y
)
)2
(Y Y = )2 (Y Y )2
+ (Y Y )2
R 2 :
R = (Y Y ) 2 = 1 ((Y Y ))2
( )2
) )2
2 Y Y Y Y
)(6
(Maximum )
) (6 )value Valeur maximale
R 2 = 1.0 Y = Y
)( ))(Minimum value Valeur minimale
R 2 = 0.0 Y = Y Y
X . Y
R 2 = m r ) (Sign -Signe
) (Direction - Direction
.
)Correlation -
(Correlation ) (Relationship - Relation
) (Covariance - Covariance
X . Y
) (Deviation scores - Deviations de la moyenne b0 = 0.0
:
)
y = bx
:
( )
Y Y )2
R 2
( = 1
) Y
2
Y
:
37
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
) ( y bx
2
R2 = 1 y2
y 2 ( y bx )2
= R2 y2
y 2 y 2 + 2 b xy b 2 x 2
= R2 y2
xy
=b x2
xy = b x 2
b xy = b 2 x 2
12
:
xy
R =b 2 2 x2
y2
= b. x2
. x2
y2
= b xxy2
b :
xy
= R2 x 2 . xy
y2
R 2 = b yx .bxy 13
xy
=r = b yx .b xy
x2 . y2
Z X ZY
r = n 1
).(Standard score Scores standards
xy
R = b
2
12
y 2
b1x1 y + b2 x 2 y
= R 116-115 .
2
y 2
13 byx Y X ) Y = F ( X bxy
X Y . X = b0 + b1Y
38
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
. ) (Causation - Causalit
) (Cause - Cause ) (Effect - Effet
) (Spurious correlation Correlation errone
) Statistical control Contrle
(statistique
r = 0.98
)(
. 14
. )The
(central limit theorem Le thorme de la limite centrale
) (The sampling distribution - Echantillonnage
).(Normally distributed Normalement distribue
) (n ) Observations -
(Observations x S X
x
x .
)Repeated random samples size n Echantillons alatoires ) (n
(rpts de taille n
)
(
X = n
%95 m 1.96 X
X X m 1.96 X
1.96 ) (critical value Valeur critique
) (table value Valeur tabule ) Areas
.(under the standard normal curve Surface sous la courbe normale
X m 1.96 X 95%
%5 .
X m 2.58 X 99%
/ ) the level of significance Le
.15 (niveau de signification
14 123-121:
15 ) (Type l error Erreur de type 1 .
1% 5% .
5% 95% .
39
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
) 30(
1.96 2.58
) (students t-distribution Distribution de Student t
) (degree of freedom Degr de libert .16
bi ) E (Y / X
)The
standard error of the sampling distribution of b Lcart-type de la
(distribution dchantillonnage
. b
:
) E [b E (b )] = (bM
2
2
M M .
) (unknown - Inconnue M .
:
= b2 SS Re sd
n k 1 . 1x 2
n K
. 17
= S
2
X
n 1
17
:
b = Y = ( + X + u) = + X + U
40
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
= b m Zc b )(1
= b m tc b )( 2
) (1 ) (n ) 30(
) (2 ) 30( .
) E (Y X Y:
) E (Y X ) = XY = + X = Y + b(X X
]) 2 = 2 [Y + b( X X
xy
) = 2 (Y ) + ( X X ) 2 (b
2 2
xy
yx
2
yx
2
) + (X X
2
= 2xy
n X2
1 ( X X )2 1 X2
= yx + = + :
X2 n X2
yx
xy
n )
Y m Z c yx
X
= =0
X2
X
= X .X = 1
X2
:
b = + U
b = U
2
X 1
E [b ] = E (U ) SS Re sd
.
2 2
2
= 2
2
= .
X X 2
n k 1
xy
41
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
)(General - Gnrale
:
- .
- .
-
.
-
.18
) .(Statistical hypotheses Hypothse statistiques
.
). (Testable hypotheses Hypothses testables
. (The null
) hypothesis: H0 Hypothse nulle: H0 (The alternative
(hypothesis: H1 Hypothse alternative: H1 .
) (Linear - Linaire
) (Nonlinear Non linaire
: 19
Y = b0 + b1 X
b0 b1 .
b0
(Intercept - Constante) b0
X = 0.0 b0 .
b0 .
) (Elasticity - Elasticit
b0
18
Fred N. kerlinger., " Foundation of behavioral Research"- 2nd ed. holt-Rinehatr and Winston, inc.,
1973,pp: 16-22
19 . ) are raised
(.to the first power
42
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
Y
= Y =
X
X
Y X
= .
Y X
Y X
=.
X Y
X X Y Y : 21
X
= b1.
Y
X
= b1.
b0 + b1 X
b1 X
=
b0 + b1 X
:
- b0 = 0.0 ) (
.
- b0
).(Elastic - Elastique
.
- b0
)inelastic -
.(inlastique
.
20 . : . :
237-236
X Y
XY 21
43
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
b1
.
b0 b1
) (Significance - signification
:
H0 : = 0
H1 : 0
.
.22
.
): 1978-1970 ( 136
: EViews
File \ New \ Workfile :
22 (F- Test) F
) (Equivalent Models
) . (Linear dependency (F- distributions) F
) (Fisher .
44
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
: :
45
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
46
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
47
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
48
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
49
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
Click the right mouse button on graph \ Choose Options \ Linear Force Thru Zero
50
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
51
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
52
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
SPSS Excel
53
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
54
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
55
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
56
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
57
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
Regression
Model Summary
ANOVAb
Sum of
Model Squares df Mean Square F Sig.
1 Regression 1.53E+09 1 1525525618 73.177 .000a
Residual 2.71E+08 13 20846952.50
Total 1.80E+09 14
a. Predictors: (Constant), GDP
b. Dependent Variable: Housing Prices
Coefficientsa
Unstandardized Standardized
Coefficients Coefficients
Model B Std. Error Beta t Sig.
1 (Constant) 5795.077 3180.041 1.822 .091
GDP 20.358 2.380 .921 8.554 .000
a. Dependent Variable: Housing Prices
58
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
Exercises
1. The following table gives data on weekly consumption expenditure (Y) (in
dollars) and weekly family income (X) (in dollars). (Le tableau suivant prsente
des donnes sur la consommation hebdomadaire (Y) (en dollars) et le revenu
familial hebdomadaire (X) (en dollars)). Table 3: Income &
consumption
a) Is this a time series regression or a cross-sectional X ($) Y ($)
regression? (est-ce une srie chronologique ou une 80 70
srie en coupe instantane ?) 100 65
b) Use the least squares method pour regress Y on X. 120 90
(Utiliser la mthode des moindres carrs pour 140 95
rgresser Y sur X) 160 110
c) How do you interpret the parameters of this 180 115
regression? (Comment interprter les paramtres de 200 120
cette rgression) 220 140
d) Test the significance of these parameters, = 0.05. 240 155
(tester la signification de ces paramtres, = 0.05) 260 150
e) Find the coefficient of determination; interpret.
(Trouver le coefficient de dtermination ; interprter)
4. In the following table we have data on CPI and S&P 500 Index for the years 1990
to 2002. (Le tableau suivant fournit des donnes annuelles sur lindice des prix
consommation (CPI) et Lindice S&P 500 sur la priode 1990 2002)
Table 4: S&P and CPI (1990-2002)
a) Plot the data on a scattergram with the S&P Year
S&P CPI
500 on the vertical axis and IPC on the (Anne)
horizontal axis. (Reprsenter graphiquement 1990 334.59 130.7
lS&P 500 sur laxe des ordonnes et lIPC sur 1991 376.18 136.2
celui des abscisses) 1992 415.74 140.3
b) What can you say about the relationship 1993 451.41 144.5
between the two indexes? What does economic 1994 460.42 148.2
theory have to say about the relationship? (Que 1995 541.72 152.4
peut-on dire de la relation entre ces deux 1996 670.50 156.9
indices? Que dit la thorie conomique sur 1997 873.43 160.5
cette relation?) 1998 1085.50 163.0
Consider the following regression model. 1999 1327.33 166.6
(Considrer le modle de rgression suivant:) 2000 1427.22 172.2
(S&P)t = 1 + 2 CPIt + ut 2001 1194.18 177.1
Use the least squares to estimate this equation and interpret the results.
(Utiliser les moindres carrs pour estimer cette quation et interprter les
rsultats)
5. The following table Gives data on the gold price, CPI, and New York Stock
Exchange (NYSE) over the period 1977-1991. (Le tableau suivant prsente des
donnes sur le prix de lor, lindice des prix la consommation et lindice du New
York Stock Exchange (NYSE) pour la priode 1977-1991)
Table 5: NYSE, CPI and Gold price (1977-1991)
a) Plot on the same scattergram the NY stock CPI Gold
Gold prices, CPI, and NYSE Exchange 1982 price in
Year
(NYSE) 31 $ per
index. (Reprsenter sur le mme Dec.1965 1984 once
(Anne)
graphique les prix de lor, = 100 = 100 (31.1g)
lindice des prix et lindice du 53.69 60.6 147.98 1977
NYSE) 53.70 65.2 193.44 1978
b) An investment is supposed to be
58.32 72.6 307.62 1979
a protection against inflation. To
68.10 82.4 612.51 1980
test this hypothesis, suppose that
74.02 90.9 459.61 1981
you decide to adjust the
68.93 96.5 376.01 1982
following model, given that the
92.63 99.6 423.83 1983
scattergram in (a) is well fitting
92.46 103.9 360.29 1984
the data (Un investissement est
suppos tre une protection 108.90 107.6 317.3 1985
contre linflation. Pour tester 136.00 109.6 367.87 1986
cette hypothse, supposez que 161.70 113.6 446.5 1987
vous dcidez dajuster le modle 149.91 118.3 436.93 1988
suivant, en supposant que le 180.02 124.0 381.28 1989
graphique de dispersion du (a) 183.46 130.7 384.04 1990
est appropri) 206.33 136.2 362.04 1991
Gold pricet = 1 + 2 CPI t + ut
NYSE indext = 1 + 2 CPI t + ut
60
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
.
.
) Simultaneous equation models
.(Modles quations simultanes
) (Multiple linear regression - Rgression linaire multiple
)(Multivariate analysis Analyse multivarie
) (Simultaneous models Modles simultans .
.
) (Partial regression coefficient Coefficient de regression partielle
.
b1
Y X 1 . bi
Y
X i ) Holding the effect
of other independent variables constant En gardant les autres variables
.(indpendantes constantes Y
X 1 X 2:
Y = 0 + 1 X1 + 2 X 2 + U
Y K :
Y = 0 + 1 X 1 + 2 X 2 + .... + K X K + U
Y X 1 . X 2
Y Y
K .
61
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
:
.
: .
:
.
: .
:
:
: :
Y X 1
X 2:
Y = b0 + b1 X 1 + b2 X 2 + e
b0 , b1 , b2
) e 2 = (Y Y .
)2
)
y = b1 x1 + b2 x 2 )(1
62
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
x 1 y = b1 x 12 + b2 x 1 x 2 )(2
) (1 x 2 :
x 2 y = b1 x 1 x 2 + b2 x 22 )(3
b1 b2 ) (2 ) (3:
x1 y x1 x 2
x2 y x 22 ) ( x1 y ) ( x 22 ) ( x 2 y )( x1 x 2
= b1 x12 x1 x 2
= ( x ) ( x ) ( x x
2 2
)2
1 2 1 2
x1 x 2 x 22
x12 x1 y
x1 x 2 x2 y ) ( x 2 y ) ( x12 ) ( x1 y )( x1 x 2
= b2 x12 x1 x 2
= ( x12 )( x22 ) ( x1 x2 )2
x1 x 2 x 22
e 2
= 2( y b1 x1 b2 x2 )( x1 ) = 0
b1
e 2
= 2( y b1 x1 b2 x 2 )( x 2 ) = 0
b2
63
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
b0
n : Y = b0 + b1 X 1 + b2 X 2
b1 X 1 b2 X 2
Y
= + +
b0 n
n n n n
Y = b0 + b1 X 1 + b2 X 2
:
b0 = Y b1 X 1 b2 X 2
x22
S 2
b1 = SSRe sd
n k 1 . ( x 2 ) ( x 2 ) ( x x )2
1 2 1 2
x12
2
=
SS Re sd
S b2 n k 1
. ( x 2 ) ( x 2 ) ( x x )2
1 2 1 2
) ( H 0 ) ( H 1 :
: .
H 0 = 1 = 2 = 3 ...... = = 0
:
) . -
(.
t :
bi
= t S bi
:
= bi m t (0.05 , n k 1 ) S bi
64
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
) (Time series data Donnes chronologiques
) Cross-section data Donnes
.(en coupes instantanes
) Z (
) X 1 : ( ) X 2
( :
)
Y = 4.20 + 3.49 X 1 0.57 X 2
b1 ) (
) (
Z
Z 3.49 ) ( .
Z
Z 349 . b2
Z
Z 57 .
X i Y ) Cetris paribus Toutes
.(choses gales par ailleurs
) Price index of common stocks Indice des prix des actions
(ordinaires ) Retained earnings Gains
(retenus )The interest rate at which
funds could be borrowed Le taux dintrt auquel les fonds peuvent tre
(emprunts ) The level of investment Le
(niveau dinvestissement ) Quarterly data
(Donnes trimestrielles :
)
Y = 2.3 + 0.091 X 1 + 1.87 X 2 + 0.02 X 3
b1
) (One-point change Un point de changement
)(Cetris paribus - Toutes choses gales par ailleurs
0.091 .
b2 b3 Y X i
.
65
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
23
: :
) (Matrices - Matrices
) Computer software -
(Logiciels .
Y X1 X2
:
Y = b0 + b1 X 1 + b2 X 2 ) (1
) (1 :
Y = b0 n + b1 X 1 + b2 X 2 )(2
) (1 X1 :
X 1Y = b0 X 1 + b1 X 12 + b2 X 1 X 2 )(3
) (1 X2 :
X 2Y = b0 X 2 + b1 X 1 X 2 + b2 X 22 ) (4
) (2 ) (3 ) (4 :
)( XX b )( XY
n X1 X2 b0 Y
X X 2
X X * b = X Y
1 1 1 2 1 1
X2 X1 X2 X22 b2 X2Y
X (Transpose
) (Vector - Vecteur X (Transpos
X .Y
) Observations
(- Observations Y X1:X2
X1 X2 Y
X 11 X 12 Y1
X 21 X 22 Y2
X 31 X 32 Y3
X 41 X 42 Y4
: X X
X X ) ( X X
1 X11 X12
1 1 1 1 n X1 X2
X X X X 1 X21 X22 X X2 X X
11 21 31 41 1 X = 1 1 2
X32 1
67
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
: X Y
24
Ward, J.H., and Jennings, E., "introduction to linear Models". Prentice Hall international, inc., 1973 pp:
26-28
68
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
X Y (X Y )
Y1
1 1 1 1 Y
X X 21 X 31 X 41 2
Y
= X 1Y
11 Y3
X 12 X 22 X 32
X 42 X 2Y
Y4
) (2 ) (3 ) (4
:
) ( X X ) b = ( X Y
) (premultiply - Prmultiplier ( X X ) 1
:
) b = ( X X ) ( X Y
1
S 2 b0 = C 00 .
SS Re sd
n k 1
S 2 b1 = C11 . SS Re sd
n k 1
S 2 b2 = C 22 . SSRe sd
n k 1
C 00 C 01 C 02
(X X ) 1 = C C C
10 11 12
C 20 C 21 C 22
e 2 = ee
:
SS T = (Y Y ) nY
:
:
)Standard partial
(regression coefficients Coefficients standardiss de rgression partielle
: 26
= R 1 . V
: ) (vector - Vecteur
).(Beta weights Pondrations Beta
: R 1 ) (inverse - Inverse .
: V ) (vector - Vecteur
.
b=
SY
SX
26 ) (interpreting - Interprtion
.
70
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
: 27
SSRe sd
S 2
b = n k 1
) X ( 1 R 2j
2
R 2j
:
R 2j = 1 1
rj
j
r
R 1 .
: :
) (Residuals - Rsidus
b1b2
.
:
Y = b0 + b1 X 1 + b2 X 2
bi Y
.
X2 X1 Y
X1 X2 .b1 X1
X2 Y X2 X1
.b2
R2
)( Y . X S
R2 :
= R2 = 1
SS Re g SS Re sd
SS T SS T )(1
Y X1 X2
:
)(
)
) SS Re sd = e 2 = e( y y ) = e ( y b1 x1 b2 x2
SS Re sd = ey b1 ex1 b2 ex 2 )(2
ex1 ex 2
:
e2
b1 = 2 ( y b1 x1 b2 x2 ) ( x1 ) = 0
2 ex1 = 0 :
e 2
b2 = 2 ( y b1 x 1 b2 x 2 )( x 2 ) = 0 :
2 ex 2 = 0 :
) (2:
SS Re sd = e y )(3
e ) (3 :
) SS Re sd = y ( y b1 x 1 b2 x 2
72
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
SS Re sd = y 2 b1 x 1 y b2 x 2 y ) (4
SS Re sd ) (1 ) (4 :
y 2 b1 x 1 y b 2 x 2 y
R2 = 1 y2
:
b1 x 1 y + b2 x2 y
= R2 y2 )(5
(Denominator - Dnominateur) y 2
R 2
. ) (Numerator - Numrateur
xy x .
R 2 : 28
b ( X Y ) nY 2
= R 2
(Y Y ) nY 2
:
(Y Y ) = Y 2
R 2
29
:
R 2 = V1 1 + V 2 2 + ...... + V k k
)
n k 1
R 2 ) (upward biased Biaise vers le haut
) ( Adjusted R2 -R2 ajuste : 30
R 2 = 1 1 R2 ([ ) n 1
n k 1
] )(6
28
Johnstom p:129
29
Kerlinger and pedhazur, p:75
R 2 R 2 30
) (6
n k 1 k = n 1
) (1 R
2
. R 2 = 1 10 )(observations - Observations
9
. R 2 = 1 n k ( N k 1) = n ( n 1) 1 = 0
R
2
R 2 R
2
n
. 0 R 1
2
) (negative - ngative
:
k = 6 n = 10 R 2 = 0.50
10 1
) R 2 = 1 (1 0 .50 = 1 (1 .50 ) = 0 .50
10 6 1
30 .
200 600.
R 2 Y 0.36
R 10 90
2
150 :
R 2
R 2 . R 2
.
:
)Evaluating the relative importance of the
independent variables in explaining the variation of the dependent variable
Evaluation de limportance relative des variables indpendante dans
.(lexplication de la variabilit de la variable dpendante
) Significance -
(Signification .
) (Equivalent models Modles quivalents
) (linear Dependency Dpendance linaire
74
Professor Dr. Charbaji at CHARBAJI Consultants. Beirut Lebanon. 009611 355046
b2 b1
) X 1 = F (x2
Y = b0 + b1 X 1 + b2 X 2 + b3 X 3
.
.
X 3
.
Y = b0 + b1 X 1 + b2 X 2
.
RF2 RR2
F:
=F
) (R R ) ( K K
2 2
1 2
)(1 R ) (n k 1
F R
2
F 1
.
k1 K 2 .
X 2 X 3 X 2
:
Y = b0 + b1 X 1 + b3 X 3
.
b1 b1
.
F .
) (Significant - Significative .
) (RF2 RR2
) semi partial correlation semi corrlation
(partielle . 31
R 2
.
Y 10
Y X 6
Y )
R 2 (
X 6 .
) Stepwise solution .
(Solution pas pas
(Statistical Packages of the Social Science) SPSS
).BMD (Biomedical Statistical Package
76