Remarque concernant l’estimation par régression lorsque la taille de la population est inconnue
2. Estimateurs par régressionRemarque concernant l’estimation par régression lorsque la taille de la population est inconnue
2. Estimateurs par régression
Sous des conditions générales de régularité
(Isaki et Fuller 1982; Montanari 1987), une approximation de l’estimateur par
régression (1.1) est
Y
˜
REG
=
Y
^
π
+
(
X
−
X
^
π
)
Τ
B
,
(
2.1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaia
WaaSbaaSqaaiaabkfacaqGfbGaae4raaqabaGccaaI9aGabmywayaa
jaWaaSbaaSqaaiabec8aWbqabaGccqGHRaWkdaqadaqaaiaahIfacq
GHsislceWHybGbaKaadaWgaaWcbaGaeqiWdahabeaaaOGaayjkaiaa
wMcaamaaCaaaleqabaGaeyiPdqfaaOGaaCOqaiaaiYcacaaMf8UaaG
zbVlaaywW7caaMf8UaaGzbVlaacIcacaaIYaGaaiOlaiaaigdacaGG
Paaaaa@5401@
où
B
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOqaaaa@37F7@
est la limite en probabilité de
B
^
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
aaaa@3807@
lorsque la taille de l’échantillon et celle de
la population tendent vers l’infini. Pour de grands échantillons, la variance
de l’estimateur par régression (1.1) peut être étudiée avec (2.1). Notons que
Y
˜
REG
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaia
WaaSbaaSqaaiaabkfacaqGfbGaae4raaqabaaaaa@3AAC@
est sans biais sous le plan de sondage
p
(
s
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamiCamaabm
aabaGaam4CaaGaayjkaiaawMcaaaaa@3AA2@
et peut être réexprimé sous la forme :
Y
˜
REG
=
X
Τ
B
+
∑
i
∈
s
d
i
E
i
,
(
2.2
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaia
WaaSbaaSqaaiaabkfacaqGfbGaae4raaqabaGccaaI9aGaaCiwamaa
CaaaleqabaGaeyiPdqfaaOGaaCOqaiabgUcaRmaaqafabeWcbaGaam
yAaiabgIGiolaadohaaeqaniabggHiLdGccaaMc8UaamizamaaBaaa
leaacaWGPbaabeaakiaadweadaWgaaWcbaGaamyAaaqabaGccaaISa
GaaGzbVlaaywW7caaMf8UaaGzbVlaaywW7caGGOaGaaGOmaiaac6ca
caaIYaGaaiykaaaa@56E5@
où
E
i
=
y
i
−
x
i
Τ
B
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyramaaBa
aaleaacaWGPbaabeaakiaai2dacaWG5bWaaSbaaSqaaiaadMgaaeqa
aOGaeyOeI0IaaCiEamaaDaaaleaacaWGPbaabaGaeyiPdqfaaOGaaC
Oqaiaac6caaaa@421A@
Une approximation de la variance par
rapport au plan de
Y
^
REG
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabkfacaqGfbGaae4raaqabaaaaa@3AAD@
peut être donnée par
AV
p
(
Y
^
REG
)
=
∑
i
∈
U
∑
j
∈
U
Δ
i
j
E
i
π
i
E
j
π
j
,
(
2.3
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaeyqaiaabA
fadaWgaaWcbaGaamiCaaqabaGcdaqadaqaaiqadMfagaqcamaaBaaa
leaacaqGsbGaaeyraiaabEeaaeqaaaGccaGLOaGaayzkaaGaaGypam
aaqafabeWcbaGaamyAaiabgIGiolaadwfaaeqaniabggHiLdGcdaae
qbqabSqaaiaadQgacqGHiiIZcaWGvbaabeqdcqGHris5aOGaaGPaVl
abfs5aenaaBaaaleaacaWGPbGaamOAaaqabaGcdaWcaaqaaiaadwea
daWgaaWcbaGaamyAaaqabaaakeaacqaHapaCdaWgaaWcbaGaamyAaa
qabaaaaOWaaSaaaeaacaWGfbWaaSbaaSqaaiaadQgaaeqaaaGcbaGa
eqiWda3aaSbaaSqaaiaadQgaaeqaaaaakiaaiYcacaaMf8UaaGzbVl
aaywW7caaMf8UaaGzbVlaacIcacaaIYaGaaiOlaiaaiodacaGGPaaa
aa@6587@
où
Δ
i
j
=
π
i
j
−
π
i
π
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaeuiLdq0aaS
baaSqaaiaadMgacaWGQbaabeaakiaai2dacqaHapaCdaWgaaWcbaGa
amyAaiaadQgaaeqaaOGaeyOeI0IaeqiWda3aaSbaaSqaaiaadMgaae
qaaOGaeqiWda3aaSbaaSqaaiaadQgaaeqaaaaa@45E2@
et
π
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaeqiWda3aaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3AF2@
est la probabilité d’inclusion du second ordre
pour les unités
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyAaaaa@381A@
et
j
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamOAaiaac6
caaaa@38CD@
Notons que
B
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOqaaaa@37F7@
peut être estimée selon
l’approche assistée par modèle (Särndal, Swensson et Wretman 1992) et
l’approche de la variance optimale (Montanari 1987). Les deux méthodes
permettent d’obtenir des estimateurs approximativement sans biais. Dans le cas
de l’approche assistée par modèle, les propriétés de base (biais et variance)
sont valides même lorsque le modèle n’est pas spécifié correctement. Sous
l’approche de la variance optimale, aucune hypothèse n’est formulée au sujet de
la variable d’intérêt.
L’estimateur assisté par modèle de Särndal et coll. (1992)
suppose un modèle de travail entre la variable d’intérêt
(
y
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaWaaeWaaeaaca
WG5baacaGLOaGaayzkaaaaaa@39B3@
et les variables auxiliaires
(
x
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaWaaeWaaeaaca
WH4baacaGLOaGaayzkaaGaaiOlaaaa@3A68@
Le modèle de travail est
désigné par
m
:
y
i
=
x
i
Τ
β
+
ε
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyBaiaaiQ
dacaWG5bWaaSbaaSqaaiaadMgaaeqaaOGaaGypaiaahIhadaqhaaWc
baGaamyAaaqaaiabgs6aubaakiaahk7acqGHRaWkcqaH1oqzdaWgaa
WcbaGaamyAaaqabaaaaa@4459@
où
β
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOSdaaa@386A@
est un vecteur de
p
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamiCaaaa@3821@
paramètres inconnus,
E
m
(
ε
i
|
x
i
)
=
0
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyramaaBa
aaleaacaWGTbaabeaakmaabmaabaWaaqGaaeaacqaH1oqzdaWgaaWc
baGaamyAaaqabaGccaaMc8oacaGLiWoacaaMc8UaaCiEamaaBaaale
aacaWGPbaabeaaaOGaayjkaiaawMcaaiaai2dacaaIWaGaaiilaaaa
@4674@
V
m
(
ε
i
|
x
i
)
=
σ
i
2
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamOvamaaBa
aaleaacaWGTbaabeaakmaabmaabaWaaqGaaeaacqaH1oqzdaWgaaWc
baGaamyAaaqabaGccaaMc8oacaGLiWoacaaMc8UaaCiEamaaBaaale
aacaWGPbaabeaaaOGaayjkaiaawMcaaiaai2dacqaHdpWCdaqhaaWc
baGaamyAaaqaaiaaikdaaaGccaGGSaaaaa@496F@
et
Cov
m
(
ε
i
,
ε
j
|
x
i
,
x
j
)
=
0,
i
≠
j
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaae4qaiaab+
gacaqG2bWaaSbaaSqaaiaad2gaaeqaaOWaaeWaaeaadaabcaqaaiab
ew7aLnaaBaaaleaacaWGPbaabeaakiaaiYcacqaH1oqzdaWgaaWcba
GaamOAaaqabaGccaaMc8oacaGLiWoacaaMc8UaaCiEamaaBaaaleaa
caWGPbaabeaakiaaiYcacaWH4bWaaSbaaSqaaiaadQgaaeqaaaGcca
GLOaGaayzkaaGaaGypaiaaicdacaaISaGaamyAaiabgcMi5kaadQga
caGGUaaaaa@5315@
Sous cette approche,
B
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOqaaaa@37F7@
dans l’équation (2.1) est
l’estimateur des moindres carrés ordinaires de
β
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOSdaaa@386A@
dans la population et est
donné par
B
GREG
=
(
∑
i
∈
U
c
i
x
i
x
i
Τ
)
−
1
(
∑
i
∈
U
c
i
x
i
y
i
)
,
(
2.4
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOqamaaBa
aaleaacaqGhbGaaeOuaiaabweacaqGhbaabeaakiaai2dadaqadaqa
amaaqafabeWcbaGaamyAaiabgIGiolaadwfaaeqaniabggHiLdGcca
aMc8Uaam4yamaaBaaaleaacaWGPbaabeaakiaahIhadaWgaaWcbaGa
amyAaaqabaGccaWH4bWaa0baaSqaaiaadMgaaeaacqGHKoavaaaaki
aawIcacaGLPaaadaahaaWcbeqaaiabgkHiTiaaigdaaaGcdaqadaqa
amaaqafabeWcbaGaamyAaiabgIGiolaadwfaaeqaniabggHiLdGcca
aMc8Uaam4yamaaBaaaleaacaWGPbaabeaakiaahIhadaWgaaWcbaGa
amyAaaqabaGccaWG5bWaaSbaaSqaaiaadMgaaeqaaaGccaGLOaGaay
zkaaGaaGilaiaaywW7caaMf8UaaGzbVlaaywW7caaMf8Uaaiikaiaa
ikdacaGGUaGaaGinaiaacMcaaaa@6951@
où
c
i
=
σ
i
−
2
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaam4yamaaBa
aaleaacaWGPbaabeaakiaai2dacqaHdpWCdaqhaaWcbaGaamyAaaqa
aiabgkHiTiaaikdaaaGccaGGUaaaaa@3F42@
Cela donne l’estimateur suivant pour le total
Y
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamywaaaa@380A@
Y
^
GREG
=
Y
^
π
+
(
X
−
X
^
π
)
Τ
B
^
GREG
,
(
2.5
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabEeacaqGsbGaaeyraiaabEeaaeqaaOGaaGypaiqa
dMfagaqcamaaBaaaleaacqaHapaCaeqaaOGaey4kaSYaaeWaaeaaca
WHybGaeyOeI0IabCiwayaajaWaaSbaaSqaaiabec8aWbqabaaakiaa
wIcacaGLPaaadaahaaWcbeqaaiabgs6aubaakiqahkeagaqcamaaBa
aaleaacaqGhbGaaeOuaiaabweacaqGhbaabeaakiaaiYcacaaMf8Ua
aGzbVlaaywW7caaMf8UaaGzbVlaacIcacaaIYaGaaiOlaiaaiwdaca
GGPaaaaa@5847@
où
B
^
GREG
=
(
∑
i
∈
s
c
i
d
i
x
i
x
i
Τ
)
−
1
(
∑
i
∈
s
c
i
d
i
x
i
y
i
)
.
(
2.6
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
WaaSbaaSqaaiaabEeacaqGsbGaaeyraiaabEeaaeqaaOGaaGypamaa
bmaabaWaaabuaeqaleaacaWGPbGaeyicI4Saam4Caaqab0GaeyyeIu
oakiaaykW7caWGJbWaaSbaaSqaaiaadMgaaeqaaOGaamizamaaBaaa
leaacaWGPbaabeaakiaahIhadaWgaaWcbaGaamyAaaqabaGccaWH4b
Waa0baaSqaaiaadMgaaeaacqGHKoavaaaakiaawIcacaGLPaaadaah
aaWcbeqaaiabgkHiTiaaigdaaaGcdaqadaqaamaaqafabeWcbaGaam
yAaiabgIGiolaadohaaeqaniabggHiLdGccaaMc8Uaam4yamaaBaaa
leaacaWGPbaabeaakiaadsgadaWgaaWcbaGaamyAaaqabaGccaWH4b
WaaSbaaSqaaiaadMgaaeqaaOGaamyEamaaBaaaleaacaWGPbaabeaa
aOGaayjkaiaawMcaaiaai6cacaaMf8UaaGzbVlaaywW7caaMf8UaaG
zbVlaacIcacaaIYaGaaiOlaiaaiAdacaGGPaaaaa@6DBB@
L’estimateur optimal de Montanari
(1987), obtenu en minimisant la variance par rapport au plan de
Y
˜
REG
=
Y
^
π
+
(
X
−
X
^
π
)
Τ
B
,
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaia
WaaSbaaSqaaiaabkfacaqGfbGaae4raaqabaGccaaI9aGabmywayaa
jaWaaSbaaSqaaiabec8aWbqabaGccqGHRaWkdaqadaqaaiaahIfacq
GHsislceWHybGbaKaadaWgaaWcbaGaeqiWdahabeaaaOGaayjkaiaa
wMcaamaaCaaaleqabaGaeyiPdqfaaOGaaCOqaiaaiYcaaaa@48B9@
est
Y
˜
OPT
=
Y
^
π
+
(
X
−
X
^
π
)
Τ
B
OPT
,
(
2.7
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaia
WaaSbaaSqaaiaab+eacaqGqbGaaeivaaqabaGccaaI9aGabmywayaa
jaWaaSbaaSqaaiabec8aWbqabaGccqGHRaWkdaqadaqaaiaahIfacq
GHsislceWHybGbaKaadaWgaaWcbaGaeqiWdahabeaaaOGaayjkaiaa
wMcaamaaCaaaleqabaGaeyiPdqfaaOGaaCOqamaaBaaaleaacaqGpb
GaaeiuaiaabsfaaeqaaOGaaGilaiaaywW7caaMf8UaaGzbVlaaywW7
caaMf8UaaiikaiaaikdacaGGUaGaaG4naiaacMcaaaa@56CE@
où
B
OPT
=
{
V
(
X
^
π
)
}
−
1
Cov
(
X
^
π
,
Y
^
π
)
=
(
∑
i
∈
U
∑
j
∈
U
Δ
i
j
x
i
π
i
x
j
Τ
π
j
)
−
1
(
∑
i
∈
U
∑
j
∈
U
Δ
i
j
x
i
π
i
y
j
π
j
)
.
(
2.8
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaqbaeaabiGaaa
qaaiaahkeadaWgaaWcbaGaae4taiaabcfacaqGubaabeaaaOqaaiaa
i2dadaGadaqaaiaadAfadaqadaqaaiqahIfagaqcamaaBaaaleaacq
aHapaCaeqaaaGccaGLOaGaayzkaaaacaGL7bGaayzFaaWaaWbaaSqa
beaacqGHsislcaaIXaaaaOGaae4qaiaab+gacaqG2bWaaeWaaeaace
WHybGbaKaadaWgaaWcbaGaeqiWdahabeaakiaaiYcaceWGzbGbaKaa
daWgaaWcbaGaeqiWdahabeaaaOGaayjkaiaawMcaaaqaaaqaaiaai2
dadaqadaqaamaaqafabeWcbaGaamyAaiabgIGiolaadwfaaeqaniab
ggHiLdGcdaaeqbqabSqaaiaadQgacqGHiiIZcaWGvbaabeqdcqGHri
s5aOGaaGPaVlabfs5aenaaBaaaleaacaWGPbGaamOAaaqabaGcdaWc
aaqaaiaahIhadaWgaaWcbaGaamyAaaqabaaakeaacqaHapaCdaWgaa
WcbaGaamyAaaqabaaaaOWaaSaaaeaacaWH4bWaa0baaSqaaiaadQga
aeaacqGHKoavaaaakeaacqaHapaCdaWgaaWcbaGaamOAaaqabaaaaa
GccaGLOaGaayzkaaWaaWbaaSqabeaacqGHsislcaaIXaaaaOWaaeWa
aeaadaaeqbqabSqaaiaadMgacqGHiiIZcaWGvbaabeqdcqGHris5aO
WaaabuaeqaleaacaWGQbGaeyicI4Saamyvaaqab0GaeyyeIuoakiaa
ykW7cqqHuoardaWgaaWcbaGaamyAaiaadQgaaeqaaOWaaSaaaeaaca
WH4bWaaSbaaSqaaiaadMgaaeqaaaGcbaGaeqiWda3aaSbaaSqaaiaa
dMgaaeqaaaaakmaalaaabaGaamyEamaaBaaaleaacaWGQbaabeaaaO
qaaiabec8aWnaaBaaaleaacaWGQbaabeaaaaaakiaawIcacaGLPaaa
caaIUaaaaiaaywW7caaMf8UaaGzbVlaaywW7caaMf8Uaaiikaiaaik
dacaGGUaGaaGioaiaacMcaaaa@971D@
L’estimateur
optimal pour le total
Y
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamywaaaa@380A@
est estimé par
Y
^
OPT
=
Y
^
π
+
(
X
−
X
^
π
)
Τ
B
^
OPT
,
(
2.9
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaab+eacaqGqbGaaeivaaqabaGccaaI9aGabmywayaa
jaWaaSbaaSqaaiabec8aWbqabaGccqGHRaWkdaqadaqaaiaahIfacq
GHsislceWHybGbaKaadaWgaaWcbaGaeqiWdahabeaaaOGaayjkaiaa
wMcaamaaCaaaleqabaGaeyiPdqfaaOGabCOqayaajaWaaSbaaSqaai
aab+eacaqGqbGaaeivaaqabaGccaaISaGaaGzbVlaaywW7caaMf8Ua
aGzbVlaaywW7caGGOaGaaGOmaiaac6cacaaI5aGaaiykaaaa@56E1@
où
B
^
OPT
=
(
∑
i
∈
s
∑
j
∈
s
Δ
i
j
π
i
j
x
i
π
i
x
j
Τ
π
j
)
−
1
(
∑
i
∈
s
∑
j
∈
s
Δ
i
j
π
i
j
x
i
π
i
y
j
π
j
)
.
(
2.10
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
WaaSbaaSqaaiaab+eacaqGqbGaaeivaaqabaGccaaI9aWaaeWaaeaa
daaeqbqabSqaaiaadMgacqGHiiIZcaWGZbaabeqdcqGHris5aOWaaa
buaeqaleaacaWGQbGaeyicI4Saam4Caaqab0GaeyyeIuoakmaalaaa
baGaeuiLdq0aaSbaaSqaaiaadMgacaWGQbaabeaaaOqaaiabec8aWn
aaBaaaleaacaWGPbGaamOAaaqabaaaaOWaaSaaaeaacaWH4bWaaSba
aSqaaiaadMgaaeqaaaGcbaGaeqiWda3aaSbaaSqaaiaadMgaaeqaaa
aakmaalaaabaGaaCiEamaaDaaaleaacaWGQbaabaGaeyiPdqfaaaGc
baGaeqiWda3aaSbaaSqaaiaadQgaaeqaaaaaaOGaayjkaiaawMcaam
aaCaaaleqabaGaeyOeI0IaaGymaaaakmaabmaabaWaaabuaeqaleaa
caWGPbGaeyicI4Saam4Caaqab0GaeyyeIuoakmaaqafabeWcbaGaam
OAaiabgIGiolaadohaaeqaniabggHiLdGcdaWcaaqaaiabfs5aenaa
BaaaleaacaWGPbGaamOAaaqabaaakeaacqaHapaCdaWgaaWcbaGaam
yAaiaadQgaaeqaaaaakmaalaaabaGaaCiEamaaBaaaleaacaWGPbaa
beaaaOqaaiabec8aWnaaBaaaleaacaWGPbaabeaaaaGcdaWcaaqaai
aadMhadaWgaaWcbaGaamOAaaqabaaakeaacqaHapaCdaWgaaWcbaGa
amOAaaqabaaaaaGccaGLOaGaayzkaaGaaGOlaiaaywW7caaMf8UaaG
zbVlaaywW7caaMf8UaaiikaiaaikdacaGGUaGaaGymaiaaicdacaGG
Paaaaa@881F@
Il est à noter que, pour que nous puissions calculer les vecteurs de
régression, la première composante qui les définit doit être inversible.
Nous pouvons nous assurer qu’elle l’est en réduisant le nombre de variables
auxiliaires qui entrent dans la régression si l’efficience de l’estimateur par
régression qui en découle n’en souffre pas trop. Par contre, si la perte
d’efficience est importante, nous pouvons inverser ces matrices singulières en
utilisant des inverses généralisés.
Comme il est mentionné dans l’introduction, les totaux de population
ne sont pas nécessairement connus pour toutes les composantes du vecteur
auxiliaire
x
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEaiaac6
caaaa@38DF@
La régression utilise
normalement les variables auxiliaires pour lesquelles un total de population
correspondant est connu. En décomposant
x
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaBa
aaleaacaWGPbaabeaaaaa@3947@
en
(
1,
x
i
*
Τ
)
Τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaWaaeWaaeaaca
aIXaGaaGilaiaahIhadaqhaaWcbaGaamyAaaqaaiaaiQcacqGHKoav
aaaakiaawIcacaGLPaaadaahaaWcbeqaaiabgs6aubaaaaa@403B@
où
x
i
*
=
(
x
2 i
, … ,
x
p i
)
Τ
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaDa
aaleaacaWGPbaabaGaaiOkaaaakiaai2dadaqadaqaaiaadIhadaWg
aaWcbaGaaGOmaiaadMgaaeqaaOGaaGilaiablAciljaaiYcacaWG4b
WaaSbaaSqaaiaadchacaWGPbaabeaaaOGaayjkaiaawMcaamaaCaaa
leqabaGaeyiPdqfaaOGaaiilaaaa@473F@
Singh et Raghunath (2011) ont
proposé un estimateur semblable au GREG qui suppose une régression fondée sur
une ordonnée à l’origine et la variable
x
*
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaCa
aaleqabaGaaGOkaaaakiaacYcaaaa@39C8@
même si seul le total de
population de
x
*
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaCa
aaleqabaGaaGOkaaaaaaa@390E@
est connu.
Si
N
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamOtaaaa@37FF@
est inconnu et que le total
de population de
x
*
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaCa
aaleqabaGaaGOkaaaaaaa@390E@
est connu, leur estimateur
est
Y
^
SREG
=
Y
^
π
+
(
X
*
−
X
^
π
*
)
Τ
B
^
2
,
GREG
,
(
2.11
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEeaaeqaaOGaaGypaiqa
dMfagaqcamaaBaaaleaacqaHapaCaeqaaOGaey4kaSYaaeWaaeaaca
WHybWaaWbaaSqabeaacaaIQaaaaOGaeyOeI0IabCiwayaajaWaa0ba
aSqaaiabec8aWbqaaiaaiQcaaaaakiaawIcacaGLPaaadaahaaWcbe
qaaiabgs6aubaakiqahkeagaqcamaaBaaaleaacaaIYaGaaiilaiaa
bEeacaqGsbGaaeyraiaabEeaaeqaaOGaaGilaiaaywW7caaMf8UaaG
zbVlaaywW7caaMf8UaaiikaiaaikdacaGGUaGaaGymaiaaigdacaGG
Paaaaa@5C16@
où
X
*
=
∑
i
∈
U
x
i
*
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiwamaaCa
aaleqabaGaaGOkaaaakiaai2dadaaeqaqabSqaaiaadMgacqGHiiIZ
caWGvbaabeqdcqGHris5aOGaaGPaVlaahIhadaqhaaWcbaGaamyAaa
qaaiaaiQcaaaaaaa@4354@
et
X
^
π
*
=
∑
i
∈
s
d
i
x
i
*
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCiwayaaja
Waa0baaSqaaiabec8aWbqaaiaaiQcaaaGccaaI9aWaaabeaeqaleaa
caWGPbGaeyicI4Saam4Caaqab0GaeyyeIuoakiaaykW7caWGKbWaaS
baaSqaaiaadMgaaeqaaOGaaCiEamaaDaaaleaacaWGPbaabaGaaGOk
aaaakiaac6caaaa@4808@
Le vecteur de régression des coefficients
estimés
B
^
2
,
GREG
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
WaaSbaaSqaaiaaikdacaGGSaGaae4raiaabkfacaqGfbGaae4raaqa
baaaaa@3CD0@
est obtenu à partir de
B
^
GREG
=
(
B
^
1
,
G
R
E
G
,
B
^
2
,
GREG
Τ
)
Τ
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
WaaSbaaSqaaiaabEeacaqGsbGaaeyraiaabEeaaeqaaOGaaGypamaa
bmaabaGabmOqayaajaWaaSbaaSqaaiaaigdacaGGSaGaai4raiaack
facaGGfbGaai4raaqabaGccaaISaGabCOqayaajaWaa0baaSqaaiaa
ikdacaGGSaGaae4raiaabkfacaqGfbGaae4raaqaaiabgs6aubaaaO
GaayjkaiaawMcaamaaCaaaleqabaGaeyiPdqfaaaaa@4D0A@
donné par (2.6). La variance
approximative par rapport au plan de
Y
^
SREG
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEeaaeqaaaaa@3B83@
prend la même forme que l’équation (2.3), où
E
i
=
y
i
−
x
i
*
Τ
B
2
,
GREG
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyramaaBa
aaleaacaWGPbaabeaakiaai2dacaWG5bWaaSbaaSqaaiaadMgaaeqa
aOGaeyOeI0IaaCiEamaaDaaaleaacaWGPbaabaGaaGOkaiabgs6aub
aakiaahkeadaWgaaWcbaGaaGOmaiaacYcacaqGhbGaaeOuaiaabwea
caqGhbaabeaakiaacYcaaaa@479F@
et
B
2,
GREG
=
{
∑
i
∈
U
c
i
(
x
i
*
−
X
¯
N
*
)
(
x
i
*
−
X
¯
N
*
)
Τ
}
−
1
∑
i
∈
U
c
i
(
x
i
*
−
X
¯
N
*
)
y
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOqamaaBa
aaleaacaaIYaGaaGilaiaabEeacaqGsbGaaeyraiaabEeaaeqaaOGa
aGypamaacmaabaWaaabuaeqaleaacaWGPbGaeyicI4Saamyvaaqab0
GaeyyeIuoakiaaykW7caWGJbWaaSbaaSqaaiaadMgaaeqaaOWaaeWa
aeaacaWH4bWaa0baaSqaaiaadMgaaeaacaaIQaaaaOGaeyOeI0IabC
iwayaaraWaa0baaSqaaiaad6eaaeaacaaIQaaaaaGccaGLOaGaayzk
aaWaaeWaaeaacaWH4bWaa0baaSqaaiaadMgaaeaacaaIQaaaaOGaey
OeI0IabCiwayaaraWaa0baaSqaaiaad6eaaeaacaaIQaaaaaGccaGL
OaGaayzkaaWaaWbaaSqabeaacqGHKoavaaaakiaawUhacaGL9baada
ahaaWcbeqaaiabgkHiTiaaigdaaaGcdaaeqbqabSqaaiaadMgacqGH
iiIZcaWGvbaabeqdcqGHris5aOGaaGPaVlaadogadaWgaaWcbaGaam
yAaaqabaGcdaqadaqaaiaahIhadaqhaaWcbaGaamyAaaqaaiaaiQca
aaGccqGHsislceWHybGbaebadaqhaaWcbaGaamOtaaqaaiaaiQcaaa
aakiaawIcacaGLPaaacaWG5bWaaSbaaSqaaiaadMgaaeqaaaaa@6FB3@
et
X
¯
N
*
=
∑
i
∈
U
x
i
*
/
N
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCiwayaara
Waa0baaSqaaiaad6eaaeaacaaIQaaaaOGaaGypamaalyaabaWaaabe
aeqaleaacaWGPbGaeyicI4Saamyvaaqab0GaeyyeIuoakiaaykW7ca
WH4bWaa0baaSqaaiaadMgaaeaacaaIQaaaaaGcbaGaamOtaaaacaGG
Uaaaaa@45E4@
Nous pouvons obtenir les propriétés de
(2.11) en notant que
Y
^
SREG
−
Y
=
Y
^
π
−
Y
+
(
X
*
−
X
^
π
*
)
Τ
B
^
2
,
GREG
=
Y
^
π
−
Y
+
(
X
*
−
X
^
π
*
)
Τ
B
2
,
GREG
+
(
X
*
−
X
^
π
*
)
Τ
(
B
^
2
,
GREG
−
B
2
,
GREG
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaqbaeaabiGaaa
qaaiqadMfagaqcamaaBaaaleaacaqGtbGaaeOuaiaabweacaqGhbaa
keqaaiabgkHiTiaadMfaaeaacaaI9aGabmywayaajaWaaSbaaSqaai
abec8aWbqabaGccqGHsislcaWGzbGaey4kaSYaaeWaaeaacaWHybWa
aWbaaSqabeaacaaIQaaaaOGaeyOeI0IabCiwayaajaWaa0baaSqaai
abec8aWbqaaiaaiQcaaaaakiaawIcacaGLPaaadaahaaWcbeqaaiab
gs6aubaakiqahkeagaqcamaaBaaaleaacaaIYaGaaiilaiaabEeaca
qGsbGaaeyraiaabEeaaOqabaaabaaabaGaaGypaiqadMfagaqcamaa
BaaaleaacqaHapaCaeqaaOGaeyOeI0IaamywaiabgUcaRmaabmaaba
GaaCiwamaaCaaaleqabaGaaGOkaaaakiabgkHiTiqahIfagaqcamaa
DaaaleaacqaHapaCaeaacaaIQaaaaaGccaGLOaGaayzkaaWaaWbaaS
qabeaacqGHKoavaaGccaWHcbWaaSbaaSqaaiaaikdacaGGSaGaae4r
aiaabkfacaqGfbGaae4raaqabaGccqGHRaWkdaqadaqaaiaahIfada
ahaaWcbeqaaiaaiQcaaaGccqGHsislceWHybGbaKaadaqhaaWcbaGa
eqiWdahabaGaaGOkaaaaaOGaayjkaiaawMcaamaaCaaaleqabaGaey
iPdqfaaOWaaeWaaeaaceWHcbGbaKaadaWgaaWcbaGaaGOmaiaacYca
caqGhbGaaeOuaiaabweacaqGhbaakeqaaiabgkHiTiaahkeadaWgaa
WcbaGaaGOmaiaacYcacaqGhbGaaeOuaiaabweacaqGhbaabeaaaOGa
ayjkaiaawMcaaiaai6caaaaaaa@8183@
Étant donné que
B
^
2,
GREG
−
B
2,
GREG
=
O
p
(
n
−
1
/
2
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
WaaSbaaSqaaiaaikdacaaISaGaae4raiaabkfacaqGfbGaae4raaqa
baGccqGHsislcaWHcbWaaSbaaSqaaiaaikdacaaISaGaae4raiaabk
facaqGfbGaae4raaqabaGccaaI9aGaam4tamaaBaaaleaacaWGWbaa
beaakmaabmaabaGaamOBamaaCaaaleqabaGaeyOeI0YaaSGbaeaaca
aIXaaabaGaaGOmaaaaaaaakiaawIcacaGLPaaaaaa@4B64@
sous certaines conditions de régularité
examinées dans Fuller (2009, chapitre 2), le dernier terme est d’ordre
plus faible. Ainsi, en ignorant les termes d’ordre plus faible, nous obtenons
l’approximation
Y
^
SREG
−
Y
≅
∑
i
∈
s
d
i
E
i
−
∑
i
∈
U
E
i
,
(
2.12
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEeaaeqaaOGaeyOeI0Ia
amywaiabgwKianaaqafabeWcbaGaamyAaiabgIGiolaadohaaeqani
abggHiLdGccaaMc8UaamizamaaBaaaleaacaWGPbaabeaakiaadwea
daWgaaWcbaGaamyAaaqabaGccqGHsisldaaeqbqabSqaaiaadMgacq
GHiiIZcaWGvbaabeqdcqGHris5aOGaaGPaVlaadweadaWgaaWcbaGa
amyAaaqabaGccaaISaGaaGzbVlaaywW7caaMf8UaaGzbVlaaywW7ca
GGOaGaaGOmaiaac6cacaaIXaGaaGOmaiaacMcaaaa@6042@
où
E
i
=
y
i
−
x
i
*
Τ
B
2,
GREG
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyramaaBa
aaleaacaWGPbaabeaakiaai2dacaWG5bWaaSbaaSqaaiaadMgaaeqa
aOGaeyOeI0IaaCiEamaaDaaaleaacaWGPbaabaGaaGOkaiabgs6aub
aakiaahkeadaWgaaWcbaGaaGOmaiaaiYcacaqGhbGaaeOuaiaabwea
caqGhbaabeaakiaac6caaaa@47A7@
Par conséquent,
Y
^
SREG
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEeaaeqaaaaa@3B83@
est approximativement sans biais sous le plan.
Nous pouvons calculer la variance asymptotique en utilisant
V
{
∑
i
∈
s
d
i
E
i
−
∑
i
∈
U
E
i
}
=
E
{
(
∑
i
∈
s
d
i
E
i
−
∑
i
∈
U
E
i
)
2
}
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamOvamaacm
aabaWaaabuaeqaleaacaWGPbGaeyicI4Saam4Caaqab0GaeyyeIuoa
kiaaykW7caWGKbWaaSbaaSqaaiaadMgaaeqaaOGaamyramaaBaaale
aacaWGPbaabeaakiabgkHiTmaaqafabeWcbaGaamyAaiabgIGiolaa
dwfaaeqaniabggHiLdGccaaMc8UaamyramaaBaaaleaacaWGPbaabe
aaaOGaay5Eaiaaw2haaiaai2dacaWGfbWaaiWaaeaadaqadaqaamaa
qafabeWcbaGaamyAaiabgIGiolaadohaaeqaniabggHiLdGccaaMc8
UaamizamaaBaaaleaacaWGPbaabeaakiaadweadaWgaaWcbaGaamyA
aaqabaGccqGHsisldaaeqbqabSqaaiaadMgacqGHiiIZcaWGvbaabe
qdcqGHris5aOGaaGPaVlaadweadaWgaaWcbaGaamyAaaqabaaakiaa
wIcacaGLPaaadaahaaWcbeqaaiaaikdaaaaakiaawUhacaGL9baaca
aIUaaaaa@6B29@
Comme nous pouvons le
voir, la variance asymptotique peut être assez importante à moins que
∑
i
∈
U
E
i
=
0.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaWaaabeaeqale
aacaWGPbGaeyicI4Saamyvaaqab0GaeyyeIuoakiaaykW7caWGfbWa
aSbaaSqaaiaadMgaaeqaaOGaaGypaiaaicdacaGGUaaaaa@4212@
Remarque 2.1 Si
y
i
=
a
+
b
x
i
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyEamaaBa
aaleaacaWGPbaabeaakiaai2dacaWGHbGaey4kaSIaamOyaiaadIha
daWgaaWcbaGaamyAaaqabaGccaGGSaaaaa@3F95@
nous avons
Y
^
SREG
−
Y
=
(
N
^
π
−
N
)
a
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEeaaeqaaOGaeyOeI0Ia
amywaiaai2dadaqadaqaaiqad6eagaqcamaaBaaaleaacqaHapaCae
qaaOGaeyOeI0IaamOtaaGaayjkaiaawMcaaiaadggacaGGSaaaaa@45D4@
ce qui implique que
V
(
Y
^
SREG
)
=
a
2
V
(
N
^
π
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamOvamaabm
aabaGabmywayaajaWaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEea
aeqaaaGccaGLOaGaayzkaaGaaGypaiaadggadaahaaWcbeqaaiaaik
daaaGccaWGwbWaaeWaaeaaceWGobGbaKaadaWgaaWcbaGaeqiWdaha
beaaaOGaayjkaiaawMcaaiaac6caaaa@467D@
Cela signifie que si
V
(
N
^
π
)
>
0
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamOvamaabm
aabaGabmOtayaajaWaaSbaaSqaaiabec8aWbqabaaakiaawIcacaGL
PaaacaaI+aGaaGimaiaacYcaaaa@3E98@
nous pouvons accroître artificiellement
a
2
V
(
N
^
π
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyyamaaCa
aaleqabaGaaGOmaaaakiaadAfadaqadaqaaiqad6eagaqcamaaBaaa
leaacqaHapaCaeqaaaGccaGLOaGaayzkaaGaaiilaaaa@3EEF@
la variance de
Y
^
SREG
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEeaaeqaaOGaaiilaaaa
@3C3D@
en choisissant des valeurs élevées de
a
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyyaiaac6
caaaa@38C4@
Il est à noter que l’estimateur par régression optimal obtenu en utilisant
x
*
=
(
x
2
,
…
,
x
p
)
Τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaCa
aaleqabaGaaGOkaaaakiaai2dadaqadaqaaiaadIhadaWgaaWcbaGa
aGOmaaqabaGccaaISaGaeSOjGSKaaGilaiaadIhadaWgaaWcbaGaam
iCaaqabaaakiaawIcacaGLPaaadaahaaWcbeqaaiabgs6aubaaaaa@43C1@
est lui aussi
approximativement sans biais sous le plan, car
Y
^
OPT
*
−
Y
=
Y
^
π
−
Y
+
(
X
*
−
X
^
π
*
)
Τ
B
^
OPT
*
=
Y
^
π
−
Y
+
(
X
*
−
X
^
π
*
)
Τ
B
OPT
*
+
(
X
*
−
X
^
π
*
)
Τ
(
B
^
OPT
*
−
B
OPT
*
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaqbaeaabiGaaa
qaaiqadMfagaqcamaaDaaaleaacaqGpbGaaeiuaiaabsfaaeaacaGG
QaaaaOGaeyOeI0Iaamywaaqaaiaai2daceWGzbGbaKaadaWgaaWcba
GaeqiWdahabeaakiabgkHiTiaadMfacqGHRaWkdaqadaqaaiaahIfa
daahaaWcbeqaaiaaiQcaaaGccqGHsislceWHybGbaKaadaqhaaWcba
GaeqiWdahabaGaaGOkaaaaaOGaayjkaiaawMcaamaaCaaaleqabaGa
eyiPdqfaaOGabCOqayaajaWaa0baaSqaaiaab+eacaqGqbGaaeivaa
qaaiaacQcaaaaakeaaaeaacaaI9aGabmywayaajaWaaSbaaSqaaiab
ec8aWbqabaGccqGHsislcaWGzbGaey4kaSYaaeWaaeaacaWHybWaaW
baaSqabeaacaaIQaaaaOGaeyOeI0IabCiwayaajaWaa0baaSqaaiab
ec8aWbqaaiaaiQcaaaaakiaawIcacaGLPaaadaahaaWcbeqaaiabgs
6aubaakiaahkeadaqhaaWcbaGaae4taiaabcfacaqGubaabaGaaiOk
aaaakiabgUcaRmaabmaabaGaaCiwamaaCaaaleqabaGaaGOkaaaaki
abgkHiTiqahIfagaqcamaaDaaaleaacqaHapaCaeaacaaIQaaaaaGc
caGLOaGaayzkaaWaaWbaaSqabeaacqGHKoavaaGcdaqadaqaaiqahk
eagaqcamaaDaaaleaacaqGpbGaaeiuaiaabsfaaeaacaGGQaaaaOGa
eyOeI0IaaCOqamaaDaaaleaacaqGpbGaaeiuaiaabsfaaeaacaGGQa
aaaaGccaGLOaGaayzkaaGaaGilaaaaaaa@7BA7@
où
B
OPT
*
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCOqamaaDa
aaleaacaqGpbGaaeiuaiaabsfaaeaacaGGQaaaaaaa@3B4E@
est obtenu en remplaçant
x
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaBa
aaleaacaWGPbaabeaaaaa@3947@
par
x
i
*
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaaCiEamaaDa
aaleaacaWGPbaabaGaaGOkaaaaaaa@39FC@
dans l’équation (2.8). Étant donné que
B
^
OPT
*
−
B
OPT
*
=
O
p
(
n
−
1
/
2
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
Waa0baaSqaaiaab+eacaqGqbGaaeivaaqaaiaaiQcaaaGccqGHsisl
caWHcbWaa0baaSqaaiaab+eacaqGqbGaaeivaaqaaiaaiQcaaaGcca
aI9aGaam4tamaaBaaaleaacaWGWbaabeaakmaabmaabaGaamOBamaa
CaaaleqabaGaeyOeI0YaaSGbaeaacaaIXaaabaGaaGOmaaaaaaaaki
aawIcacaGLPaaaaaa@4880@
sous certaines conditions de régularité
examinées dans Fuller (2009, chapitre 2), en ignorant les termes d’ordre
plus faible, nous obtenons
Y
^
OPT
*
−
Y
≅
Y
^
π
−
Y
+
(
X
*
−
X
^
π
*
)
Τ
B
OPT
*
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
Waa0baaSqaaiaab+eacaqGqbGaaeivaaqaaiaacQcaaaGccqGHsisl
caWGzbGaeyyrIaKabmywayaajaWaaSbaaSqaaiabec8aWbqabaGccq
GHsislcaWGzbGaey4kaSYaaeWaaeaacaWHybWaaWbaaSqabeaacaaI
QaaaaOGaeyOeI0IabCiwayaajaWaa0baaSqaaiabec8aWbqaaiaaiQ
caaaaakiaawIcacaGLPaaadaahaaWcbeqaaiabgs6aubaakiaahkea
daqhaaWcbaGaae4taiaabcfacaqGubaabaGaaiOkaaaakiaai6caaa
a@5283@
La variance asymptotique de
Y
^
OPT
*
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
Waa0baaSqaaiaab+eacaqGqbGaaeivaaqaaiaacQcaaaaaaa@3B71@
est plus faible que celle de
Y
^
SREG
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaabofacaqGsbGaaeyraiaabEeaaeqaaOGaaiilaaaa
@3C3D@
car l’estimateur optimal
minimise la variance asymptotique dans la classe d’estimateurs de la forme
Y
^
B
=
Y
^
π
+
(
X
*
−
X
^
π
*
)
Τ
B
^
(
2.13
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabmywayaaja
WaaSbaaSqaaiaadkeaaeqaaOGaaGypaiqadMfagaqcamaaBaaaleaa
cqaHapaCaeqaaOGaey4kaSYaaeWaaeaacaWHybWaaWbaaSqabeaaca
aIQaaaaOGaeyOeI0IabCiwayaajaWaa0baaSqaaiabec8aWbqaaiaa
iQcaaaaakiaawIcacaGLPaaadaahaaWcbeqaaiabgs6aubaakiqahk
eagaqcaiaaywW7caaMf8UaaGzbVlaaywW7caaMf8Uaaiikaiaaikda
caGGUaGaaGymaiaaiodacaGGPaaaaa@5419@
indexée
par
B
^
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9Ff0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGabCOqayaaja
GaaiOlaaaa@38B9@
ISSN : 1712-5685
Politique de rédaction
Techniques d ’enquête publie des articles sur les divers aspects des méthodes statistiques qui intéressent un organisme statistique comme, par exemple, les problèmes de conception découlant de contraintes d’ordre pratique, l’utilisation de différentes sources de données et de méthodes de collecte, les erreurs dans les enquêtes, l’évaluation des enquêtes, la recherche sur les méthodes d’enquête, l’analyse des séries chronologiques, la désaisonnalisation, les études démographiques, l’intégration de données statistiques, les méthodes d’estimation et d’analyse de données et le développement de systèmes généralisés. Une importance particulière est accordée à l’élaboration et à l’évaluation de méthodes qui ont été utilisées pour la collecte de données ou appliquées à des données réelles. Tous les articles seront soumis à une critique, mais les auteurs demeurent responsables du contenu de leur texte et les opinions émises dans la revue ne sont pas nécessairement celles du comité de rédaction ni de Statistique Canada.
Présentation de textes pour la revue
Techniques d ’enquête est publiée en version électronique deux fois l’an. Les auteurs désirant faire paraître un article sont invités à le faire parvenir en français ou en anglais en format électronique et préférablement en Word au rédacteur en chef, (statcan.smj-rte.statcan@canada.ca , Statistique Canada, 150 Promenade du Pré Tunney, Ottawa, (Ontario), Canada, K1A 0T6). Pour les instructions sur le format, veuillez consulter les directives présentées dans la revue ou sur le site web (www.statcan.gc.ca/Techniquesdenquete).
Note de reconnaissance
Le succès du système statistique du Canada repose sur un partenariat bien établi entre Statistique Canada et la population, les entreprises, les administrations canadiennes et les autres organismes. Sans cette collaboration et cette bonne volonté, il serait impossible de produire des statistiques précises et actuelles.
Normes de service à la clientèle
Statistique Canada s'engage à fournir à ses clients des services rapides, fiables et courtois. À cet égard, notre organisme s'est doté de normes de service à la clientèle qui doivent être observées par les employés lorsqu'ils offrent des services à la clientèle.
Droit d'auteur
Publication autorisée par le ministre responsable de Statistique Canada.
© Ministre de l'Industrie, 2016
L'utilisation de la présente publication est assujettie aux modalités de l'Entente de licence ouverte de Statistique Canada .
N° 12-001-X au catalogue
Périodicité : Semi-annuel
Ottawa
Date de modification :
2016-06-22