Estimation sur petits domaines fondée sur un modèle sous échantillonnage informatif
4. Étude en simulationEstimation sur petits domaines fondée sur un modèle sous échantillonnage informatif
4. Étude en simulation
4.1 Exécution
Une approche plan de
sondage-modèle (pm) a été utilisée pour l’étude
en simulation en générant des données pour les
N
=
∑
i
=
1
M
N
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGobGaey
ypa0ZaaabmaeaacaWGobWaaSbaaSqaaiaadMgaaeqaaaqaaiaadMga
cqGH9aqpcaaIXaaabaGaamytaaqdcqGHris5aaaa@41A3@
unités de la population conformément
à un modèle spécifié, puis en sélectionnant un échantillon selon un plan de
sondage spécifié. Le processus de génération des données de la population, puis
de sélection d’un échantillon a été répété
R
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbaaaa@393C@
fois. Nous décrivons
maintenant les étapes de l’exécution du processus. Les données de population,
y
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG5bWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacYcaaaa@3C26@
pour
M
=
99
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbGaey
ypa0JaaGyoaiaaiMdaaaa@3BC3@
domaines et
N
i
=
100
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGobWaaS
baaSqaaiaadMgaaeqaaOGaeyypa0JaaGymaiaaicdacaaIWaaaaa@3D91@
unités dans chaque domaine
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbaaaa@3953@
ont été générées au moyen du
simple modèle de régression linéaire à erreurs emboîtées
y
i j
=
β
0
+
β
1
x
i j
+
ν
i
+
e
i j
; i = 1 , … , 99 ; j = 1 , … , 100 , ( 4.1 )
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG5bWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9iabek7aInaaBaaaleaa
caaIWaaabeaakiabgUcaRiabek7aInaaBaaaleaacaaIXaaabeaaki
aadIhadaWgaaWcbaGaamyAaiaadQgaaeqaaOGaey4kaSIaeqyVd42a
aSbaaSqaaiaadMgaaeqaaOGaey4kaSIaamyzamaaBaaaleaacaWGPb
GaamOAaaqabaGccaGG7aGaaeiiaiaadMgacqGH9aqpcaaIXaGaaiil
aiablAciljaacYcacaaI5aGaaGyoaiaacUdacaqGGaGaamOAaiabg2
da9iaaigdacaGGSaGaeSOjGSKaaiilaiaaigdacaaIWaGaaGimaiaa
cYcacaaMf8UaaGzbVlaaywW7caaMf8UaaGzbVlaacIcacaaI0aGaai
OlaiaaigdacaGGPaaaaa@6A03@
où
β
0
=
1
,
β
1
=
1
,
v
i
∼
iid
N
(
0
,
σ
v
2
=
0,5
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHYoGyda
WgaaWcbaGaaGimaaqabaGccqGH9aqpcaaIXaGaaiilaiabek7aInaa
BaaaleaacaaIXaaabeaakiabg2da9iaaigdacaGGSaGaamODamaaBa
aaleaacaWGPbaabeaakmaaxacabaGaeSipIOdaleqabaGaaeyAaiaa
bMgacaqGKbaaaOGaamOtamaabmaabaGaaGimaiaacYcacqaHdpWCda
qhaaWcbaGaamODaaqaaiaaikdaaaGccqGH9aqpcaqGWaGaaeilaiaa
bwdaaiaawIcacaGLPaaaaaa@535B@
et indépendants des
e
i
j
∼
iid
N
(
0
,
σ
e
2
=
2
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGLbWaaS
baaSqaaiaadMgacaWGQbaabeaakmaaxacabaGaeSipIOdaleqabaGa
aeyAaiaabMgacaqGKbaaaOGaamOtamaabmaabaGaaGimaiaacYcacq
aHdpWCdaqhaaWcbaGaamyzaaqaaiaaikdaaaGccqGH9aqpcaaIYaaa
caGLOaGaayzkaaGaaiOlaaaa@4977@
Les valeurs de
x
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9pC0xbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamiEamaaBa
aaleaacaWGPbGaamOAaaqabaaaaa@3B5B@
dans la population
ont été générées à partir d’une loi gamma de moyenne 10 et de variance 50, et maintenues
fixes pendant la simulation des valeurs de
y
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9pC0xbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyEamaaBa
aaleaacaWGPbGaamOAaaqabaaaaa@3B5C@
de la population
au moyen de (4.1).
Nous avons considéré différentes tailles d’échantillon dans les domaines
en fixant
n
i
=
5
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaeyypa0JaaGynaaaa@3C41@
pour les 33 premiers domaines,
n
i
=
7
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaeyypa0JaaG4naaaa@3C43@
pour les 33 domaines suivants
et
n
i
=
9
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaeyypa0JaaGyoaaaa@3C45@
pour les 33 derniers
domaines. L’objectif était d’étudier l’effet de tailles d’échantillon inégales
sur le choix de la variable d’augmentation
g
i
j
=
g
(
p
j
|
i
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9iaadEgadaqadaqaaiaa
dchadaWgaaWcbaWaaqGaaeaacaWGQbaacaGLiWoacaWGPbaabeaaaO
GaayjkaiaawMcaaiaac6caaaa@442F@
Des échantillons de tailles
spécifiées,
n
i
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaaiilaaaa@3B2C@
ont été sélectionnés dans les
domaines avec probabilités proportionnelles aux tailles spécifiées,
b
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGIbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacYcaaaa@3C0F@
en utilisant la méthode d’échantillonnage de Rao-Sampford (Rao 1965 et Sampford 1967) avec
probabilités inégales et sans remise. Cette dernière méthode fait en sorte que les
probabilités d’inclusion
π
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
WgaaWcbaWaaqGaaeaacaWGQbaacaGLiWoacaWGPbaabeaaaaa@3DC1@
soient proportionnelles aux
tailles
b
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGIbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacYcaaaa@3C0F@
c’est-à-dire,
π
j
|
i
=
n
i
b
i
j
/
B
i
=
n
i
p
j
|
i
,
j
=
1
,
…
,
N
i
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
WgaaWcbaWaaqGaaeaacaWGQbaacaGLiWoacaWGPbaabeaakiabg2da
9maalyaabaGaamOBamaaBaaaleaacaWGPbaabeaakiaadkgadaWgaa
WcbaGaamyAaiaadQgaaeqaaaGcbaGaamOqamaaBaaaleaacaWGPbaa
beaakiabg2da9iaad6gadaWgaaWcbaGaamyAaaqabaGccaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaOGaaiil
aiaadQgacqGH9aqpcaaIXaGaaiilaiablAciljaacYcacaWGobWaaS
baaSqaaiaadMgaaeqaaOGaaiilaaaa@5627@
où
B
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGcbWaaS
baaSqaaiaadMgaaeqaaaaa@3A46@
est le total des
b
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGIbWaaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3B55@
dans le domaine
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbGaai
Olaaaa@3A05@
Nous avons considéré deux choix distincts des tailles
b
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGIbWaaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3B55@
dans l’étude en simulation. Comme
premier choix, nous avons utilisé
b
i
j
=
exp
[
{
−
(
y
i
j
−
β
0
−
β
1
x
i
j
)
/
σ
e
+
δ
i
j
/
5
}
/
3
]
=
exp
[
{
−
(
v
i
+
e
i
j
)
/
σ
e
+
δ
i
j
/
5
}
/
3
]
,
(
4.2
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaafaqaaeOada
aabaGaamOyamaaBaaaleaacaWGPbGaamOAaaqabaaakeaacqGH9aqp
aeaaciGGLbGaaiiEaiaacchadaWadeqaamaalyaabaWaaiWaaeaacq
GHsisldaWcgaqaamaabmqabaGaamyEamaaBaaaleaacaWGPbGaamOA
aaqabaGccqGHsislcqaHYoGydaWgaaWcbaGaaGimaaqabaGccqGHsi
slcqaHYoGydaWgaaWcbaGaaGymaaqabaGccaWG4bWaaSbaaSqaaiaa
dMgacaWGQbaabeaaaOGaayjkaiaawMcaaaqaaiabeo8aZnaaBaaale
aacaWGLbaabeaaaaGccqGHRaWkdaWcgaqaaiabes7aKnaaBaaaleaa
caWGPbGaamOAaaqabaaakeaacaaI1aaaaaGaay5Eaiaaw2haaaqaai
aaiodaaaaacaGLBbGaayzxaaaabaaabaGaeyypa0dabaGaciyzaiaa
cIhacaGGWbWaamWabeaadaWcgaqaamaacmaabaGaeyOeI0YaaSGbae
aadaqadeqaaiaadAhadaWgaaWcbaGaamyAaaqabaGccqGHRaWkcaWG
LbWaaSbaaSqaaiaadMgacaWGQbaabeaaaOGaayjkaiaawMcaaaqaai
abeo8aZnaaBaaaleaacaWGLbaabeaaaaGccqGHRaWkdaWcgaqaaiab
es7aKnaaBaaaleaacaWGPbGaamOAaaqabaaakeaacaaI1aaaaaGaay
5Eaiaaw2haaaqaaiaaiodaaaaacaGLBbGaayzxaaGaaiilaaaacaaM
f8UaaGzbVlaaywW7caaMf8UaaGzbVlaaywW7caaMf8UaaGzbVlaacI
cacaaI0aGaaiOlaiaaikdacaGGPaaaaa@871F@
où
δ
i
j
∼
iid
N
(
0,1
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaH0oazda
WgaaWcbaGaamyAaiaadQgaaeqaaOWaaCbiaeaacqWI8iIoaSqabeaa
caqGPbGaaeyAaiaabsgaaaGccaWGobWaaeWaaeaacaqGWaGaaeilai
aabgdaaiaawIcacaGLPaaacaGGUaaaaa@457C@
Les mesures de
taille (4.2) sont équivalentes à celles utilisées par Pfeffermann et Sverchkov (2007)
dans leur étude en simulation et satisfont la relation (1.2) sur les poids
w
j
|
i
=
π
j
|
i
−
1
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaGccqGH9aqp
cqaHapaCdaqhaaWcbaWaaqGaaeaacaWGQbaacaGLiWoacaWGPbaaba
GaeyOeI0IaaGymaaaakiaac6caaaa@45D1@
À l’instar de l’approche
PS, nous avons tronqué les effets de domaine
v
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG2bWaaS
baaSqaaiaadMgaaeqaaaaa@3A7A@
et les
erreurs au niveau de l’unité
e
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGLbWaaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3B58@
à
±
2,5
σ
v
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqGHXcqSca
qGYaGaaeilaiaabwdacqaHdpWCdaWgaaWcbaGaamODaaqabaaaaa@3F59@
et
±
2,5
σ
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqGHXcqSca
qGYaGaaeilaiaabwdacqaHdpWCdaWgaaWcbaGaamyzaaqabaaaaa@3F48@
pour éviter des
probabilités de sélection extrêmes.
Le deuxième choix de mesures de taille, à l’exemple d’Asparouhov (2006), comprend deux types distincts
de mesure de taille, à savoir une mesure invariante (I) et une mesure non invariante
(NI). Dans le cas invariant,
b
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGIbWaaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3B55@
est indépendante de
v
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG2bWaaS
baaSqaaiaadMgaaeqaaaaa@3A7A@
sachant
x
i
j
;
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4bWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacUdaaaa@3C38@
sinon, la mesure est dite non
invariante. Les mesures de taille invariantes sont données par
b
i
j
=
[
1
+
exp
{
−
τ
(
1
α
e
i
j
+
1
−
1
α
2
e
i
j
*
)
}
]
−
1
.
(
4.3
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGIbWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9maadmaabaGaaGymaiab
gUcaRiGacwgacaGG4bGaaiiCamaacmaabaGaeyOeI0IaeqiXdq3aae
WaaeaadaWcaaqaaiaaigdaaeaacqaHXoqyaaGaamyzamaaBaaaleaa
caWGPbGaamOAaaqabaGccqGHRaWkdaGcaaqaaiaaigdacqGHsislda
WcaaqaaiaaigdaaeaacqaHXoqydaahaaWcbeqaaiaaikdaaaaaaaqa
baGccaaMe8UaamyzamaaDaaaleaacaWGPbGaamOAaaqaaiaacQcaaa
aakiaawIcacaGLPaaaaiaawUhacaGL9baaaiaawUfacaGLDbaadaah
aaWcbeqaaiabgkHiTiaaigdaaaGccaGGUaGaaGzbVlaaywW7caaMf8
UaaGzbVlaaywW7caGGOaGaaGinaiaac6cacaaIZaGaaiykaaaa@67AE@
Les
mesures de taille non invariantes sont telles que
b
i
j
=
[
1
+
exp
{
−
τ
(
1
α
(
v
i
+
e
i
j
)
+
1
−
1
α
2
(
v
i
*
+
e
i
j
*
)
)
}
]
−
1
.
(
4.4
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGIbWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9maadmaabaGaaGymaiab
gUcaRiGacwgacaGG4bGaaiiCamaacmaabaGaeyOeI0IaeqiXdq3aae
WaaeaadaWcaaqaaiaaigdaaeaacqaHXoqyaaWaaeWaaeaacaWG2bWa
aSbaaSqaaiaadMgaaeqaaOGaey4kaSIaamyzamaaBaaaleaacaWGPb
GaamOAaaqabaaakiaawIcacaGLPaaacqGHRaWkdaGcaaqaaiaaigda
cqGHsisldaWcaaqaaiaaigdaaeaacqaHXoqydaahaaWcbeqaaiaaik
daaaaaaaqabaGcdaqadaqaaiaadAhadaqhaaWcbaGaamyAaaqaaiaa
cQcaaaGccqGHRaWkcaWGLbWaa0baaSqaaiaadMgacaWGQbaabaGaai
OkaaaaaOGaayjkaiaawMcaaaGaayjkaiaawMcaaaGaay5Eaiaaw2ha
aaGaay5waiaaw2faamaaCaaaleqabaGaeyOeI0IaaGymaaaakiaac6
cacaaMf8UaaGzbVlaaywW7caaMf8UaaGzbVlaacIcacaaI0aGaaiOl
aiaaisdacaGGPaaaaa@6FE5@
Le coefficient
τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDaa
a@3A2A@
dans (4.3) et
(4.4), qui est choisi égal à 0,5, fait en sorte que la variation des poids
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
ne soit pas trop
grande dans une exécution de la simulation. La paire aléatoire
(
v
i
*
,
e
i
j
*
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aadAhadaqhaaWcbaGaamyAaaqaaiaacQcaaaGccaGGSaGaamyzamaa
DaaaleaacaWGPbGaamOAaaqaaiaacQcaaaaakiaawIcacaGLPaaaaa
a@4118@
a été générée indépendamment
de
(
v
i
,
e
i
j
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aadAhadaWgaaWcbaGaamyAaaqabaGccaGGSaGaamyzamaaBaaaleaa
caWGPbGaamOAaaqabaaakiaawIcacaGLPaaaaaa@3FBA@
à partir des
mêmes lois que celle de
v
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG2bWaaS
baaSqaaiaadMgaaeqaaaaa@3A7A@
et
e
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGLbWaaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3B58@
pour s’assurer
que la variation des poids soit comparable entre les divers niveaux de
α
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyca
GGUaaaaa@3AB6@
Les probabilités
π
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
WgaaWcbaGaamOAamaaeeaabaGaamyAaaGaay5bSdaabeaaaaa@3DBF@
qui étaient
supérieures à un ont été fixées à un, et les probabilités des unités restantes
ont été recalculées. Les valeurs de
α
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9pC0xbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaeqySdegaaa@39F4@
dans (4.3) et (4.4), choisies
égales à 1, 2, 3 ou
∞
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqGHEisPca
GGSaaaaa@3A86@
contrôlent
le niveau d’informativité (caractère informatif). L’accroissement
de la valeur de
α
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyaa
a@3A04@
réduit l’informativité,
α
=
∞
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqycq
GH9aqpcqGHEisPaaa@3C7B@
correspondant à un échantillonnage non
informatif. Diverses dépendances ont été introduites dans les simulations comme
il suit, afin d’accroître la précision des comparaisons entre les différents
estimateurs : les quatre composantes de l’erreur
(
v
i
,
e
i
j
,
v
i
*
,
e
i
j
*
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aadAhadaWgaaWcbaGaamyAaaqabaGccaGGSaGaamyzamaaBaaaleaa
caWGPbGaamOAaaqabaGccaGGSaGaamODamaaDaaaleaacaWGPbaaba
GaaiOkaaaakiaacYcacaWGLbWaa0baaSqaaiaadMgacaWGQbaabaGa
aiOkaaaaaOGaayjkaiaawMcaaaaa@4794@
ont toutes été générées
pour commencer. Les valeurs de
y
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9pC0xbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaamyEaaaa@3953@
dans la population,
ainsi que les probabilités de sélection invariantes et non invariantes, ont
ensuite été générées à partir de ces erreurs. Pour une population générée donnée,
huit échantillons ont été sélectionnés : un échantillon invariant et un
échantillon non invariant pour chaque valeur de
α
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyaa
a@3A04@
prise en
considération.
Il convient de souligner que les poids
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
obtenus au moyen des mesures
de taille (4.3) et (4.4) peuvent ne pas satisfaire la condition (1.2) de l’approche
PS. Nous avons néanmoins ajusté (1.2) à ces poids pour calculer
b
^
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGIbGbaK
aacaGGSaaaaa@3A0C@
qui est nécessaire afin de calculer
l’estimateur corrigé du biais
Y
¯
^
i
PS
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaabcfacaqGtbaaaOGaaiOlaaaa
@3CEA@
En utilisant l’approche plan
de sondage-modèle (pm), nous avons généré
R
=
1
000
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbGaey
ypa0JaaeymaiaaykW7caqGWaGaaeimaiaabcdaaaa@3E9A@
échantillons sous les mesures de taille (4.2)
et sous les mesures de taille (4.3) et (4.4). Pour chaque échantillon simulé
r
(
r
=
1
,
…
,
R
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGYbWaae
WaaeaacaWGYbGaeyypa0JaaGymaiaacYcacqWIMaYscaGGSaGaamOu
aaGaayjkaiaawMcaaiaacYcaaaa@41A6@
nous avons calculé les
estimations
Y
¯
^
i
H
(
r
)
,
Y
¯
^
i
(
a
)
H
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaadIeadaqadaqaaiaadkhaaiaa
wIcacaGLPaaaaaGccaGGSaGabmywayaaryaajaWaa0baaSqaaiaadM
gadaqadaqaaiaadggaaiaawIcacaGLPaaaaeaacaWGibWaaeWaaeaa
caWGYbaacaGLOaGaayzkaaaaaaaa@4668@
et
Y
¯
^
i
PS
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaabcfacaqGtbWaaeWaaeaacaWG
YbaacaGLOaGaayzkaaaaaaaa@3EAE@
pour chaque petit domaine
i
;
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbGaai
4oaaaa@3A12@
pour la méthode YR, nous
avons calculé
μ
^
i
YR
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaaeywaiaabkfadaqadaqaaiaadkha
aiaawIcacaGLPaaaaaaaaa@3F77@
et
μ
^
i
(
a
)
YR
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaaeywaiaabkfadaqadaqaaiaadkhaaiaawIcacaGLPaaaaaaaaa@41E6@
seulement. En outre, nous avons calculé les estimations de l’EQM ,
eqm
(
μ
^
i
H
)
(
r
)
,
eqm
(
μ
^
i
(
a
)
H
)
(
r
)
,
eqm
(
μ
^
i
YR
)
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgaaeaa
caWGibaaaaGccaGLOaGaayzkaaWaaWbaaSqabeaadaqadaqaaiaadk
haaiaawIcacaGLPaaaaaGccaGGSaGaaeyzaiaabghacaqGTbWaaeWa
aeaacuaH8oqBgaqcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaaca
GLOaGaayzkaaaabaGaamisaaaaaOGaayjkaiaawMcaamaaCaaaleqa
baWaaeWaaeaacaWGYbaacaGLOaGaayzkaaaaaOGaaiilaiaabwgaca
qGXbGaaeyBamaabmaabaGafqiVd0MbaKaadaqhaaWcbaGaamyAaaqa
aiaabMfacaqGsbaaaaGccaGLOaGaayzkaaWaaWbaaSqabeaadaqada
qaaiaadkhaaiaawIcacaGLPaaaaaaaaa@5D5A@
et
eqm
(
μ
^
i
(
a
)
YR
)
(
r
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgadaqa
daqaaiaadggaaiaawIcacaGLPaaaaeaacaqGzbGaaeOuaaaaaOGaay
jkaiaawMcaamaaCaaaleqabaWaaeWaaeaacaWGYbaacaGLOaGaayzk
aaaaaOGaaiilaaaa@472C@
associées à
μ
^
i
H
,
μ
^
i
(
a
)
H
,
μ
^
i
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaamisaaaakiaacYcacuaH8oqBgaqc
amaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaaba
GaamisaaaakiaacYcacuaH8oqBgaqcamaaDaaaleaacaWGPbaabaGa
aeywaiaabkfaaaaaaa@4836@
et
μ
^
i
(
a
)
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaaeywaiaabkfaaaaaaa@3F66@
. Comme nous l’avons mentionné
plus haut, nous n’avons pas inclus l’estimateur bootstrap de l’EQM de
Y
¯
^
i
PS
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaabcfacaqGtbaaaOGaaiilaaaa
@3CE8@
proposé par Pfeffermann et Sverchkov (2007), dans
l’étude en simulation. En outre, pour simplifier, nous n’avons pas inclus les
estimateurs de l’EQM de
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaadIeaaaaaaa@3B52@
et
Y
¯
^
i
(
a
)
H
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAamaabmaabaGaamyyaaGaayjkaiaawMca
aaqaaiaadIeaaaGccaGGSaaaaa@3E7B@
parce que ces estimateurs donnent des
résultats comparables à
μ
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaamisaaaaaaa@3C13@
et
μ
^
i
(
a
)
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaamisaaaaaaa@3E82@
en ce qui concerne l’EQM .
Nous avons considéré les mesures de performance qui suivent pour un estimateur
donné, disons de la moyenne de petit domaine
Y
¯
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
badaWgaaWcbaGaamyAaaqabaGccaGGUaaaaa@3B31@
Le biais absolu moyen
(
BA
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOqaiaabgeaaaaacaGLOaGaayzkaaaaaa@3B88@
est mesuré par
BA
¯
=
1
M
∑
i
=
1
M
BA
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaiabg2da9maalaaabaGaaGymaaqaaiaad2eaaaWa
aabCaeaacaqGcbGaaeyqamaaBaaaleaacaWGPbaabeaaaeaacaWGPb
Gaeyypa0JaaGymaaqaaiaad2eaa0GaeyyeIuoaaaa@44FC@
avec
BA
i
=
|
1
R
∑
r
=
1
R
(
Y
¯
^
i
(
r
)
−
Y
¯
i
(
r
)
)
|
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGcbGaae
yqamaaBaaaleaacaWGPbaabeaakiabg2da9maaemaabaGaaGPaVpaa
laaabaGaaGymaaqaaiaadkfaaaWaaabCaeaadaqadaqaaiqadMfaga
qegaqcamaaDaaaleaacaWGPbaabaWaaeWaaeaacaWGYbaacaGLOaGa
ayzkaaaaaOGaeyOeI0IabmywayaaraWaa0baaSqaaiaadMgaaeaada
qadaqaaiaadkhaaiaawIcacaGLPaaaaaaakiaawIcacaGLPaaacaaM
c8oaleaacaWGYbGaeyypa0JaaGymaaqaaiaadkfaa0GaeyyeIuoaaO
Gaay5bSlaawIa7aaaa@5587@
où
Y
¯
^
i
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaamaabmaabaGaamOCaaGaayjkaiaa
wMcaaaaaaaa@3D05@
et
Y
¯
i
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
badaqhaaWcbaGaamyAaaqaamaabmaabaGaamOCaaGaayjkaiaawMca
aaaaaaa@3CF6@
sont les valeurs
de
Y
¯
^
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaWgaaWcbaGaamyAaaqabaaaaa@3A84@
et
Y
¯
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
badaWgaaWcbaGaamyAaaqabaaaaa@3A75@
pour les
r
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGYbWaaW
baaSqabeaacaqGLbaaaaaa@3A71@
échantillon et population simulés. L’efficacité d’un
estimateur
Y
¯
^
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaWgaaWcbaGaamyAaaqabaaaaa@3A84@
est mesurée par la racine carrée moyenne de l’EQM
REQM
¯
=
1
M
∑
i
=
1
M
1
R
∑
r
=
1
R
(
Y
¯
^
i
(
r
)
−
Y
¯
i
(
r
)
)
2
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaGaeyypa0ZaaSaaaeaacaaIXaaa
baGaamytaaaadaaeWbqaamaakaaabaWaaSaaaeaacaaIXaaabaGaam
OuaaaadaaeWbqaamaabmaabaGabmywayaaryaajaWaa0baaSqaaiaa
dMgaaeaadaqadaqaaiaadkhaaiaawIcacaGLPaaaaaGccqGHsislce
WGzbGbaebadaqhaaWcbaGaamyAaaqaamaabmaabaGaamOCaaGaayjk
aiaawMcaaaaaaOGaayjkaiaawMcaamaaCaaaleqabaGaaGOmaaaaae
aacaWGYbGaeyypa0JaaGymaaqaaiaadkfaa0GaeyyeIuoaaSqabaaa
baGaamyAaiabg2da9iaaigdaaeaacaWGnbaaniabggHiLdGccaGGUa
aaaa@58F4@
En ce qui concerne la performance des estimateurs de l’EQM ,
eqm
(
μ
^
i
H
)
,
eqm
(
μ
^
i
(
a
)
H
)
,
eqm
(
μ
^
i
YR
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgaaeaa
caWGibaaaaGccaGLOaGaayzkaaGaaiilaiaabwgacaqGXbGaaeyBam
aabmaabaGafqiVd0MbaKaadaqhaaWcbaGaamyAamaabmaabaGaamyy
aaGaayjkaiaawMcaaaqaaiaadIeaaaaakiaawIcacaGLPaaacaGGSa
GaaeyzaiaabghacaqGTbWaaeWaaeaacuaH8oqBgaqcamaaDaaaleaa
caWGPbaabaGaaeywaiaabkfaaaaakiaawIcacaGLPaaaaaa@553F@
et
eqm
(
μ
^
i
(
a
)
YR
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgadaqa
daqaaiaadggaaiaawIcacaGLPaaaaeaacaqGzbGaaeOuaaaaaOGaay
jkaiaawMcaaiaacYcaaaa@4475@
nous avons d’abord calculé
des mesures fiables des EQM en passant de
R
=
1
000
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbGaey
ypa0JaaeymaiaaykW7caqGWaGaaeimaiaabcdaaaa@3E9A@
à
T
=
10
000
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGubGaey
ypa0JaaeymaiaabcdacaaMc8UaaeimaiaabcdacaqGWaaaaa@3F4F@
échantillons simulés. L’EQM d’un estimateur
μ
^
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaBaaaleaacaWGPbaabeaaaaa@3B45@
est alors calculée comme il
suit
EQM
(
μ
^
i
)
=
1
T
∑
t
=
1
T
(
μ
^
i
(
t
)
−
Y
¯
i
(
t
)
)
2
,
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGfbGaae
yuaiaab2eadaqadaqaaiqbeY7aTzaajaWaaSbaaSqaaiaadMgaaeqa
aaGccaGLOaGaayzkaaGaeyypa0ZaaSaaaeaacaaIXaaabaGaamivaa
aadaaeWbqaamaabmaabaGafqiVd0MbaKaadaqhaaWcbaGaamyAaaqa
amaabmaabaGaamiDaaGaayjkaiaawMcaaaaakiabgkHiTiqadMfaga
qeamaaDaaaleaacaWGPbaabaWaaeWaaeaacaWG0baacaGLOaGaayzk
aaaaaaGccaGLOaGaayzkaaWaaWbaaSqabeaacaaIYaaaaaqaaiaads
hacqGH9aqpcaaIXaaabaGaamivaaqdcqGHris5aOGaaiilaaaa@55DA@
où
μ
^
i
(
t
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaWaaeWaaeaacaWG0baacaGLOaGaayzk
aaaaaaaa@3DC8@
Y
¯
i
(
t
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
badaqhaaWcbaGaamyAaaqaamaabmaabaGaamiDaaGaayjkaiaawMca
aaaaaaa@3CF8@
désignent les
valeurs de
μ
^
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaBaaaleaacaWGPbaabeaaaaa@3B45@
et
Y
¯
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
badaWgaaWcbaGaamyAaaqabaaaaa@3A75@
pour les
t
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG0bWaaW
baaSqabeaacaqGLbaaaaaa@3A73@
échantillon et population simulés. Pour l’estimation de
l’EQM , nous avons gardé les
R
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbaaaa@393C@
échantillons simulés originaux et calculé les
valeurs prévues
E
[
eqm
(
μ
^
i
)
]
=
R
−
1
∑
r
=
1
R
eqm
(
μ
^
i
)
(
r
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGfbWaam
WaaeaacaqGLbGaaeyCaiaab2gadaqadaqaaiqbeY7aTzaajaWaaSba
aSqaaiaadMgaaeqaaaGccaGLOaGaayzkaaaacaGLBbGaayzxaaGaey
ypa0JaamOuamaaCaaaleqabaGaeyOeI0IaaGymaaaakmaaqadabaGa
aeyzaiaabghacaqGTbWaaeWaaeaacuaH8oqBgaqcamaaBaaaleaaca
WGPbaabeaaaOGaayjkaiaawMcaamaaCaaaleqabaWaaeWaaeaacaWG
YbaacaGLOaGaayzkaaaaaaqaaiaadkhacqGH9aqpcaaIXaaabaGaam
OuaaqdcqGHris5aOGaaiilaaaa@5648@
où
eqm
(
μ
^
i
)
(
r
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaaSbaaSqaaiaadMgaaeqa
aaGccaGLOaGaayzkaaWaaWbaaSqabeaadaqadaqaaiaadkhaaiaawI
cacaGLPaaaaaaaaa@4251@
désigne la
valeur de l’estimation de l’EQM pour le
r
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGYbWaaW
baaSqabeaacaqGLbaaaaaa@3A71@
échantillon simulé. Le biais relatif absolu moyen
(
BRA
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOqaiaabkfacaqGbbaaaaGaayjkaiaawMcaaaaa@3C5D@
d’un estimateur
de l’EQM
eqm
(
μ
^
i
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaaSbaaSqaaiaadMgaaeqa
aaGccaGLOaGaayzkaaaaaa@3FA4@
est alors
calculé selon
BRA
¯
[
eqm
(
μ
^
i
)
]
=
M
−
1
∑
i
=
1
M
|
E
[
eqm
(
μ
^
i
)
]
EQM
(
μ
^
i
)
−
1
|
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaadaWadaqaaiaabwgacaqGXbGaaeyBamaa
bmaabaGafqiVd0MbaKaadaWgaaWcbaGaamyAaaqabaaakiaawIcaca
GLPaaaaiaawUfacaGLDbaacqGH9aqpcaWGnbWaaWbaaSqabeaacqGH
sislcaaIXaaaaOWaaabCaeaadaabdaqaaiaaykW7daWcaaqaaiaadw
eadaWadaqaaiaabwgacaqGXbGaaeyBamaabmaabaGafqiVd0MbaKaa
daWgaaWcbaGaamyAaaqabaaakiaawIcacaGLPaaaaiaawUfacaGLDb
aaaeaacaqGfbGaaeyuaiaab2eadaqadaqaaiqbeY7aTzaajaWaaSba
aSqaaiaadMgaaeqaaaGccaGLOaGaayzkaaaaaiabgkHiTiaaigdaca
aMc8oacaGLhWUaayjcSdaaleaacaWGPbGaeyypa0JaaGymaaqaaiaa
d2eaa0GaeyyeIuoakiaac6caaaa@6704@
4.2
Résultats sous les mesures de taille de Pfeffermann et Sverchkov
Le tableau 4.1 donne les résultats des simulations concernant
le biais absolu moyen
(
BA
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOqaiaabgeaaaaacaGLOaGaayzkaaaaaa@3B88@
et la racine carrée moyenne de l’erreur quadratique moyenne
(
REQM
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOuaiaabweacaqGrbGaaeytaaaaaiaawIcacaGLPaaa
aaa@3D40@
des estimateurs
Y
¯
^
i
H
,
Y
¯
^
i
(
a
)
H
,
μ
^
i
YR
,
μ
^
i
(
a
)
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaadIeaaaGccaGGSaGabmywayaa
ryaajaWaa0baaSqaaiaadMgadaqadaqaaiaadggaaiaawIcacaGLPa
aaaeaacaWGibaaaOGaaiilaiqbeY7aTzaajaWaa0baaSqaaiaadMga
aeaacaqGzbGaaeOuaaaakiaacYcacuaH8oqBgaqcamaaDaaaleaaca
WGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaabaGaaeywaiaabkfa
aaaaaa@4E6F@
et
Y
¯
^
i
PS
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaabcfacaqGtbaaaaaa@3C2E@
sous les mesures de taille
(4.2) de l’approche PS. Le biais relatif absolu moyen
(
BRA
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOqaiaabkfacaqGbbaaaaGaayjkaiaawMcaaaaa@3C5D@
des estimateurs de l’EQM ,
eqm
(
μ
^
i
H
)
,
eqm
(
μ
^
i
(
a
)
H
)
,
eqm
(
μ
^
i
YR
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgaaeaa
caWGibaaaaGccaGLOaGaayzkaaGaaiilaiaabwgacaqGXbGaaeyBam
aabmaabaGafqiVd0MbaKaadaqhaaWcbaGaamyAamaabmaabaGaamyy
aaGaayjkaiaawMcaaaqaaiaadIeaaaaakiaawIcacaGLPaaacaGGSa
GaaeyzaiaabghacaqGTbWaaeWaaeaacuaH8oqBgaqcamaaDaaaleaa
caWGPbaabaGaaeywaiaabkfaaaaakiaawIcacaGLPaaaaaa@553F@
et
eqm
(
μ
^
i
(
a
)
YR
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgadaqa
daqaaiaadggaaiaawIcacaGLPaaaaeaacaqGzbGaaeOuaaaaaOGaay
jkaiaawMcaaiaacYcaaaa@4475@
est également présenté. Quatre
choix différents de la variable d’augmentation
g
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3B5A@
ont été étudiés :
p
j
|
i
,
w
j
|
i
,
n
i
w
j
|
i
=
p
j
|
i
−
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaGccaGGSaGa
am4DamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaO
Gaaiilaiaad6gadaWgaaWcbaGaamyAaaqabaGccaWG3bWaaSbaaSqa
amaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaGccqGH9aqpcaWGWb
Waa0baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqaaiabgkHi
Tiaaigdaaaaaaa@5107@
et
log
p
j
|
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaciGGSbGaai
4BaiaacEgacaWGWbWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGa
amyAaaqabaGccaGGUaaaaa@4085@
L’estimateur bootstrap de
EQM
(
Y
¯
^
i
PS
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGfbGaae
yuaiaab2eadaqadaqaaiqadMfagaqegaqcamaaDaaaleaacaWGPbaa
baGaaeiuaiaabofaaaaakiaawIcacaGLPaaaaaa@402D@
proposé par Pfeffermann et Sverchkov (2007) n’est
pas inclus dans notre étude, parce que la simulation bootstrap est très
gourmande en ressources informatiques.
Le tableau 4.1 montre que le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
de l’estimateur EBLUP
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaadIeaaaaaaa@3B52@
est grand (= 0,456) comparativement
à l’EBLUP avec modèle augmenté correspondant,
Y
¯
^
i
(
a
)
H
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAamaabmaabaGaamyyaaGaayjkaiaawMca
aaqaaiaadIeaaaGccaGGSaaaaa@3E7B@
pour les quatre choix de
g
i
j
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaac6caaaa@3C16@
En outre, le choix
g
i
j
=
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9iaadEhadaWgaaWcbaWa
aqGaaeaacaWGQbaacaGLiWoacaWGPbaabeaaaaa@4105@
donne un
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
plus grand que les trois autres
choix (0,131 comparativement à 0,042 ou moins). Le pseudo-EBLUP habituel,
μ
^
i
YR
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaaeywaiaabkfaaaGccaGGSaaaaa@3DB1@
donne des résultats
étonnamment bons
(
BA
¯
=
0,044
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOqaiaabgeaaaGaeyypa0JaaeimaiaabYcacaqGWaGa
aeinaiaabsdaaiaawIcacaGLPaaacaGGSaaaaa@40C1@
même s’il a été obtenu sous
l’hypothèse d’un échantillonnage non informatif. Cette bonne performance est
peut-être due à l’utilisation de poids dans
μ
^
i
YR
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaaeywaiaabkfaaaGccaGGUaaaaa@3DB3@
Le pseudo-EBLUP augmenté,
μ
^
i
(
a
)
YR
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaaeywaiaabkfaaaGccaGGSaaaaa@4020@
entraîne une réduction supplémentaire
du
BA
¯
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaiaac6caaaa@3AB1@
L’estimateur PS,
Y
¯
^
i
PS
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaabcfacaqGtbaaaOGaaiilaaaa
@3CE8@
donne de bons résultats
comparativement à
Y
¯
^
i
(
a
)
H
:
BA
¯
=
0,033
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAamaabmaabaGaamyyaaGaayjkaiaawMca
aaqaaiaadIeaaaGccaGG6aWaa0aaaeaacaqGcbGaaeyqaaaacqGH9a
qpcaqGWaGaaeilaiaabcdacaqGZaGaae4maiaac6caaaa@455C@
Si l’on examine la
REQM
¯
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaGaaiilaaaa@3C67@
le tableau 4.1 montre
que
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaadIeaaaaaaa@3B52@
possède la plus grande valeur
(= 0,617) en raison de son grand
BA
¯
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaiaacYcaaaa@3AAF@
suivi par
μ
^
i
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaaeywaiaabkfaaaaaaa@3CF7@
et
Y
¯
^
i
PS
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaabcfacaqGtbaaaaaa@3C2E@
pour lesquels les valeurs
sont 0,442 et 0,416, respectivement. Par ailleurs, les estimateurs avec modèle
augmenté donnent des résultats significativement meilleurs que
Y
¯
^
i
PS
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaabcfacaqGtbaaaaaa@3C2E@
et
μ
^
i
YR
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaaeywaiaabkfaaaGccaGGUaaaaa@3DB3@
Par exemple, le choix
g
i
j
=
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9iaadchadaWgaaWcbaWa
aqGaaeaacaWGQbaacaGLiWoacaWGPbaabeaaaaa@40FE@
donne
REQM
¯
=
0,151
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaGaeyypa0JaaeimaiaabYcacaqG
XaGaaeynaiaabgdacaGGUaaaaa@40F1@
Parmi les quatre choix de
g
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacYcaaaa@3C14@
le choix
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
donne la plus grande
REQM
¯
(
=
0,242
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaWaaeWaaeaacqGH9aqpcaqGWaGa
aeilaiaabkdacaqG0aGaaeOmaaGaayjkaiaawMcaaiaac6caaaa@427B@
Nous avons également calculé
le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
et la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
des estimateurs EBLUP
approximatifs
μ
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaamisaaaaaaa@3C13@
et
μ
^
i
(
a
)
H
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaamisaaaakiaac6caaaa@3F3E@
Nous avons constaté que les
valeurs sont presque les mêmes que les valeurs correspondantes pour
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAaaqaaiaadIeaaaaaaa@3B52@
et
Y
¯
^
i
(
a
)
H
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWGzbGbae
HbaKaadaqhaaWcbaGaamyAamaabmaabaGaamyyaaGaayjkaiaawMca
aaqaaiaadIeaaaGccaGGUaaaaa@3E7D@
Enfin, en ce qui concerne l’estimation de l’EQM ,
eqm
(
μ
^
i
H
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgaaeaa
caWGibaaaaGccaGLOaGaayzkaaaaaa@4072@
produit le
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD3@
le plus grand, soit
53,1 % comparativement à 3,8 % pour
μ
^
i
YR
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaaeywaiaabkfaaaGccaGGSaaaaa@3DB1@
quoique la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
de
μ
^
i
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbaabaGaaeywaiaabkfaaaaaaa@3CF7@
est plus grande que celle de
μ
^
i
(
a
)
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaamisaaaaaaa@3E82@
en se basant sur
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3CF9@
ou
n
i
w
j
|
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaam4DamaaBaaaleaadaabcaqaaiaadQga
aiaawIa7aiaadMgaaeqaaOGaaiOlaaaa@3FD3@
Les estimateurs de l’EQM
eqm
(
μ
^
i
(
a
)
H
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgadaqa
daqaaiaadggaaiaawIcacaGLPaaaaeaacaWGibaaaaGccaGLOaGaay
zkaaaaaa@42E1@
et
eqm
(
μ
^
i
(
a
)
YR
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGLbGaae
yCaiaab2gadaqadaqaaiqbeY7aTzaajaWaa0baaSqaaiaadMgadaqa
daqaaiaadggaaiaawIcacaGLPaaaaeaacaqGzbGaaeOuaaaaaOGaay
jkaiaawMcaaaaa@43C5@
donnent un petit
BRA
¯
(
<
7
%
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaadaqadaqaaabaaaaaaaaapeGaeyipaWJa
aG4naiaaysW7caGGLaaapaGaayjkaiaawMcaaiaacYcaaaa@4137@
sauf pour le choix
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
qui donne
BRA
¯
=
62,6
%
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaacqGH9aqpcaqG2aGaaeOmaiaabYcacaqG
2aGaaGjbVlaacwcaaaa@40E6@
pour
μ
^
i
(
a
)
H
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaamisaaaaaaa@3E82@
et
BRA
¯
=
39,6
%
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaacqGH9aqpcaqGZaGaaeyoaiaabYcacaqG
2aGaaGjbVlaacwcaaaa@40EA@
pour
μ
^
i
(
a
)
YR
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaH8oqBga
qcamaaDaaaleaacaWGPbWaaeWaaeaacaWGHbaacaGLOaGaayzkaaaa
baGaaeywaiaabkfaaaGccaGGUaaaaa@4022@
Tableau 4.1
Biais absolu moyen
(
BA
¯
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqipu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meqabeqadiqaceGabeqabeWabeqaeeaakeaadaqadaqaam
aanaaabaGaaeOqaiaabgeaaaaacaGLOaGaayzkaaGaaiilaaaa@3C32@
REQM moyenne
(
REQM
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqipu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meqabeqadiqaceGabeqabeWabeqaeeaakeaadaqadaqaam
aanaaabaGaaeOuaiaabweacaqGrbGaaeytaaaaaiaawIcacaGLPaaa
aaa@3D3A@
des estimateurs et biais relatif absolu moyen en pourcentage
(
BRA
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqipu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meqabeqadiqaceGabeqabeWabeqaeeaakeaadaqadaqaam
aanaaabaGaaeOqaiaabkfacaqGbbaaaaGaayjkaiaawMcaaaaa@3C57@
des estimateurs de l’EQM : mesures de taille de Pfeffermann et Sverchkov
Sommaire du tableau
Le tableau montre les résultats de Biais absolu moyen XXXX et REQM moyenne XXXX des estimateurs et biais relatif absolu moyen en pourcentage XXXX des estimateurs de l’EQM : mesures de taille de Pfeffermann et Sverchkov (4.2) . Les données sont présentées selon Mesure de performance (titres de rangée) et EBLUP , pseudo-EBLUP , PS et XXXX, calculées selon XXXX unités de mesure (figurant comme en-tête de colonne).
Mesure de performance
EBLUP
pseudo-EBLUP
PS
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgaaeaacaWGibaaaaaa@3C25@
Y
¯
^
i
(
a
)
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgadaqadaqaaiaacggaaiaawIcacaGLPaaa
aeaacaWGibaaaaaa@3E93@
μ
^
i
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAaaqaaiaabMfacaqGsbaaaaaa@3DCA@
μ
^
i
(
a
)
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAamaabmaabaGaaiyyaaGaayjkaiaawMcaaaqa
aiaabMfacaqGsbaaaaaa@4038@
Y
¯
^
i
PS
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgaaeaacaqGqbGaae4uaaaaaaa@3D01@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCC@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FEA@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD3@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409C@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCC@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FEA@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD3@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409C@
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@3C22@
0,456
0,042
0,004
0,131
0,003
0,044
0,007
0,004
0,044
0,003
0,033
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3DDA@
0,617
0,151
0,147
0,242
0,101
0,442
0,157
0,156
0,207
0,106
0,416
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3CF6@
en % (eqm)
53,1
3,7
6,7
62,6
6,9
3,8
4,1
5,2
39,6
6,7
Cette cellule ne contient aucune données
4.3 Sélection de la variable d’augmentation
À la présente section, nous illustrons la sélection de la variable
d’augmentation en générant des données pour les
N
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGobaaaa@3938@
unités de la population au
moyen du modèle (4.1), puis en sélectionnant un échantillon dans la population de
données selon la méthode de Rao-Sampford
en utilisant les mesures de taille (4.2). En posant que
u
i
j
=
v
i
+
e
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG1bWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9iaadAhadaWgaaWcbaGa
amyAaaqabaGccqGHRaWkcaWGLbWaaSbaaSqaaiaadMgacaWGQbaabe
aakiaacYcaaaa@4326@
nous avons ajusté le modèle
y
i
j
=
β
0
+
β
1
x
i
j
+
u
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG5bWaaS
baaSqaaiaadMgacaWGQbaabeaakiabg2da9iabek7aInaaBaaaleaa
caaIWaaabeaakiabgUcaRiabek7aInaaBaaaleaacaaIXaaabeaaki
aadIhadaWgaaWcbaGaamyAaiaadQgaaeqaaOGaey4kaSIaamyDamaa
BaaaleaacaWGPbGaamOAaaqabaaaaa@4976@
aux données d’échantillon par
la méthode des moindres carrés ordinaires (MCO) et obtenu les résidus
u
˜
i
j
=
y
i
j
−
β
˜
0
−
β
˜
1
x
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWG1bGbaG
aadaWgaaWcbaGaamyAaiaadQgaaeqaaOGaeyypa0JaamyEamaaBaaa
leaacaWGPbGaamOAaaqabaGccqGHsislcuaHYoGygaacamaaBaaale
aacaaIWaaabeaakiabgkHiTiqbek7aIzaaiaWaaSbaaSqaaiaaigda
aeqaaOGaamiEamaaBaaaleaacaWGPbGaamOAaaqabaGccaGGSaaaaa@4A73@
où
β
˜
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHYoGyga
acamaaBaaaleaacaaIWaaabeaaaaa@3AFB@
et
β
˜
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHYoGyga
acamaaBaaaleaacaaIXaaabeaaaaa@3AFC@
sont les estimateurs par les
MCO de
β
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHYoGyda
WgaaWcbaGaaGimaaqabaaaaa@3AEC@
et
β
1
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHYoGyda
WgaaWcbaGaaGymaaqabaGccaGGSaaaaa@3BA7@
respectivement.
La figure 4.1 donne les graphiques des résidus de
(
u
˜
i
j
,
p
j
|
i
)
,
(
u
˜
i
j
,
log
p
j
|
i
)
,
(
u
˜
i
j
,
n
i
w
j
|
i
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
qadwhagaacamaaBaaaleaacaWGPbGaamOAaaqabaGccaGGSaGaamiC
amaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaGcca
GLOaGaayzkaaGaaiilamaabmaabaGabmyDayaaiaWaaSbaaSqaaiaa
dMgacaWGQbaabeaakiaacYcaciGGSbGaai4BaiaacEgacaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaakiaawIca
caGLPaaacaGGSaWaaeWaaeaaceWG1bGbaGaadaWgaaWcbaGaamyAai
aadQgaaeqaaOGaaiilaiaad6gadaWgaaWcbaGaamyAaaqabaGccaWG
3bWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaki
aawIcacaGLPaaaaaa@5C8C@
et
(
u
˜
i
j
,
w
j
|
i
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
qadwhagaacamaaBaaaleaacaWGPbGaamOAaaqabaGccaGGSaGaam4D
amaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaGcca
GLOaGaayzkaaGaaiOlaaaa@4311@
Les quatre graphiques
indiquent clairement que l’échantillonnage est informatif. Les relations
linéaires entre
u
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG1bWaaS
baaSqaaiaadMgacaWGQbaabeaaaaa@3B68@
et les deux choix
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3CF9@
et
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaciGGSbGaai
4BaiaacEgacaWGWbWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGa
amyAaaqabaaaaa@3FC9@
donnent à penser que n’importe
lequel de ces choix devrait donner de bons résultats. Le choix
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
indique une certaine non-linéarité et une plus
grande dispersion du diagramme des résidus, et ce choix est, parmi les quatre,
celui qui a produit la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
la plus grande, comme le montre le tableau 4.1. Le choix
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaam4DamaaBaaaleaadaabcaqaaiaadQga
aiaawIa7aiaadMgaaeqaaaaa@3F17@
révèle aussi une certaine
non-linéarité, mais une dispersion moins importante du graphique des résidus.
Description de la figure 4.1
La figure 4.1 présente quatre nuages de points donnant la relation entre
u
˜
i j
=
y
i j
−
β
˜
0
−
β
˜
1
x
i j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbba9=b0P0RWFb9fq0FXxbbf9Ff0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmyDayaaia
WaaSbaaSqaaiaadMgacaWGQbaabeaakiabg2da9iaadMhadaWgaaWc
baGaamyAaiaadQgaaeqaaOGaeyOeI0IafqOSdiMbaGaadaWgaaWcba
GaaGimaaqabaGccqGHsislcuaHYoGygaacamaaBaaaleaacaaIXaaa
beaakiaadIhadaWgaaWcbaGaamyAaiaadQgaaeqaaaaa@494D@
(en ordonnée) et quatre valeurs de
g
i j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbba9=b0P0RWFb9fq0FXxbbf9Ff0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaam4zamaaBa
aaleaacaWGPbGaamOAaaqabaaaaa@3AEE@
(en abscisse), soient
p
j | i
, log
p
j | i
,
n
i
w
j | i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbba9=b0P0RWFb9fq0FXxbbf9Ff0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaOGaaiilaiaa
bccaciGGSbGaai4BaiaacEgacaWGWbWaaSbaaSqaamaaeiaabaGaam
OAaaGaayjcSdGaamyAaaqabaGccaGGSaGaaeiiaiaad6gadaWgaaWc
baGaamyAaaqabaGccaWG3bWaaSbaaSqaamaaeiaabaGaamOAaaGaay
jcSdGaamyAaaqabaaaaa@4D5D@
et
w
j | i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbba9=b0P0RWFb9fq0FXxbbf9Ff0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaOGaaiOlaaaa
@3D50@
Une relation linéaire apparaît dans le nuage de points présentant les couples
(
u
˜
i j
,
p
j | i
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbba9=b0P0RWFb9fq0FXxbbf9Ff0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaeWaaeaace
WG1bGbaGaadaWgaaWcbaGaamyAaiaadQgaaeqaaOGaaiilaiaadcha
daWgaaWcbaWaaqGaaeaacaWGQbaacaGLiWoacaWGPbaabeaaaOGaay
jkaiaawMcaaaaa@41EC@
et dans celui présentant les couples
(
u
˜
i j
, log
p
j | i
) .
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbba9=b0P0RWFb9fq0FXxbbf9Ff0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaeWaaeaace
WG1bGbaGaadaWgaaWcbaGaamyAaiaadQgaaeqaaOGaaiilaiGacYga
caGGVbGaai4zaiaadchadaWgaaWcbaWaaqGaaeaacaWGQbaacaGLiW
oacaWGPbaabeaaaOGaayjkaiaawMcaaiaac6caaaa@456E@
Les deux autres nuages de points montrent une relation non-linéaire et une pente décroissante. La dispersion du graphe des
(
u
˜
i j
,
w
j | i
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbba9=b0P0RWFb9fq0FXxbbf9Ff0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaeWaaeaace
WG1bGbaGaadaWgaaWcbaGaamyAaiaadQgaaeqaaOGaaiilaiaadEha
daWgaaWcbaWaaqGaaeaacaWGQbaacaGLiWoacaWGPbaabeaaaOGaay
jkaiaawMcaaaaa@41F3@
est particulièrement prononcée.
Nous avons également ajusté le modèle augmenté (1.4) avec
g
(
p
j
|
i
)
=
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaae
WaaeaacaWGWbWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyA
aaqabaaakiaawIcacaGLPaaacqGH9aqpcaWGWbWaaSbaaSqaamaaei
aabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@4512@
et calculé les résidus par les MCO
u
˜
0
i
j
=
y
i
j
−
β
˜
00
−
β
˜
01
x
i
j
−
δ
˜
0
p
j
|
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWG1bGbaG
aadaWgaaWcbaGaaGimaiaadMgacaWGQbaabeaakiabg2da9iaadMha
daWgaaWcbaGaamyAaiaadQgaaeqaaOGaeyOeI0IafqOSdiMbaGaada
WgaaWcbaGaaGimaiaaicdaaeqaaOGaeyOeI0IafqOSdiMbaGaadaWg
aaWcbaGaaGimaiaaigdaaeqaaOGaamiEamaaBaaaleaacaWGPbGaam
OAaaqabaGccqGHsislcuaH0oazgaacamaaBaaaleaacaaIWaaabeaa
kiaadchadaWgaaWcbaWaaqGaaeaacaWGQbaacaGLiWoacaWGPbaabe
aakiaac6caaaa@54D2@
Tous les résidus
u
˜
0
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWG1bGbaG
aadaWgaaWcbaGaaGimaiaadMgacaWGQbaabeaaaaa@3C31@
sont inférieurs à 2,0 en valeur
absolue, ce qui laisse entendre que le modèle augmenté est adéquat.
4.4 Résultats sous les mesures de taille d’Asparouhov
Le tableau 4.2 donne les résultats des simulations pour le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
sous les mesures de taille
(4.3) et (4.4) d’Asparouhov. Il montre, comme
le tableau 4.1 pour les mesures de taille de Pfeffermann et Sverchkov, que
le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
de l’EBLUP est grand (0,437 pour
la mesure de taille invariante (I) et 0,440 pour la mesure de taille non invariante
(NI)) quand la variable d’augmentation,
g
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacYcaaaa@3C14@
n’est pas incluse dans le
modèle et que l’échantillonnage est très informatif
(
α
=
1
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iaaigdaaiaawIcacaGLPaaacaGGUaaaaa@3E00@
De plus, le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
diminue à mesure que
α
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyaa
a@3A04@
augmente. Par ailleurs, sous
le même modèle, le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
associé au pseudo-EBLUP est beaucoup plus
faible : 0,048 pour I et 0,047 pour NI quand
α
=
1
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqycq
GH9aqpcaaIXaGaaiilaaaa@3C75@
et le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
diminue quand
α
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyaa
a@3A04@
augmente. L’estimateur de Pfeffermann
et Sverchkov sous le même modèle présente aussi un
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
plus faible (environ 0,01), quel
que soit le choix de la valeur de
α
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyca
GGUaaaaa@3AB6@
L’inclusion de
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3CF9@
ou
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaam4DamaaBaaaleaadaabcaqaaiaadQga
aiaawIa7aiaadMgaaeqaaaaa@3F17@
ou
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaciGGSbGaai
4BaiaacEgacaWGWbWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGa
amyAaaqabaaaaa@3FC9@
comme variable d’augmentation
dans le modèle donne aussi lieu à un petit
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
pour l’EBLUP (0,02 ou moins),
quelle que soit la valeur de
α
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyca
GGUaaaaa@3AB6@
Par ailleurs, le choix
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
comme variable d’augmentation
entraîne un plus grand
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
(0,14 pour
α
=
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqycq
GH9aqpcaaIXaaaaa@3BC5@
et
2 ) ,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqacaqaai
aaikdaaiaawMcaaiaacYcaaaa@3A99@
sauf en cas d’échantillonnage non
informatif
(
α
=
∞
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iabg6HiLcGaayjkaiaawMcaaiaac6caaaa@3EB6@
Ces résultats médiocres pour
le choix
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
sont probablement dus au fait
que
w
j
|
i
=
(
n
i
p
j
|
i
)
−
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaGccqGH9aqp
daqadaqaaiaad6gadaWgaaWcbaGaamyAaaqabaGccaWGWbWaaSbaaS
qaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaakiaawIcacaGL
PaaadaahaaWcbeqaaiabgkHiTiaaigdaaaaaaa@4823@
dépend des tailles d’échantillon de domaine,
n
i
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaaiilaaaa@3B2C@
quand celles-ci ne sont pas
égales, contrairement aux autres choix de
g
i
j
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaac6caaaa@3C16@
Le pseudo-EBLUP donne des
résultats comparables à l’EBLUP sous le modèle augmenté en ce qui concerne le
BA
¯
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaiaac6caaaa@3AB1@
Tableau 4.2
Biais absolu moyen
(
BA
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqipu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meqabeqadiqaceGabeqabeWabeqaeeaakeaadaqadaqaam
aanaaabaGaaeOqaiaabgeaaaaacaGLOaGaayzkaaaaaa@3B82@
des estimateurs sous les mesures de taille d’Asparouhov : invariante (I) et non invariante (NI)
Sommaire du tableau
Le tableau montre les résultats de Biais absolu moyen XXXX des estimateurs sous les mesures de taille d’Asparouhov : invariante (I) et non invariante (NI). Les données sont présentées selon XXXX (titres de rangée) et Mesure de taille, EBLUP , pseudo-EBLUP , PS et XXXX, calculées selon XXXX unités de mesure (figurant comme en-tête de colonne).
α
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaeqySdegaaa@3AD6@
Mesure de taille
EBLUP
pseudo-EBLUP
PS
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgaaeaacaWGibaaaaaa@3C24@
Y
¯
^
i
(
a
)
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgadaqadaqaaiaacggaaiaawIcacaGLPaaa
aeaacaWGibaaaaaa@3E92@
μ
^
i
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAaaqaaiaabMfacaqGsbaaaaaa@3DC9@
μ
^
i
(
a
)
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAamaabmaabaGaaiyyaaGaayjkaiaawMcaaaqa
aiaabMfacaqGsbaaaaaa@4037@
Y
¯
^
i
PS
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgaaeaacaqGqbGaae4uaaaaaaa@3D00@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCB@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FE9@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD2@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409B@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCB@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FE9@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD2@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409B@
1
I
0,437
0,001
0,005
0,140
0,022
0,048
0,001
0,006
0,057
0,005
0,012
NI
0,440
0,007
0,007
0,145
0,021
0,047
0,003
0,007
0,064
0,005
0,013
2
I
0,217
0,009
0,010
0,137
0,014
0,024
0,010
0,010
0,098
0,010
0,012
NI
0,217
0,011
0,009
0,136
0,011
0,024
0,009
0,010
0,098
0,010
0,012
3
I
0,145
0,010
0,010
0,101
0,011
0,017
0,010
0,010
0,075
0,010
0,011
NI
0,144
0,011
0,011
0,099
0,012
0,016
0,010
0,011
0,074
0,011
0,011
∞
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaeyOhIukaaa@3A9E@
I
0,011
0,011
0,011
0,011
0,011
0,012
0,011
0,011
0,012
0,011
0,011
NI
0,010
0,010
0,010
0,010
0,010
0,010
0,010
0,010
0,010
0,010
0,010
Le tableau 4.3 présente les résultats des simulations concernant
la racine carrée moyenne de l’erreur
quadratique moyenne
(
REQM
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOuaiaabweacaqGrbGaaeytaaaaaiaawIcacaGLPaaa
aaa@3D40@
lorsque l’on utilise les mesures
de taille (4.3) et (4.4) d’Asparouhov. Les
résultats montrent que l’EBLUP , fondé sur le modèle (1.4) sans la variable
d’augmentation
g
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacYcaaaa@3C14@
est celui dont la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
est la plus grande (0,596 pour
I et 0,619 pour NI ) quand l’échantillonnage est très informatif
(
α
=
1
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iaaigdaaiaawIcacaGLPaaacaGGUaaaaa@3E00@
La
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
diminue progressivement
jusqu’à environ 0,42 au fur et à mesure que l’échantillonnage devient non informatif
(
α
=
∞
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iabg6HiLcGaayjkaiaawMcaaiaac6caaaa@3EB6@
Par ailleurs, la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
du pseudo-EBLUP (sans le
terme
g
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9pC0xbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaam4zamaaBa
aaleaacaWGPbGaamOAaaqabaaaaa@3B4A@
dans le modèle) et de l’estimateur PS ne dépend
pas de
α
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyca
GGSaaaaa@3AB4@
et est considérablement réduite : la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
du pseudo-EBLUP est de
l’ordre de 0,44 et la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
de l’estimateur PS est
un peu plus petite, de l’ordre de 0,42. L’augmentation de la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
des estimateurs pseudo-EBLUP
et PS par rapport à l’estimateur EBLUP sous échantillonnage non informatif
(
α
=
∞
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iabg6HiLcGaayjkaiaawMcaaaaa@3E04@
est également faible. Par
ailleurs, l’utilisation de l’EBLUP et du pseudo-EBLUP sous le modèle augmenté donne
lieu à une réduction importante de l’EQM quand l’échantillonnage est très informatif
(
α
=
1
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iaaigdaaiaawIcacaGLPaaacaGGSaaaaa@3DFE@
particulièrement pour les choix
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3CF9@
et
log
p
j
|
i
:
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaciGGSbGaai
4BaiaacEgacaWGWbWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGa
amyAaaqabaGccaGG6aWaa0aaaeaacaqGsbGaaeyraiaabgfacaqGnb
aaaaaa@43E3@
inférieure à 0,15. Le choix
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
donne une plus grande
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
(autour de 0,29) quand
α
=
1
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqycq
GH9aqpcaaIXaGaaiilaaaa@3C75@
mais elle demeure néanmoins nettement
plus petite que la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
pour le pseudo-EBLUP sans le
terme
g
i
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9pC0xbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaam4zamaaBa
aaleaacaWGPbGaamOAaaqabaaaaa@3B4A@
et l’estimateur PS. À
mesure que
α
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyaa
a@3A04@
augmente, la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
devient à peu près la même pour
les estimateurs EBLUP (sous le modèle augmenté), pseudo-EBLUP et PS.
Tableau 4.3
Racine carrée moyenne de l’erreur quadratique moyenne
(
REQM
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqipu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meqabeqadiqaceGabeqabeWabeqaeeaakeaadaqadaqaam
aanaaabaGaaeOuaiaabweacaqGrbGaaeytaaaaaiaawIcacaGLPaaa
aaa@3D3A@
des estimateurs sous les mesures de taille d’Asparouhov : invariante (I) et non invariante (NI)
Sommaire du tableau
Le tableau montre les résultats de Racine carrée moyenne de l’erreur quadratique moyenne XXXX des estimateurs sous les mesures de taille d’Asparouhov : invariante (I) et non invariante (NI). Les données sont présentées selon XXXX (titres de rangée) et Mesure de taille, EBLUP , pseudo-EBLUP , PS et XXXX, calculées selon XXXX unités de mesure (figurant comme en-tête de colonne).
α
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaeqySdegaaa@3AD6@
Mesure de taille
EBLUP
pseudo-EBLUP
PS
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgaaeaacaWGibaaaaaa@3C24@
Y
¯
^
i
(
a
)
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgadaqadaqaaiaacggaaiaawIcacaGLPaaa
aeaacaWGibaaaaaa@3E92@
μ
^
i
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAaaqaaiaabMfacaqGsbaaaaaa@3DC9@
μ
^
i
(
a
)
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAamaabmaabaGaaiyyaaGaayjkaiaawMcaaaqa
aiaabMfacaqGsbaaaaaa@4037@
Y
¯
^
i
PS
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgaaeaacaqGqbGaae4uaaaaaaa@3D00@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCB@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FE9@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD2@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409B@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCB@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FE9@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD2@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409B@
1
I
0,596
0,039
0,203
0,281
0,108
0,454
0,040
0,223
0,258
0,112
0,406
NI
0,619
0,110
0,205
0,295
0,135
0,457
0,092
0,235
0,273
0,136
0,435
2
I
0,468
0,377
0,385
0,418
0,379
0,436
0,391
0,398
0,415
0,392
0,416
NI
0,474
0,375
0,378
0,414
0,374
0,438
0,392
0,396
0,413
0,391
0,423
3
I
0,439
0,400
0,403
0,420
0,401
0,432
0,414
0,417
0,425
0,415
0,415
NI
0,443
0,400
0,401
0,418
0,399
0,435
0,416
0,416
0,425
0,415
0,420
∞
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaeyOhIukaaa@3A9E@
I
0,417
0,418
0,418
0,418
0,418
0,431
0,431
0,431
0,432
0,431
0,418
NI
0,418
0,418
0,418
0,419
0,418
0,432
0,432
0,432
0,433
0,432
0,418
Le tableau 4.4 donne les résultats des simulations concernant
le biais relatif absolu moyen
(
BRA
¯
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaam
aanaaabaGaaeOqaiaabkfacaqGbbaaaaGaayjkaiaawMcaaaaa@3C5D@
des estimateurs de l’EQM sous
les mesures de taille (4.3) et (4.4) d’Asparouhov.
Il montre que le
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD4@
de l’estimateur de l’EQM de
l’EBLUP , basé sur le modèle sans la variable d’augmentation
g
i
j
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacYcaaaa@3C14@
est très grand quand l’échantillonnage
est très informatif
(
α
=
1
)
:
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iaaigdaaiaawIcacaGLPaaacaGG6aaaaa@3E0C@
52,8 % pour I et 59,1 % pour NI . Le
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD4@
diminue progressivement jusqu’à
environ 5 % sous échantillonnage non informatif
(
α
=
∞
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abeg7aHjabg2da9iabg6HiLcGaayjkaiaawMcaaiaac6caaaa@3EB6@
L’utilisation de
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaciGGSbGaai
4BaiaacEgacaWGWbWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGa
amyAaaqabaaaaa@3FC9@
comme variable d’augmentation
donne lieu à une grande réduction du
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD4@
(
<
9
%
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
abgYda8iaaiMdacaaMe8UaaiyjaaGaayjkaiaawMcaaaaa@3DEB@
et les choix
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3CF9@
et
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbWaaS
baaSqaaiaadMgaaeqaaOGaam4DamaaBaaaleaadaabcaqaaiaadQga
aiaawIa7aiaadMgaaeqaaaaa@3F17@
donnent aussi de bons
résultats en ce qui concerne le
BRA
¯
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaacaGGSaaaaa@3B84@
sauf dans le cas de NI et
α
=
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqycq
GH9aqpcaaIXaaaaa@3BC5@
où le BRA vaut 18,5 % et
12,9 %, respectivement. De nouveau,
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG3bWaaS
baaSqaamaaeiaabaGaamOAaaGaayjcSdGaamyAaaqabaaaaa@3D00@
n’est pas un bon choix parce
qu’il donne un
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD4@
aussi grand que 40 % quand
α
=
1.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqycq
GH9aqpcaaIXaGaaiOlaaaa@3C77@
L’estimateur de l’EQM associé
au pseudo-EBLUP (sans
g
i
j
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGNbWaaS
baaSqaaiaadMgacaWGQbaabeaakiaacMcaaaa@3C11@
donne aussi de bons
résultats, sauf pour NI et
α
=
1
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqycq
GH9aqpcaaIXaGaaiilaaaa@3C75@
qui mènent à un
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD4@
de 19,5 %. L’utilisation
de
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaciGGSbGaai
4BaiaacEgacaWGWbWaaSbaaSqaamaaeiaabaGaamOAaaGaayjcSdGa
amyAaaqabaaaaa@3FC9@
comme variable auxiliaire produit
un
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD4@
inférieur à 8 % pour l’estimateur
de l’EQM associé au pseudo-EBLUP . Nous n’avons pas inclus l’estimateur
bootstrap de l’EQM de Pfeffermann et Sverchkov dans notre étude.
Dans l’ensemble, notre étude en simulation indique que l’utilisation
de modèles augmentés sous échantillonnage informatif produit des estimateurs
EBLUP qui donnent de bons résultats en ce qui concerne le
BA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGbbaaaaaa@39FF@
et la
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
des estimateurs, et le
BRA
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkeacaqGsbGaaeyqaaaaaaa@3AD4@
des estimateurs de l’EQM , à
condition que la variable d’augmentation soit choisie convenablement. Les
estimateurs corrigés du biais de Pfeffermann et Sverchkov donnent aussi de bons
résultats, quoiqu’ils produisent une plus grande
REQM
¯
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqdaaqaai
aabkfacaqGfbGaaeyuaiaab2eaaaaaaa@3BB7@
sous les mesures de taille
(4.2) de Pfeffermann et Sverchkov. Les estimateurs pseudo-EBLUP (sans la variable
d’augmentation) sont également bons et une amélioration supplémentaire peut
être obtenue sous les modèles augmentés.
Tableau 4.4
Biais relatif moyen (%) des estimateurs de l’EQM sous les mesures de taille d’Asparouhov : invariante (I) et non invariante (NI)
Sommaire du tableau
Le tableau montre les résultats de Biais relatif moyen (%) des estimateurs de l’EQM sous les mesures de taille d’Asparouhov : invariante (I) et non invariante (NI). Les données sont présentées selon XXXX (titres de rangée) et Mesure de taille, EBLUP , pseudo-EBLUP et XXXX(figurant comme en-tête de colonne).
α
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaeqySdegaaa@3AD6@
Mesure de taille
EBLUP
pseudo-EBLUP
Y
¯
^
i
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgaaeaacaWGibaaaaaa@3C24@
Y
¯
^
i
(
a
)
H
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGabmywayaary
aajaWaa0baaSqaaiaadMgadaqadaqaaiaacggaaiaawIcacaGLPaaa
aeaacaWGibaaaaaa@3E92@
μ
^
i
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAaaqaaiaabMfacaqGsbaaaaaa@3DC9@
μ
^
i
(
a
)
YR
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGafqiVd0MbaK
aadaqhaaWcbaGaamyAamaabmaabaGaaiyyaaGaayjkaiaawMcaaaqa
aiaabMfacaqGsbaaaaaa@4037@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCB@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FE9@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD2@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409B@
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamiCamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DCB@
n
i
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaamOBamaaBa
aaleaacaWGPbaabeaakiaadEhadaWgaaWcbaWaaqGaaeaacaWGQbaa
caGLiWoacaWGPbaabeaaaaa@3FE9@
w
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaam4DamaaBa
aaleaadaabcaqaaiaadQgaaiaawIa7aiaadMgaaeqaaaaa@3DD2@
log
p
j
|
i
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbeqabeWaceGabiqabeqabmqabeabbaGcbaGaciiBaiaac+
gacaGGNbGaamiCamaaBaaaleaadaabcaqaaiaadQgaaiaawIa7aiaa
dMgaaeqaaaaa@409B@
1
I
52,8
6,5
4,8
39,8
3,3
11,7
6,6
7,8
19,2
6,2
NI
59,1
18,5
12,9
39,4
7,8
19,5
26,0
10,2
16,6
6,0
2
I
19,4
6,0
5,5
10,7
5,9
3,9
6,3
6,0
7,3
6,4
NI
22,6
8,8
8,0
11,3
8,6
4,2
6,7
6,0
7,4
6,7
3
I
7,1
5,5
5,5
5,3
5,5
4,4
6,0
6,3
7,2
6,3
NI
8,9
7,3
7,0
5,9
7,2
4,0
7,1
7,0
7,3
7,2
∞
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGak0Jf9crFfpeea0xh9v8qiW7rqqrpipeea0xe9Lq=Je9
vqaqFeFr0xbbG8FaYPYRWFb9fi0FXxbbf9=e0dfrpm0dXdHqVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaGaeyOhIukaaa@3A9E@
I
5,1
5,1
5,0
5,0
5,1
5,1
5,2
5,3
5,3
5,2
NI
5,0
4,9
4,9
4,9
4,9
4,9
5,0
5,1
5,1
5,0
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, 2015
L'utilisation de la présente publication est assujettie aux modalités de l'Entente de licence ouverte de Statistique Canada .
No 12-001-X au catalogue
Périodicité : Semi-annuel
Ottawa
Date de modification :
2017-09-20