Combiner l’échantillonnage par dépistage de liens et l’échantillonnage en grappes pour estimer la taille d’une population cachée en présence de probabilités de lien hétérogènes
3. Estimateurs du maximum de vraisemblance de
τ
1
,
τ
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiVv0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9
vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x
fr=xb9adbeqabeWaceGabiqabeqabmqabeabbaGcbaGaeqiXdq3aaS
baaSqaaiaaigdaaeqaaOGaaiilaiabes8a0naaBaaaleaacaaIYaaa
beaaaaa@3C52@
et
τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiVv0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9
vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x
fr=xb9adbeqabeWaceGabiqabeqabmqabeabbaGcbaGaeqiXdqhaaa@3804@
Combiner l’échantillonnage par dépistage de liens et l’échantillonnage en grappes pour estimer la taille d’une population cachée en présence de probabilités de lien hétérogènes
3. Estimateurs du maximum de vraisemblance de
τ
1
,
τ
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiVv0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9
vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x
fr=xb9adbeqabeWaceGabiqabeqabmqabeabbaGcbaGaeqiXdq3aaS
baaSqaaiaaigdaaeqaaOGaaiilaiabes8a0naaBaaaleaacaaIYaaa
beaaaaa@3C52@
et
τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiVv0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9
vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x
fr=xb9adbeqabeWaceGabiqabeqabmqabeabbaGcbaGaeqiXdqhaaa@3804@
3.1 Modèles
probabilistes
Pour construire les EMV des
τ
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDca
GGSaaaaa@3ADA@
nous devons spécifier des
modèles pour les variables observées. Donc, comme dans Félix-Medina et Thompson (2004), nous supposons que les nombres
M
1
,
…
,
M
N
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaaigdaaeqaaOGaaGilaiablAciljaaiYcacaWGnbWaaSba
aSqaaiaad6eaaeqaaaaa@3E87@
de personnes qui
appartiennent aux sites
A
1
,
…
,
A
N
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaaigdaaeqaaOGaaGilaiablAciljaaiYcacaWGbbWaaSba
aSqaaiaad6eaaeqaaaaa@3E6F@
sont des variables aléatoires
de Poisson indépendantes de moyenne
λ
1
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaH7oaBda
WgaaWcbaGaaGymaaqabaGccaGGUaaaaa@3BBC@
Par conséquent, la loi
conditionnelle conjointe de
(
M
1
,
…
,
M
n
,
τ
1
−
M
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aad2eadaWgaaWcbaGaaGymaaqabaGccaaISaGaeSOjGSKaaGilaiaa
d2eadaWgaaWcbaGaamOBaaqabaGccaaISaGaeqiXdq3aaSbaaSqaai
aaigdaaeqaaOGaeyOeI0IaamytaaGaayjkaiaawMcaaaaa@4565@
sachant que
∑
1
N
M
i
=
τ
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaaeWaqabS
qaaiaaigdaaeaacaWGobaaniabggHiLdGccaWGnbWaaSbaaSqaaiaa
dMgaaeqaaOGaeyypa0JaeqiXdq3aaSbaaSqaaiaaigdaaeqaaaaa@41A8@
est une loi multinomiale dont
la fonction de masse de probabilité (fmp) est :
f (
m
1
, … ,
m
n
,
τ
1
− m
) =
τ
1
!
∏
1
n
m
i
! (
τ
1
− m
) !
(
1
N
)
m
(
1 −
n
N
)
τ
1
− m
. ( 3.1 )
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGMbWaae
WaaeaacaWGTbWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiablAciljaa
iYcacaWGTbWaaSbaaSqaaiaad6gaaeqaaOGaaGilaiabes8a0naaBa
aaleaacaaIXaaabeaakiabgkHiTiaad2gaaiaawIcacaGLPaaacqGH
9aqpdaWcaaqaaiabes8a0naaBaaaleaacaaIXaaabeaakiaacgcaae
aadaqeWbqabSqaaiaaigdaaeaacaWGUbaaniabg+GivdGccaWGTbWa
aSbaaSqaaiaadMgaaeqaaOGaaiyiamaabmaabaGaeqiXdq3aaSbaaS
qaaiaaigdaaeqaaOGaeyOeI0IaamyBaaGaayjkaiaawMcaaiaacgca
aaWaaeWaaeaadaWcaaqaaiaaigdaaeaacaWGobaaaaGaayjkaiaawM
caamaaCaaaleqabaGaamyBaaaakmaabmaabaGaaGymaiabgkHiTmaa
laaabaGaamOBaaqaaiaad6eaaaaacaGLOaGaayzkaaWaaWbaaSqabe
aacqaHepaDdaWgaaadbaGaaGymaaqabaWccqGHsislcaWGTbaaaOGa
aGOlaiaaywW7caaMf8UaaGzbVlaaywW7caaMf8Uaaiikaiaaiodaca
GGUaGaaGymaiaacMcaaaa@72AE@
Pour
modéliser les liens entre les membres de la population et les sites
échantillonnés, nous définissons les variables aléatoires suivantes :
X
i j
(
k
)
= 1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGybWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaeyypa0JaaGymaaaa@3F90@
si la personne
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbaaaa@3954@
dans
U
k
−
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0IaamyqamaaBaaaleaacaWGPbaa
beaaaaa@3D32@
est liée au site
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaaaa@3A45@
et
X
i j
(
k
)
= 0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGybWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaeyypa0JaaGimaaaa@3F8F@
si
j
∈
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbGaey
icI4SaamyqamaaBaaaleaacaWGPbaabeaaaaa@3CB8@
ou que la
personne n’est pas liée à
A
i
, j = 1 , … ,
τ
k
, i = 1 , … , n .
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaaiilaiaadQgacqGH9aqpcaaIXaGaaiil
aiablAciljaaiYcacqaHepaDdaWgaaWcbaGaam4AaaqabaGccaGGSa
GaamyAaiabg2da9iaaigdacaGGSaGaeSOjGSKaaGilaiaad6gacaGG
Uaaaaa@4AAE@
Nous supposons
que, étant donné l’échantillon
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGtbWaaS
baaSqaaiaadgeaaeqaaaaa@3A2F@
de sites, les
X
i
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGybWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaaaa@3DC5@
sont des
variables aléatoires de Bernoulli de moyenne
p
i
j
(
k
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaaiilaaaa@3E97@
où la probabilité de lien
p
i
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaaaa@3DDD@
satisfait le modèle
de Rasch suivant :
p
i j
(
k
)
= Pr (
X
i j
(
k
)
= 1 |
β
j
(
k
)
,
S
A
) =
exp (
α
i
(
k
)
+
β
j
(
k
)
)
1 + exp (
α
i
(
k
)
+
β
j
(
k
)
)
, j ∈
U
k
−
A
i
; i = 1 , … , n . ( 3.2 )
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaeyypa0JaciiuaiaackhadaqadaqaaiaadIfadaqhaaWcba
GaamyAaiaadQgaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaGc
cqGH9aqpcaaIXaWaaqqaaeaacaaMc8UaeqOSdi2aa0baaSqaaiaadQ
gaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaGccaaISaGaam4u
amaaBaaaleaacaWGbbaabeaaaOGaay5bSdaacaGLOaGaayzkaaGaey
ypa0ZaaSaaaeaaciGGLbGaaiiEaiaacchadaqadaqaaiabeg7aHnaa
DaaaleaacaWGPbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzkaaaaaO
Gaey4kaSIaeqOSdi2aa0baaSqaaiaadQgaaeaadaqadaqaaiaadUga
aiaawIcacaGLPaaaaaaakiaawIcacaGLPaaaaeaacaaIXaGaey4kaS
IaciyzaiaacIhacaGGWbWaaeWaaeaacqaHXoqydaqhaaWcbaGaamyA
aaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaakiabgUcaRiabek
7aInaaDaaaleaacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaaGccaGLOaGaayzkaaaaaiaaiYcacaaMi8UaaeiiaiaadQgacq
GHiiIZcaWGvbWaaSbaaSqaaiaadUgaaeqaaOGaeyOeI0Iaamyqamaa
BaaaleaacaWGPbaabeaakiaacUdacaaMi8UaaeiiaiaadMgacqGH9a
qpcaaIXaGaaiilaiablAciljaaiYcacaWGUbGaaGOlaiaaywW7caaM
f8UaaGzbVlaaywW7caGGOaGaaG4maiaac6cacaaIYaGaaiykaaaa@9453@
Il
convient de souligner que ce modèle a été pris en considération par Coull et Agresti (1999) dans le contexte de l’échantillonnage
par capture-recapture. Dans ce modèle,
α
i
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHXoqyda
qhaaWcbaGaamyAaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaa
aaa@3D98@
est un effet
fixe (non aléatoire) qui représente la possibilité qu’a la grappe
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaaaa@3A45@
de former des
liens avec les personnes comprises dans
U
k
−
A
i
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0IaamyqamaaBaaaleaacaWGPbaa
beaakiaacYcaaaa@3DEC@
et
β
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHYoGyda
qhaaWcbaGaamOAaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaa
aaa@3D9B@
est un effet
aléatoire qui représente la propension de la personne
j
∈
U
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbGaey
icI4SaamyvamaaBaaaleaacaWGRbaabeaaaaa@3CCE@
à être liée à
une grappe. Nous supposons que
β
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHYoGyda
qhaaWcbaGaamOAaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaa
aaa@3D9B@
suit une loi
normale de moyenne 0 et de variance inconnue
σ
k
2
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHdpWCda
qhaaWcbaGaam4AaaqaaiaaikdaaaGccaGGSaaaaa@3CBB@
et que ces
variables sont indépendantes. Le paramètre
σ
k
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHdpWCda
qhaaWcbaGaam4Aaaqaaiaaikdaaaaaaa@3C01@
détermine le degré d’hétérogénéité
des
p
i
j
(
k
)
:
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaaiOoaaaa@3EA5@
de grandes valeurs
de
σ
k
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHdpWCda
qhaaWcbaGaam4Aaaqaaiaaikdaaaaaaa@3C01@
impliquent un haut degré d’hétérogénéité.
Avant de conclure la présente sous-section,
nous formulerons certains commentaires au sujet des modèles supposés. Premièrement,
la loi multinomiale des
M
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaadMgaaeqaaaaa@3A51@
observés (qui est celle utilisée dans la
fonction de vraisemblance) implique que les personnes sont distribuées
indépendamment et avec probabilités égales sur les sites de la base de sondage.
Cette hypothèse est difficile à satisfaire en pratique; cependant, comme nous
le montrerons plus loin, la fonction de vraisemblance dépend des
M
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaadMgaaeqaaaaa@3A51@
essentiellement par la voie
de leur somme
M
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbaaaa@3937@
et, puisque
N
M
/
n
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaWcgaqaai
aad6eacaWGnbaabaGaamOBaaaaaaa@3B13@
est un estimateur de
τ
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGymaaqabaaaaa@3B11@
fondé sur le plan de sondage,
c’est-à-dire qu’il s’agit d’un estimateur exempt d’une loi, il s’ensuit que
l’EMV de
τ
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGymaaqabaaaaa@3B11@
sera également robuste aux écarts
par rapport à la loi multinomiale des
M
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaadMgaaeqaaOGaaeOlaaaa@3B0C@
Néanmoins, les écarts par rapport à ce modèle auront
une incidence sur la performance des estimateurs de variance et des intervalles
de confiance calculés sous cette hypothèse. Deuxièmement, le modèle de Rasch donné par (3.2) implique ce qui
suit : i) la probabilité de lien
p
i
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaaaa@3DDD@
dépend uniquement de deux
effets : la sociabilité des personnes dans la grappe
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaaaa@3A45@
et celle de la personne
j
∈
U
k
−
A
i
;
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbGaey
icI4SaamyvamaaBaaaleaacaWGRbaabeaakiabgkHiTiaadgeadaWg
aaWcbaGaamyAaaqabaGccaGG7aaaaa@406E@
ii) les deux effets sont
additifs, et iii) pour tout site
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaaaa@3A45@
dans la base de sondage et toute
personne
j ∈ U −
A
i
,
p
i j
(
k
)
> 0.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbGaey
icI4SaamyvaiabgkHiTiaadgeadaWgaaWcbaGaamyAaaqabaGccaGG
SaGaaGjbVlaadchadaqhaaWcbaGaamyAaiaadQgaaeaadaqadaqaai
aadUgaaiaawIcacaGLPaaaaaGccaaMe8UaaeOpaiaaysW7caaIWaGa
aiOlaaaa@4B8F@
Le modèle (3.2) est un cas particulier
d’un modèle mixte linéaire généralisé. (Voir Agresti
2002, section 2.1, pour un bref examen de ce type de modèle.) Par conséquent,
nous pourrions incorporer les structures de réseau des personnes comprises dans
la grappe
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaaaa@3A45@
et de la personne
j
∈
U
k
−
A
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbGaey
icI4SaamyvamaaBaaaleaacaWGRbaabeaakiabgkHiTiaadgeadaWg
aaWcbaGaamyAaaqabaaaaa@3FA5@
pour modéliser la probabilité
de lien
p
i
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaaaa@3DDD@
en étendant le modèle (3.2) à
un modèle qui comprend les covariables associées à la personne
j
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbaaaa@3954@
et à la grappe
A
i
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaaiilaaaa@3AFF@
ainsi que leurs termes d’interaction.
Cependant, en utilisant un modèle plus général que (3.2), nous rendrons le problème
d’inférence beaucoup plus difficile que celui que nous devons résoudre dans la
présente étude. Donc, malgré la simplicité relative du modèle (3.2), nous nous
attendons à ce qu’il rende compte de l’hétérogénéité des probabilités de lien et
qu’il nous permette de faire des inférences au sujet des
τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDaa
a@3A2A@
dont au moins l’ordre de
grandeur est correct.
3.2 Fonction
de vraisemblance
Le moyen le plus facile de construire
la fonction de vraisemblance consiste à la factoriser en différentes composantes.
L’une d’elles est associée à la probabilité de sélectionner l’échantillon
initial
S
0
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGtbWaaS
baaSqaaiaaicdaaeqaaOGaaiilaaaa@3ADD@
qui est donnée par la loi
multinomiale (3.1), c’est-à-dire
L
MULT
(
τ
1
) ∝
τ
1
!
(
τ
1
− m
) !
(
1 − n / N
)
τ
1
− m
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaab2eacaqGvbGaaeitaiaabsfaaeqaaOWaaeWaaeaacqaH
epaDdaWgaaWcbaGaaGymaaqabaaakiaawIcacaGLPaaacqGHDisTda
Wcaaqaaiabes8a0naaBaaaleaacaaIXaaabeaakiaacgcaaeaadaqa
daqaaiabes8a0naaBaaaleaacaaIXaaabeaakiabgkHiTiaad2gaai
aawIcacaGLPaaacaGGHaaaamaabmaabaWaaSGbaeaacaaIXaGaeyOe
I0IaamOBaaqaaiaad6eaaaaacaGLOaGaayzkaaWaaWbaaSqabeaacq
aHepaDdaWgaaadbaGaaGymaaqabaWccqGHsislcaWGTbaaaOGaaGOl
aaaa@5739@
Deux autres composantes sont associées aux
probabilités conditionnelles des configurations de liens entre les personnes dans
U
k
−
S
0
, k = 1 , 2 ,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0Iaam4uamaaBaaaleaacaaIWaaa
beaakiaacYcacaWGRbGaeyypa0JaaGymaiaacYcacaaIYaGaaiilaa
aa@4297@
et les grappes
A
i
∈
S
A
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaakiaacYcaaaa@3E57@
sachant
S
A
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGtbWaaS
baaSqaaiaadgeaaeqaaOGaaiOlaaaa@3AEB@
Pour obtenir ces facteurs, nous
devons calculer les probabilités de certains événements. Soit
X
j
(
k
)
= (
X
1 j
(
k
)
, … ,
X
n j
(
k
)
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHybWaa0
baaSqaaiaadQgaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaGc
cqGH9aqpdaqadaqaaiaadIfadaqhaaWcbaGaaGymaiaadQgaaeaada
qadaqaaiaadUgaaiaawIcacaGLPaaaaaGccaaISaGaeSOjGSKaaGil
aiaadIfadaqhaaWcbaGaamOBaiaadQgaaeaadaqadaqaaiaadUgaai
aawIcacaGLPaaaaaaakiaawIcacaGLPaaaaaa@4CA8@
le vecteur de dimension
n
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbaaaa@3958@
de variables indicatrices de
lien
X
i
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGybWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaaaa@3DC5@
associées à la
j
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbWaaW
baaSqabeaacaqGLbaaaaaa@3A69@
personne dans
U
k
−
S
0
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0Iaam4uamaaBaaaleaacaaIWaaa
beaakiaac6caaaa@3DCC@
Notons que
X
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHybWaa0
baaSqaaiaadQgaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaaa
aa@3CDB@
indique quelles grappes
A
i
∈
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaaaaa@3D9D@
sont liées à cette personne. Soit
x = (
x
1
, … ,
x
n
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4bGaey
ypa0ZaaeWaaeaacaWG4bWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiab
lAciljaaiYcacaWG4bWaaSbaaSqaaiaad6gaaeqaaaGccaGLOaGaay
zkaaaaaa@4297@
un vecteur dont le
i
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbWaaW
baaSqabeaacaqGLbaaaaaa@3A68@
élément est
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaaIWaaaaa@391F@
ou
1 , i = 1 , … , n .
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaaIXaGaai
ilaiaadMgacqGH9aqpcaaIXaGaaiilaiablAciljaaiYcacaWGUbGa
aiOlaaaa@40AC@
En raison des hypothèses que
nous avons faites au sujet des lois des variables
X
i
j
(
k
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGybWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaaeilaaaa@3E7E@
la probabilité conditionnelle,
sachant
β
j
(
k
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHYoGyda
qhaaWcbaGaamOAaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaa
kiaacYcaaaa@3E55@
que
X
j
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHybWaa0
baaSqaaiaadQgaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaaa
aa@3CDB@
soit égale à
x
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4bGaai
ilaaaa@3A16@
c’est-à-dire la probabilité que
la
j
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGQbWaaW
baaSqabeaacaqGLbaaaaaa@3A69@
personne soit liée uniquement
aux grappes
A
i
∈
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaaaaa@3D9D@
de manière telle que le
i
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbWaaW
baaSqabeaacaqGLbaaaaaa@3A68@
élément
x
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWG4bWaaS
baaSqaaiaadMgaaeqaaaaa@3A7C@
de
x
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4baaaa@3966@
soit égal à 1, est
Pr (
X
j
(
k
)
= x |
β
j
(
k
)
,
S
A
) =
∏
i = 1
n
[
p
i j
(
k
)
]
x
i
[
1 −
p
i j
(
k
)
]
1 −
x
i
=
∏
i = 1
n
exp [
x
i
(
α
i
(
k
)
+
β
j
(
k
)
) ]
1 + exp (
α
i
(
k
)
+
β
j
(
k
)
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaciGGqbGaai
OCamaabmaabaGaaCiwamaaDaaaleaacaWGQbaabaWaaeWaaeaacaWG
RbaacaGLOaGaayzkaaaaaOGaeyypa0JaaCiEamaaeeaabaGaaGPaVl
abek7aInaaDaaaleaacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGa
ayzkaaaaaOGaaGilaiaadofadaWgaaWcbaGaamyqaaqabaaakiaawE
a7aaGaayjkaiaawMcaaiabg2da9maarahabeWcbaGaamyAaiabg2da
9iaaigdaaeaacaWGUbaaniabg+GivdGcdaWadaqaaiaadchadaqhaa
WcbaGaamyAaiaadQgaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaa
aaaakiaawUfacaGLDbaadaahaaWcbeqaaiaadIhadaWgaaadbaGaam
yAaaqabaaaaOWaamWaaeaacaaIXaGaeyOeI0IaamiCamaaDaaaleaa
caWGPbGaamOAaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaaaO
Gaay5waiaaw2faamaaCaaaleqabaGaaGymaiabgkHiTiaadIhadaWg
aaadbaGaamyAaaqabaaaaOGaeyypa0ZaaebCaeqaleaacaWGPbGaey
ypa0JaaGymaaqaaiaad6gaa0Gaey4dIunakmaalaaabaGaciyzaiaa
cIhacaGGWbWaamWaaeaacaWG4bWaaSbaaSqaaiaadMgaaeqaaOWaae
WaaeaacqaHXoqydaqhaaWcbaGaamyAaaqaamaabmaabaGaam4AaaGa
ayjkaiaawMcaaaaakiabgUcaRiabek7aInaaDaaaleaacaWGQbaaba
WaaeWaaeaacaWGRbaacaGLOaGaayzkaaaaaaGccaGLOaGaayzkaaaa
caGLBbGaayzxaaaabaGaaGymaiabgUcaRiGacwgacaGG4bGaaiiCam
aabmaabaGaeqySde2aa0baaSqaaiaadMgaaeaadaqadaqaaiaadUga
aiaawIcacaGLPaaaaaGccqGHRaWkcqaHYoGydaqhaaWcbaGaamOAaa
qaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaaaOGaayjkaiaawMca
aaaacaaIUaaaaa@97C5@
Donc,
la probabilité que le vecteur de variables indicatrices de lien associé à une personne
sélectionnée aléatoirement dans
U
k
−
S
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0Iaam4uamaaBaaaleaacaaIWaaa
beaaaaa@3D10@
soit égal à
x
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4baaaa@3966@
est
π
x
(
k
)
(
α
k
,
σ
k
) =
∫
∏
i = 1
n
exp [
x
i
(
α
i
(
k
)
+
σ
k
z
) ]
1 + exp (
α
i
(
k
)
+
σ
k
z
)
ϕ (
z
) d z ,
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
qhaaWcbaGaaCiEaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaa
kmaabmaabaGaaCySdmaaBaaaleaacaWGRbaabeaakiaaiYcacqaHdp
WCdaWgaaWcbaGaam4AaaqabaaakiaawIcacaGLPaaacqGH9aqpdaWd
baqabSqabeqaniabgUIiYdGcdaqeWbqabSqaaiaadMgacqGH9aqpca
aIXaaabaGaamOBaaqdcqGHpis1aOWaaSaaaeaaciGGLbGaaiiEaiaa
cchadaWadaqaaiaadIhadaWgaaWcbaGaamyAaaqabaGcdaqadaqaai
abeg7aHnaaDaaaleaacaWGPbaabaWaaeWaaeaacaWGRbaacaGLOaGa
ayzkaaaaaOGaey4kaSIaeq4Wdm3aaSbaaSqaaiaadUgaaeqaaOGaam
OEaaGaayjkaiaawMcaaaGaay5waiaaw2faaaqaaiaaigdacqGHRaWk
ciGGLbGaaiiEaiaacchadaqadaqaaiabeg7aHnaaDaaaleaacaWGPb
aabaWaaeWaaeaacaWGRbaacaGLOaGaayzkaaaaaOGaey4kaSIaeq4W
dm3aaSbaaSqaaiaadUgaaeqaaOGaamOEaaGaayjkaiaawMcaaaaacq
aHvpGzdaqadaqaaiaadQhaaiaawIcacaGLPaaacaWGKbGaamOEaiaa
iYcaaaa@77CB@
où
α
k
= (
α
1
(
k
)
, … ,
α
n
(
k
)
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHXoWaaS
baaSqaaiaadUgaaeqaaOGaeyypa0ZaaeWaaeaacqaHXoqydaqhaaWc
baGaaGymaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaakiaaiY
cacqWIMaYscaaISaGaeqySde2aa0baaSqaaiaad6gaaeaadaqadaqa
aiaadUgaaiaawIcacaGLPaaaaaaakiaawIcacaGLPaaaaaa@4A31@
et
ϕ ( ⋅ )
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHvpGzca
GGOaGaeyyXICTaaiykaaaa@3DD0@
désignent la densité
de probabilité de la loi normale centrée réduite
[
N
(
0
,
1
)
]
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaWadaqaai
aab6eadaqadaqaaiaaicdacaGGSaGaaGymaaGaayjkaiaawMcaaaGa
ay5waiaaw2faaiaac6caaaa@3F88@
Comme dans Coull et Agresti (1999), au lieu d’utiliser
π
x
(
k
)
(
α
k
,
σ
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
qhaaWcbaGaaCiEaaqaamaabmaabaGaam4AaaGaayjkaiaawMcaaaaa
kmaabmaabaGaaCySdmaaBaaaleaacaWGRbaabeaakiaaiYcacqaHdp
WCdaWgaaWcbaGaam4AaaqabaaakiaawIcacaGLPaaaaaa@455E@
dans la fonction de vraisemblance,
nous utilisons son approximation par quadrature gaussienne
π
˜
x
(
k
)
(
α
k
,
σ
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHapaCga
acamaaDaaaleaacaWH4baabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOWaaeWaaeaacaWHXoWaaSbaaSqaaiaadUgaaeqaaOGaaGilai
abeo8aZnaaBaaaleaacaWGRbaabeaaaOGaayjkaiaawMcaaaaa@456D@
donnée par
π
˜
x
(
k
)
(
α
k
,
σ
k
) =
∑
t = 1
q
∏
i = 1
n
exp [
x
i
(
α
i
(
k
)
+
σ
k
z
t
) ]
1 + exp (
α
i
(
k
)
+
σ
k
z
t
)
ν
t
, ( 3.3 )
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHapaCga
acamaaDaaaleaacaWH4baabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOWaaeWaaeaacaWHXoWaaSbaaSqaaiaadUgaaeqaaOGaaGilai
abeo8aZnaaBaaaleaacaWGRbaabeaaaOGaayjkaiaawMcaaiabg2da
9maaqahabeWcbaGaamiDaiabg2da9iaaigdaaeaacaWGXbaaniabgg
HiLdGcdaqeWbqabSqaaiaadMgacqGH9aqpcaaIXaaabaGaamOBaaqd
cqGHpis1aOWaaSaaaeaaciGGLbGaaiiEaiaacchadaWadaqaaiaadI
hadaWgaaWcbaGaamyAaaqabaGcdaqadaqaaiabeg7aHnaaDaaaleaa
caWGPbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzkaaaaaOGaey4kaS
Iaeq4Wdm3aaSbaaSqaaiaadUgaaeqaaOGaamOEamaaBaaaleaacaWG
0baabeaaaOGaayjkaiaawMcaaaGaay5waiaaw2faaaqaaiaaigdacq
GHRaWkciGGLbGaaiiEaiaacchadaqadaqaaiabeg7aHnaaDaaaleaa
caWGPbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzkaaaaaOGaey4kaS
Iaeq4Wdm3aaSbaaSqaaiaadUgaaeqaaOGaamOEamaaBaaaleaacaWG
0baabeaaaOGaayjkaiaawMcaaaaacqaH9oGBdaWgaaWcbaGaamiDaa
qabaGccaaISaGaaGzbVlaaywW7caaMf8UaaGzbVlaaywW7caGGOaGa
aG4maiaac6cacaaIZaGaaiykaaaa@8629@
où
q
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGXbaaaa@395B@
est une constante
fixée, et
{
z
t
}
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aadQhadaWgaaWcbaGaamiDaaqabaaakiaawUhacaGL9baaaaa@3CC4@
et les valeurs de
{
ν
t
}
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
abe27aUnaaBaaaleaacaWG0baabeaaaOGaay5Eaiaaw2haaaaa@3D7D@
sont tirées de
tables.
Nous pouvons maintenant calculer les
deux facteurs susmentionnés de la fonction de vraisemblance. Soit
Ω = {
(
x
1
, … ,
x
n
) :
x
i
= 0 , 1 ; i = 1 , … , n } ,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqqHPoWvcq
GH9aqpdaGadaqaamaabmaabaGaamiEamaaBaaaleaacaaIXaaabeaa
kiaaiYcacqWIMaYscaaISaGaamiEamaaBaaaleaacaWGUbaabeaaaO
GaayjkaiaawMcaaiaacQdacaWG4bWaaSbaaSqaaiaadMgaaeqaaOGa
eyypa0JaaGimaiaacYcacaaIXaGaai4oaiaadMgacqGH9aqpcaaIXa
GaaiilaiablAciljaaiYcacaWGUbaacaGL7bGaayzFaaGaaiilaaaa
@52F8@
l’ensemble de tous les vecteurs
de dimension
n
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbaaaa@3958@
tels que chacun de leurs éléments
est 0 ou 1. Pour
x = (
x
1
, … ,
x
n
) ∈ Ω ,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4bGaey
ypa0ZaaeWaaeaacaWG4bWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiab
lAciljaaiYcacaWG4bWaaSbaaSqaaiaad6gaaeqaaaGccaGLOaGaay
zkaaGaeyicI4SaeuyQdCLaaiilaaaa@4659@
soit
R
x
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaa0
baaSqaaiaahIhaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaaa
aa@3CE3@
la variable aléatoire qui indique
le nombre de personnes distinctes dans
U
k
−
S
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0Iaam4uamaaBaaaleaacaaIWaaa
beaaaaa@3D10@
dont les vecteurs de variables
indicatrices de lien sont égaux à
x
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4bGaai
Olaaaa@3A18@
Enfin, soit
R
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaaS
baaSqaaiaadUgaaeqaaaaa@3A58@
la variable aléatoire qui
indique le nombre de personnes distinctes dans
U
k
−
S
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0Iaam4uamaaBaaaleaacaaIWaaa
beaaaaa@3D10@
qui sont liées à au moins une
grappe
A
i
∈
S
A
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaakiaac6caaaa@3E59@
Notons que
R
k
=
∑
x ∈ Ω − { 0 }
R
x
(
k
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaaS
baaSqaaiaadUgaaeqaaOGaeyypa0ZaaabeaeaacaWGsbWaa0baaSqa
aiaahIhaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaaabaGaaC
iEaiabgIGiolabfM6axjabgkHiTiaabUhacaWHWaGaaeyFaaqab0Ga
eyyeIuoakiaacYcaaaa@4A2F@
où
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHWaaaaa@391E@
désigne le vecteur de
dimension
n
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGUbaaaa@3958@
de zéros.
En raison des hypothèses que nous avons
émises au sujet des lois des variables
X
i j
( k )
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGybWaa0
baaSqaaiaadMgacaWGQbaabaGaaiikaiaadUgacaGGPaaaaOGaaiil
aaaa@3E4F@
la loi de probabilité conjointe
conditionnelle des variables
{
R
x
( 1 )
}
x ∈ Ω − { 0 }
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aadkfadaqhaaWcbaGaaCiEaaqaaiaacIcacaaIXaGaaiykaaaaaOGa
ay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4SaeuyQdCLaeyOeI0
YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeaaaaa@46CF@
et
τ
1
−
m
−
R
1
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGymaaqabaGccqGHsislcaWGTbGaeyOeI0IaamOuamaa
BaaaleaacaaIXaaabeaakiaacYcaaaa@405F@
sachant que
{
M
i
=
m
i
}
i = 1
n
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aad2eadaWgaaWcbaGaamyAaaqabaGccqGH9aqpcaWGTbWaaSbaaSqa
aiaadMgaaeqaaaGccaGL7bGaayzFaaWaa0baaSqaaiaadMgacqGH9a
qpcaaIXaaabaGaamOBaaaakiaacYcaaaa@4431@
est une loi multinomiale de paramètre
de taille
τ
1
−
m
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGymaaqabaGccqGHsislcaWGTbaaaa@3CFA@
et de probabilités
{
π
x
(
1
)
(
α
1
,
σ
1
)
}
x
∈
Ω
−
{
0
}
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
abec8aWnaaDaaaleaacaWH4baabaWaaeWaaeaacaaIXaaacaGLOaGa
ayzkaaaaaOWaaeWaaeaacaWHXoWaaSbaaSqaaiaaigdaaeqaaOGaaG
ilaiabeo8aZnaaBaaaleaacaaIXaaabeaaaOGaayjkaiaawMcaaaGa
ay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4SaeuyQdCLaeyOeI0
YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeaaaaa@4F06@
et
π
0
(
1
)
(
α
1
,
σ
1
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
qhaaWcbaGaaCimaaqaamaabmaabaGaaGymaaGaayjkaiaawMcaaaaa
kmaabmaabaGaaCySdmaaBaaaleaacaaIXaaabeaakiaaiYcacqaHdp
WCdaWgaaWcbaGaaGymaaqabaaakiaawIcacaGLPaaacaGGSaaaaa@4527@
et celle des variables
{
R
x
(
2
)
}
x
∈
Ω
−
{
0
}
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aadkfadaqhaaWcbaGaaCiEaaqaamaabmaabaGaaGOmaaGaayjkaiaa
wMcaaaaaaOGaay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4Saeu
yQdCLaeyOeI0YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeaaaaa@4700@
et
τ
2
−
R
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGOmaaqabaGccqGHsislcaWGsbWaaSbaaSqaaiaaikda
aeqaaaaa@3DC8@
est une loi multinomiale de paramètre
de taille
τ
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGOmaaqabaaaaa@3B12@
et de probabilités
{
π
x
(
2
)
(
α
2
,
σ
2
)
}
x
∈
Ω
−
{
0
}
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
abec8aWnaaDaaaleaacaWH4baabaWaaeWaaeaacaaIYaaacaGLOaGa
ayzkaaaaaOWaaeWaaeaacaWHXoWaaSbaaSqaaiaaikdaaeqaaOGaaG
ilaiabeo8aZnaaBaaaleaacaaIYaaabeaaaOGaayjkaiaawMcaaaGa
ay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4SaeuyQdCLaeyOeI0
YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeaaaaa@4F09@
et
π
0
(
2
)
(
α
2
,
σ
2
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
qhaaWcbaGaaCimaaqaamaabmaabaGaaGOmaaGaayjkaiaawMcaaaaa
kmaabmaabaGaaCySdmaaBaaaleaacaaIYaaabeaakiaaiYcacqaHdp
WCdaWgaaWcbaGaaGOmaaqabaaakiaawIcacaGLPaaacaGGUaaaaa@452C@
Par conséquent, les facteurs associés aux
probabilités des configurations des liens entre les personnes dans
U
k
−
S
0
, k = 1 , 2 ,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGvbWaaS
baaSqaaiaadUgaaeqaaOGaeyOeI0Iaam4uamaaBaaaleaacaaIWaaa
beaakiaacYcacaWGRbGaeyypa0JaaGymaiaacYcacaaIYaGaaiilaa
aa@4297@
et les grappes
A
i
∈
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaaaaa@3D9D@
sont
L
1
(
τ
1
,
α
1
,
σ
1
) ∝
(
τ
1
− m
) !
(
τ
1
− m −
r
1
) !
∏
x ∈ Ω − { 0 }
[
π
˜
x
(
1
)
(
α
1
,
σ
1
) ]
r
x
(
1
)
[
π
˜
0
(
1
)
(
α
1
,
σ
1
) ]
τ
1
− m −
r
1
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaigdaaeqaaOWaaeWaaeaacqaHepaDdaWgaaWcbaGaaGym
aaqabaGccaaISaGaaCySdmaaBaaaleaacaaIXaaabeaakiaaiYcacq
aHdpWCdaWgaaWcbaGaaGymaaqabaaakiaawIcacaGLPaaacqGHDisT
daWcaaqaamaabmaabaGaeqiXdq3aaSbaaSqaaiaaigdaaeqaaOGaey
OeI0IaamyBaaGaayjkaiaawMcaaiaacgcaaeaadaqadaqaaiabes8a
0naaBaaaleaacaaIXaaabeaakiabgkHiTiaad2gacqGHsislcaWGYb
WaaSbaaSqaaiaaigdaaeqaaaGccaGLOaGaayzkaaGaaiyiaaaadaqe
qbqabSqaaiaahIhacqGHiiIZcqqHPoWvcqGHsisldaGadaqaaiaahc
daaiaawUhacaGL9baaaeqaniabg+GivdGcdaWadaqaaiqbec8aWzaa
iaWaa0baaSqaaiaahIhaaeaadaqadaqaaiaaigdaaiaawIcacaGLPa
aaaaGcdaqadaqaaiaahg7adaWgaaWcbaGaaGymaaqabaGccaaISaGa
eq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOaGaayzkaaaacaGLBb
GaayzxaaWaaWbaaSqabeaacaWGYbWaa0baaWqaaiaahIhaaeaadaqa
daqaaiaaigdaaiaawIcacaGLPaaaaaaaaOWaamWaaeaacuaHapaCga
acamaaDaaaleaacaWHWaaabaWaaeWaaeaacaaIXaaacaGLOaGaayzk
aaaaaOWaaeWaaeaacaWHXoWaaSbaaSqaaiaaigdaaeqaaOGaaGilai
abeo8aZnaaBaaaleaacaaIXaaabeaaaOGaayjkaiaawMcaaaGaay5w
aiaaw2faamaaCaaaleqabaGaeqiXdq3aaSbaaWqaaiaaigdaaeqaaS
GaeyOeI0IaamyBaiabgkHiTiaadkhadaWgaaadbaGaaGymaaqabaaa
aaaa@8944@
et
L
2
(
τ
2
,
α
2
,
σ
2
) ∝
τ
2
!
(
τ
2
−
r
2
) !
∏
x ∈ Ω − { 0 }
[
π
˜
x
(
2
)
(
α
2
,
σ
2
) ]
r
x
(
2
)
[
π
˜
0
(
2
)
(
α
2
,
σ
2
) ]
τ
2
−
r
2
.
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaikdaaeqaaOWaaeWaaeaacqaHepaDdaWgaaWcbaGaaGOm
aaqabaGccaaISaGaaCySdmaaBaaaleaacaaIYaaabeaakiaaiYcacq
aHdpWCdaWgaaWcbaGaaGOmaaqabaaakiaawIcacaGLPaaacqGHDisT
daWcaaqaaiabes8a0naaBaaaleaacaaIYaaabeaakiaacgcaaeaada
qadaqaaiabes8a0naaBaaaleaacaaIYaaabeaakiabgkHiTiaadkha
daWgaaWcbaGaaGOmaaqabaaakiaawIcacaGLPaaacaGGHaaaamaara
fabeWcbaGaaCiEaiabgIGiolabfM6axjabgkHiTmaacmaabaGaaCim
aaGaay5Eaiaaw2haaaqab0Gaey4dIunakmaadmaabaGafqiWdaNbaG
aadaqhaaWcbaGaaCiEaaqaamaabmaabaGaaGOmaaGaayjkaiaawMca
aaaakmaabmaabaGaaCySdmaaBaaaleaacaaIYaaabeaakiaaiYcacq
aHdpWCdaWgaaWcbaGaaGOmaaqabaaakiaawIcacaGLPaaaaiaawUfa
caGLDbaadaahaaWcbeqaaiaadkhadaqhaaadbaGaaCiEaaqaamaabm
aabaGaaGOmaaGaayjkaiaawMcaaaaaaaGcdaWadaqaaiqbec8aWzaa
iaWaa0baaSqaaiaahcdaaeaadaqadaqaaiaaikdaaiaawIcacaGLPa
aaaaGcdaqadaqaaiaahg7adaWgaaWcbaGaaGOmaaqabaGccaaISaGa
eq4Wdm3aaSbaaSqaaiaaikdaaeqaaaGccaGLOaGaayzkaaaacaGLBb
GaayzxaaWaaWbaaSqabeaacqaHepaDdaWgaaadbaGaaGOmaaqabaWc
cqGHsislcaWGYbWaaSbaaWqaaiaaikdaaeqaaaaakiaai6caaaa@82F0@
La dernière composante de la fonction
de vraisemblance est associée à la probabilité conditionnelle, sachant
S
A
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGtbWaaS
baaSqaaiaadgeaaeqaaOGaaiilaaaa@3AE9@
de la configuration des liens entre les personnes
dans
S
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGtbWaaS
baaSqaaiaaicdaaeqaaaaa@3A23@
et les grappes
A
i
∈
S
A
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaakiaac6caaaa@3E59@
Pour calculer ce facteur,
observons pour commencer que, par définition des variables indicatrices
X
i
j
(
k
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGybWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaaiilaaaa@3E7F@
le
i
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbWaaW
baaSqabeaacaqGLbaaaaaa@3A68@
élément du vecteur des variables indicatrices
de lien associé à une personne dans
A
i
∈
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaaaaa@3D9D@
est égal à zéro. Donc, soit
Ω
− i
= {
(
x
1
, … ,
x
i − 1
,
x
i + 1
, … ,
x
n
) :
x
j
= 0 , 1 , j ≠ i , j = 1 , … , n } ,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqqHPoWvda
WgaaWcbaGaeyOeI0IaamyAaaqabaGccqGH9aqpdaGadaqaamaabmaa
baGaamiEamaaBaaaleaacaaIXaaabeaakiaaiYcacqWIMaYscaaISa
GaamiEamaaBaaaleaacaWGPbGaeyOeI0IaaGymaaqabaGccaaISaGa
amiEamaaBaaaleaacaWGPbGaey4kaSIaaGymaaqabaGccaaISaGaeS
OjGSKaaGilaiaadIhadaWgaaWcbaGaamOBaaqabaaakiaawIcacaGL
PaaacaGG6aGaamiEamaaBaaaleaacaWGQbaabeaakiabg2da9iaaic
dacaGGSaGaaGymaiaacYcacaqGGaGaamOAaiabgcMi5kaadMgacaaI
SaGaaeiiaiaadQgacqGH9aqpcaaIXaGaaiilaiablAciljaaiYcaca
WGUbaacaGL7bGaayzFaaGaaiilaaaa@6567@
c’est-à-dire l’ensemble de tous
les vecteurs de dimension
(
n
−
1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aad6gacqGHsislcaaIXaaacaGLOaGaayzkaaaaaa@3C89@
obtenu à partir des vecteurs dans
Ω
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqqHPoWvaa
a@39F3@
en omettant leur
i
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbWaaW
baaSqabeaacaqGLbaaaaaa@3A68@
coordonnée. Pour
x = (
x
1
, … ,
x
i − 1
,
x
i + 1
, … ,
x
n
) ∈
Ω
− i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4bGaey
ypa0ZaaeWaaeaacaWG4bWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiab
lAciljaaiYcacaWG4bWaaSbaaSqaaiaadMgacqGHsislcaaIXaaabe
aakiaaiYcacaWG4bWaaSbaaSqaaiaadMgacqGHRaWkcaaIXaaabeaa
kiaaiYcacqWIMaYscaaISaGaamiEamaaBaaaleaacaWGUbaabeaaaO
GaayjkaiaawMcaaiabgIGiolabfM6axnaaBaaaleaacqGHsislcaWG
Pbaabeaaaaa@527B@
, soit
R
x
(
A
i
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaa0
baaSqaaiaahIhaaeaadaqadaqaaiaadgeadaWgaaadbaGaamyAaaqa
baaaliaawIcacaGLPaaaaaaaaa@3DDF@
la variable aléatoire qui indique
le nombre de personnes distinctes dans
A
i
∈
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaaaaa@3D9D@
dont les vecteurs de variables
indicatrices de lien, quand la
i
e
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGPbWaaW
baaSqabeaacaqGLbaaaaaa@3A68@
coordonnée est omise, sont
égaux à
x
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWH4bGaai
Olaaaa@3A18@
Enfin, soit
R
(
A
i
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaaW
baaSqabeaadaqadaqaaiaadgeadaWgaaadbaGaamyAaaqabaaaliaa
wIcacaGLPaaaaaaaaa@3CDE@
la variable aléatoire qui indique
le nombre de personnes distinctes dans
A
i
∈
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaaaaa@3D9D@
qui sont liées à au moins un
site
A
j
∈
S
A
,
j
≠
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadQgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaakiaacYcacaWGQbGaeyiyIKRaamyAaiaac6caaaa@42AE@
Notons que
R
(
A
i
)
=
∑
x ∈
Ω
− i
− { 0 }
R
x
(
A
i
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaaW
baaSqabeaadaqadaqaaiaadgeadaWgaaadbaGaamyAaaqabaaaliaa
wIcacaGLPaaaaaGccqGH9aqpdaaeqaqaaiaadkfadaqhaaWcbaGaaC
iEaaqaamaabmaabaGaamyqamaaBaaameaacaWGPbaabeaaaSGaayjk
aiaawMcaaaaaaeaacaWH4bGaeyicI4SaeuyQdC1aaSbaaWqaaiabgk
HiTiaadMgaaeqaaSGaeyOeI0YaaiWaaeaacaWHWaaacaGL7bGaayzF
aaaabeqdcqGHris5aOGaaiilaaaa@4FF7@
où
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHWaaaaa@391E@
désigne le vecteur de
dimension
(
n
−
1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aad6gacqGHsislcaaIXaaacaGLOaGaayzkaaaaaa@3C89@
de zéros. Alors, comme dans les
cas précédents, la loi de probabilité conjointe conditionnelle des variables
{
R
x
(
A
i
)
}
x
∈
Ω
−
i
−
{
0
}
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aadkfadaqhaaWcbaGaaCiEaaqaamaabmaabaGaamyqamaaBaaameaa
caWGPbaabeaaaSGaayjkaiaawMcaaaaaaOGaay5Eaiaaw2haamaaBa
aaleaacaWH4bGaeyicI4SaeuyQdC1aaSbaaWqaaiabgkHiTiaadMga
aeqaaSGaeyOeI0YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeaaaa
a@4A43@
et
m
i
−
R
(
A
i
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGTbWaaS
baaSqaaiaadMgaaeqaaOGaeyOeI0IaamOuamaaCaaaleqabaWaaeWa
aeaacaWGbbWaaSbaaWqaaiaadMgaaeqaaaWccaGLOaGaayzkaaaaaO
Gaaiilaaaa@409B@
sachant que
{
M
i
=
m
i
}
i = 1
n
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aad2eadaWgaaWcbaGaamyAaaqabaGccqGH9aqpcaWGTbWaaSbaaSqa
aiaadMgaaeqaaaGccaGL7bGaayzFaaWaa0baaSqaaiaadMgacqGH9a
qpcaaIXaaabaGaamOBaaaakiaacYcaaaa@4431@
est une loi multinomiale de paramètre
de taille
m
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGTbWaaS
baaSqaaiaadMgaaeqaaaaa@3A71@
et de probabilités
{
π
x
(
A
i
)
(
α
1
(
−
i
)
,
σ
1
)
}
x
∈
Ω
−
i
−
{
0
}
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
abec8aWnaaDaaaleaacaWH4baabaWaaeWaaeaacaWGbbWaaSbaaWqa
aiaadMgaaeqaaaWccaGLOaGaayzkaaaaaOWaaeWaaeaacaWHXoWaa0
baaSqaaiaaigdaaeaadaqadaqaaiabgkHiTiaadMgaaiaawIcacaGL
PaaaaaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOa
GaayzkaaaacaGL7bGaayzFaaWaaSbaaSqaaiaahIhacqGHiiIZcqqH
PoWvdaWgaaadbaGaeyOeI0IaamyAaaqabaWccqGHsisldaGadaqaai
aahcdaaiaawUhacaGL9baaaeqaaaaa@55AF@
et
π
0
(
A
i
)
(
α
1
(
−
i
)
,
σ
1
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
qhaaWcbaGaaCimaaqaamaabmaabaGaamyqamaaBaaameaacaWGPbaa
beaaaSGaayjkaiaawMcaaaaakmaabmaabaGaaCySdmaaDaaaleaaca
aIXaaabaWaaeWaaeaacqGHsislcaWGPbaacaGLOaGaayzkaaaaaOGa
aGilaiabeo8aZnaaBaaaleaacaaIXaaabeaaaOGaayjkaiaawMcaai
aacYcaaaa@49BD@
où
α
1
(
− i
)
= (
α
1
(
1
)
, … ,
α
i − 1
(
1
)
,
α
i + 1
(
1
)
, … ,
α
n
(
1
)
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHXoWaa0
baaSqaaiaaigdaaeaadaqadaqaaiabgkHiTiaadMgaaiaawIcacaGL
PaaaaaGccqGH9aqpdaqadaqaaiabeg7aHnaaDaaaleaacaaIXaaaba
WaaeWaaeaacaaIXaaacaGLOaGaayzkaaaaaOGaaGilaiablAciljaa
iYcacqaHXoqydaqhaaWcbaGaamyAaiabgkHiTiaaigdaaeaadaqada
qaaiaaigdaaiaawIcacaGLPaaaaaGccaaISaGaeqySde2aa0baaSqa
aiaadMgacqGHRaWkcaaIXaaabaWaaeWaaeaacaaIXaaacaGLOaGaay
zkaaaaaOGaaGilaiablAciljaaiYcacqaHXoqydaqhaaWcbaGaamOB
aaqaamaabmaabaGaaGymaaGaayjkaiaawMcaaaaaaOGaayjkaiaawM
caaaaa@5D90@
et
π
x
(
A
i
)
(
α
1
(
− i
)
,
σ
1
) =
∫
∏
j ≠ i
n
exp [
x
j
(
α
j
(
1
)
+
σ
1
z
) ]
1 + exp (
α
j
(
1
)
+
σ
1
z
)
ϕ (
z
) d z .
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
qhaaWcbaGaaCiEaaqaamaabmaabaGaamyqamaaBaaameaacaWGPbaa
beaaaSGaayjkaiaawMcaaaaakmaabmaabaGaaCySdmaaDaaaleaaca
aIXaaabaWaaeWaaeaacqGHsislcaWGPbaacaGLOaGaayzkaaaaaOGa
aGilaiabeo8aZnaaBaaaleaacaaIXaaabeaaaOGaayjkaiaawMcaai
abg2da9maapeaabeWcbeqab0Gaey4kIipakmaarahabeWcbaGaamOA
aiabgcMi5kaadMgaaeaacaWGUbaaniabg+GivdGcdaWcaaqaaiGacw
gacaGG4bGaaiiCamaadmaabaGaamiEamaaBaaaleaacaWGQbaabeaa
kmaabmaabaGaeqySde2aa0baaSqaaiaadQgaaeaadaqadaqaaiaaig
daaiaawIcacaGLPaaaaaGccqGHRaWkcqaHdpWCdaWgaaWcbaGaaGym
aaqabaGccaWG6baacaGLOaGaayzkaaaacaGLBbGaayzxaaaabaGaaG
ymaiabgUcaRiGacwgacaGG4bGaaiiCamaabmaabaGaeqySde2aa0ba
aSqaaiaadQgaaeaadaqadaqaaiaaigdaaiaawIcacaGLPaaaaaGccq
GHRaWkcqaHdpWCdaWgaaWcbaGaaGymaaqabaGccaWG6baacaGLOaGa
ayzkaaaaaiabew9aMnaabmaabaGaamOEaaGaayjkaiaawMcaaiaads
gacaWG6bGaaGOlaaaa@7BE8@
Par conséquent, la probabilité de la configuration
des liens entre les personnes dans
S
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGtbWaaS
baaSqaaiaaicdaaeqaaaaa@3A23@
et les grappes
A
j
∈
S
A
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadQgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaaaaa@3D9E@
est donnée par le produit des
probabilités multinomiales précédentes (une pour chaque
A
i
∈
S
A
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGbbWaaS
baaSqaaiaadMgaaeqaaOGaeyicI4Saam4uamaaBaaaleaacaWGbbaa
beaakiaacMcacaGGSaaaaa@3F04@
et conséquemment le facteur de
la vraisemblance qui est associé à cette probabilité est
L
0
(
α
1
,
σ
1
) ∝
∏
i = 1
n
∏
x ∈
Ω
− i
− { 0 }
[
π
˜
x
(
A
i
)
(
α
1
(
− i
)
,
σ
1
) ]
r
x
(
A
i
)
[
π
˜
0
(
A
i
)
(
α
1
(
− i
)
,
σ
1
) ]
m
i
−
r
(
A
i
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaicdaaeqaaOWaaeWaaeaacaWHXoWaaSbaaSqaaiaaigda
aeqaaOGaaGilaiabeo8aZnaaBaaaleaacaaIXaaabeaaaOGaayjkai
aawMcaaiabg2Hi1oaarahabeWcbaGaamyAaiabg2da9iaaigdaaeaa
caWGUbaaniabg+GivdGcdaqeqbqabSqaaiaahIhacqGHiiIZcqqHPo
WvdaWgaaadbaGaeyOeI0IaamyAaaqabaWccqGHsisldaGadaqaaiaa
hcdaaiaawUhacaGL9baaaeqaniabg+GivdGcdaWadaqaaiqbec8aWz
aaiaWaa0baaSqaaiaahIhaaeaadaqadaqaaiaadgeadaWgaaadbaGa
amyAaaqabaaaliaawIcacaGLPaaaaaGcdaqadaqaaiaahg7adaqhaa
WcbaGaaGymaaqaamaabmaabaGaeyOeI0IaamyAaaGaayjkaiaawMca
aaaakiaaiYcacqaHdpWCdaWgaaWcbaGaaGymaaqabaaakiaawIcaca
GLPaaaaiaawUfacaGLDbaadaahaaWcbeqaaiaadkhadaqhaaadbaGa
aCiEaaqaamaabmaabaGaamyqamaaBaaabaGaamyAaaqabaaacaGLOa
GaayzkaaaaaaaakmaadmaabaGafqiWdaNbaGaadaqhaaWcbaGaaCim
aaqaamaabmaabaGaamyqamaaBaaameaacaWGPbaabeaaaSGaayjkai
aawMcaaaaakmaabmaabaGaaCySdmaaDaaaleaacaaIXaaabaWaaeWa
aeaacqGHsislcaWGPbaacaGLOaGaayzkaaaaaOGaaGilaiabeo8aZn
aaBaaaleaacaaIXaaabeaaaOGaayjkaiaawMcaaaGaay5waiaaw2fa
amaaCaaaleqabaGaamyBamaaBaaameaacaWGPbaabeaaliabgkHiTi
aadkhadaahaaadbeqaamaabmaabaGaamyqamaaBaaabaGaamyAaaqa
baaacaGLOaGaayzkaaaaaaaakiaacYcaaaa@8882@
où
π
˜
x
(
A
i
)
(
α
1
(
− i
)
,
σ
1
) =
∑
t = 1
q
∏
j ≠ i
n
exp [
x
j
(
α
j
(
1
)
+
σ
1
z
t
) ]
1 + exp (
α
j
(
1
)
+
σ
1
z
t
)
ν
t
, ( 3.4 )
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHapaCga
acamaaDaaaleaacaWH4baabaWaaeWaaeaacaWGbbWaaSbaaWqaaiaa
dMgaaeqaaaWccaGLOaGaayzkaaaaaOWaaeWaaeaacaWHXoWaa0baaS
qaaiaaigdaaeaadaqadaqaaiabgkHiTiaadMgaaiaawIcacaGLPaaa
aaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOaGaay
zkaaGaeyypa0ZaaabCaeqaleaacaWG0bGaeyypa0JaaGymaaqaaiaa
dghaa0GaeyyeIuoakmaarahabeWcbaGaamOAaiabgcMi5kaadMgaae
aacaWGUbaaniabg+GivdGcdaWcaaqaaiGacwgacaGG4bGaaiiCamaa
dmaabaGaamiEamaaBaaaleaacaWGQbaabeaakmaabmaabaGaeqySde
2aa0baaSqaaiaadQgaaeaadaqadaqaaiaaigdaaiaawIcacaGLPaaa
aaGccqGHRaWkcqaHdpWCdaWgaaWcbaGaaGymaaqabaGccaWG6bWaaS
baaSqaaiaadshaaeqaaaGccaGLOaGaayzkaaaacaGLBbGaayzxaaaa
baGaaGymaiabgUcaRiGacwgacaGG4bGaaiiCamaabmaabaGaeqySde
2aa0baaSqaaiaadQgaaeaadaqadaqaaiaaigdaaiaawIcacaGLPaaa
aaGccqGHRaWkcqaHdpWCdaWgaaWcbaGaaGymaaqabaGccaWG6bWaaS
baaSqaaiaadshaaeqaaaGccaGLOaGaayzkaaaaaiabe27aUnaaBaaa
leaacaWG0baabeaakiaaiYcacaaMf8UaaGzbVlaaywW7caaMf8UaaG
zbVlaacIcacaaIZaGaaiOlaiaaisdacaGGPaaaaa@8A45@
est l’approximation
par quadrature gaussienne de la probabilité
π
x
(
A
i
)
(
α
1
(
−
i
)
,
σ
1
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHapaCda
qhaaWcbaGaaCiEaaqaamaabmaabaGaamyqamaaBaaameaacaWGPbaa
beaaaSGaayjkaiaawMcaaaaakmaabmaabaGaaCySdmaaDaaaleaaca
aIXaaabaWaaeWaaeaacqGHsislcaWGPbaacaGLOaGaayzkaaaaaOGa
aGilaiabeo8aZnaaBaaaleaacaaIXaaabeaaaOGaayjkaiaawMcaai
aac6caaaa@4A07@
Des résultats
qui précèdent, il découle que la fonction de vraisemblance est donnée par
L (
τ
1
,
τ
2
,
α
1
,
α
2
,
σ
1
,
σ
2
) =
L
(
1
)
(
τ
1
,
α
1
,
σ
1
)
L
(
2
)
(
τ
2
,
α
2
,
σ
2
) ,
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaae
WaaeaacqaHepaDdaWgaaWcbaGaaGymaaqabaGccaaISaGaeqiXdq3a
aSbaaSqaaiaaikdaaeqaaOGaaGilaiabeg7aHnaaBaaaleaacaaIXa
aabeaakiaaiYcacqaHXoqydaWgaaWcbaGaaGOmaaqabaGccaaISaGa
eq4Wdm3aaSbaaSqaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBaaale
aacaaIYaaabeaaaOGaayjkaiaawMcaaiabg2da9iaadYeadaWgaaWc
baWaaeWaaeaacaaIXaaacaGLOaGaayzkaaaabeaakmaabmaabaGaeq
iXdq3aaSbaaSqaaiaaigdaaeqaaOGaaGilaiaahg7adaWgaaWcbaGa
aGymaaqabaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGcca
GLOaGaayzkaaGaamitamaaBaaaleaadaqadaqaaiaaikdaaiaawIca
caGLPaaaaeqaaOWaaeWaaeaacqaHepaDdaWgaaWcbaGaaGOmaaqaba
GccaaISaGaaCySdmaaBaaaleaacaaIYaaabeaakiaaiYcacqaHdpWC
daWgaaWcbaGaaGOmaaqabaaakiaawIcacaGLPaaacaaISaaaaa@6BB3@
où
L
(
1
)
(
τ
1
,
α
1
,
σ
1
) =
L
MULT
(
τ
1
)
L
1
(
τ
1
,
α
1
,
σ
1
)
L
0
(
α
1
,
σ
1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaamaabmaabaGaaGymaaGaayjkaiaawMcaaaqabaGcdaqadaqa
aiabes8a0naaBaaaleaacaaIXaaabeaakiaaiYcacaWHXoWaaSbaaS
qaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBaaaleaacaaIXaaabeaa
aOGaayjkaiaawMcaaiabg2da9iaadYeadaWgaaWcbaGaaeytaiaabw
facaqGmbGaaeivaaqabaGcdaqadaqaaiabes8a0naaBaaaleaacaaI
XaaabeaaaOGaayjkaiaawMcaaiaadYeadaWgaaWcbaGaaGymaaqaba
Gcdaqadaqaaiabes8a0naaBaaaleaacaaIXaaabeaakiaaiYcacaWH
XoWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBaaaleaaca
aIXaaabeaaaOGaayjkaiaawMcaaiaadYeadaWgaaWcbaGaaGimaaqa
baGcdaqadaqaaiaahg7adaWgaaWcbaGaaGymaaqabaGccaaISaGaeq
4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOaGaayzkaaaaaa@6507@
et
L
(
2
)
(
τ
2
,
α
2
,
σ
2
) =
L
2
(
τ
2
,
α
2
,
σ
2
) .
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaamaabmaabaGaaGOmaaGaayjkaiaawMcaaaqabaGcdaqadaqa
aiabes8a0naaBaaaleaacaaIYaaabeaakiaaiYcacaWHXoWaaSbaaS
qaaiaaikdaaeqaaOGaaGilaiabeo8aZnaaBaaaleaacaaIYaaabeaa
aOGaayjkaiaawMcaaiabg2da9iaadYeadaWgaaWcbaGaaGOmaaqaba
Gcdaqadaqaaiabes8a0naaBaaaleaacaaIYaaabeaakiaaiYcacaWH
XoWaaSbaaSqaaiaaikdaaeqaaOGaaGilaiabeo8aZnaaBaaaleaaca
aIYaaabeaaaOGaayjkaiaawMcaaiaai6caaaa@5451@
Dans les commentaires à la fin de la
sous-section 3.1, nous avons indiqué que la fonction de vraisemblance dépend
des
M
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaadMgaaeqaaaaa@3A51@
essentiellement par la voie
de leur somme
M
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbGaai
Olaaaa@39E9@
On peut le constater en
remarquant que seul le facteur
L
0
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaicdaaeqaaaaa@3A1C@
dépend directement des
M
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaadMgaaeqaaOGaaeOlaaaa@3B0C@
Les facteurs
L
MULT
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaab2eacaqGvbGaaeitaiaabsfaaeqaaaaa@3CB0@
et
L
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaigdaaeqaaaaa@3A1D@
dépendent des
M
i
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaadMgaaeqaaaaa@3A51@
par la voie de
M
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbGaai
ilaaaa@39E7@
tandis que le facteur
L
(
2
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaamaabmaabaGaaGOmaaGaayjkaiaawMcaaaqabaaaaa@3BA7@
ne dépend pas des
M
i
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGnbWaaS
baaSqaaiaadMgaaeqaaOGaaiOlaaaa@3B0D@
3.3 Estimateurs
du maximum de vraisemblance inconditionnel
La maximisation numérique de la fonction
de vraisemblance
L
(
τ
1
,
τ
2
,
α
1
,
α
2
,
σ
1
,
σ
2
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaae
WaaeaacqaHepaDdaWgaaWcbaGaaGymaaqabaGccaaISaGaeqiXdq3a
aSbaaSqaaiaaikdaaeqaaOGaaGilaiabeg7aHnaaBaaaleaacaaIXa
aabeaakiaaiYcacqaHXoqydaWgaaWcbaGaaGOmaaqabaGccaaISaGa
eq4Wdm3aaSbaaSqaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBaaale
aacaaIYaaabeaaaOGaayjkaiaawMcaaaaa@4E44@
par rapport aux paramètres donne
les estimateurs du maximum de vraisemblance ordinaire ou inconditionnel (EMVI)
τ
^
k
(
U
)
,
α
^
k
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaOGaaiilaiqbeg7aHzaajaWaa0baaSqaaiaadUgaaeaadaqada
qaaiaadwfaaiaawIcacaGLPaaaaaaaaa@43A3@
et
σ
^
k
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHdpWCga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaaaa@3DB8@
de
τ
k
,
α
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaam4AaaqabaGccaGGSaGaeqySde2aaSbaaSqaaiaadUga
aeqaaaaa@3EBB@
et
σ
k
, k = 1 , 2.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHdpWCda
WgaaWcbaGaam4AaaqabaGccaGGSaGaam4Aaiabg2da9iaaigdacaGG
SaGaaGOmaiaac6caaaa@40CD@
Par conséquent, l’EMVI de
τ =
τ
1
+
τ
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDcq
GH9aqpcqaHepaDdaWgaaWcbaGaaGymaaqabaGccqGHRaWkcqaHepaD
daWgaaWcbaGaaGOmaaqabaaaaa@4175@
est
τ
^
(
U
)
=
τ
^
1
(
U
)
+
τ
^
2
(
U
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaCaaaleqabaWaaeWaaeaacaWGvbaacaGLOaGaayzkaaaaaOGa
eyypa0JafqiXdqNbaKaadaqhaaWcbaGaaGymaaqaamaabmaabaGaam
yvaaGaayjkaiaawMcaaaaakiabgUcaRiqbes8a0zaajaWaa0baaSqa
aiaaikdaaeaadaqadaqaaiaadwfaaiaawIcacaGLPaaaaaGccaGGUa
aaaa@49C3@
Il n’existe pas d'expressions analytiques pour les EMVI ; cependant, en utilisant l’approximation
asymptotique
∂ ln (
τ
k
!
) /
∂
τ
k
≈ ln (
τ
k
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaWcgaqaai
abgkGi2kGacYgacaGGUbWaaeWaaeaacqaHepaDdaWgaaWcbaGaam4A
aaqabaGccaGGHaaacaGLOaGaayzkaaaabaGaeyOaIyRaeqiXdq3aaS
baaSqaaiaadUgaaeqaaOGaeyisISRaciiBaiaac6gadaqadaqaaiab
es8a0naaBaaaleaacaWGRbaabeaaaOGaayjkaiaawMcaaaaacaGGSa
aaaa@4DE8@
nous obtenons les
approximations suivantes de
τ
^
1
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIXaaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaaaa@3D85@
et
τ
^
2
(
U
)
:
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIYaaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaOGaaiOoaaaa@3E4E@
τ
^
1
(
U
)
=
M +
R
1
1 − (
1 − n / N
)
π
˜
0
(
1
)
(
α
^
1
(
U
)
,
σ
^
1
(
U
)
)
et
τ
^
2
(
U
)
=
R
2
1 −
π
˜
0
(
2
)
(
α
^
2
(
U
)
,
σ
^
2
(
U
)
)
. ( 3.5 )
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIXaaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaOGaeyypa0ZaaSaaaeaacaWGnbGaey4kaSIaamOuamaaBaaale
aacaaIXaaabeaaaOqaaiaaigdacqGHsisldaqadaqaamaalyaabaGa
aGymaiabgkHiTiaad6gaaeaacaWGobaaaaGaayjkaiaawMcaaiqbec
8aWzaaiaWaa0baaSqaaiaahcdaaeaadaqadaqaaiaaigdaaiaawIca
caGLPaaaaaGcdaqadaqaaiqahg7agaqcamaaDaaaleaacaaIXaaaba
WaaeWaaeaacaWGvbaacaGLOaGaayzkaaaaaOGaaGilaiqbeo8aZzaa
jaWaa0baaSqaaiaaigdaaeaadaqadaqaaiaadwfaaiaawIcacaGLPa
aaaaaakiaawIcacaGLPaaaaaGaaeiiaiaabccacaqGGaGaaeyzaiaa
bshacaqGGaGaaeiiaiaabccacuaHepaDgaqcamaaDaaaleaacaaIYa
aabaWaaeWaaeaacaWGvbaacaGLOaGaayzkaaaaaOGaeyypa0ZaaSaa
aeaacaWGsbWaaSbaaSqaaiaaikdaaeqaaaGcbaGaaGymaiabgkHiTi
qbec8aWzaaiaWaa0baaSqaaiaahcdaaeaadaqadaqaaiaaikdaaiaa
wIcacaGLPaaaaaGcdaqadaqaaiqahg7agaqcamaaDaaaleaacaaIYa
aabaWaaeWaaeaacaWGvbaacaGLOaGaayzkaaaaaOGaaGilaiqbeo8a
ZzaajaWaa0baaSqaaiaaikdaaeaadaqadaqaaiaadwfaaiaawIcaca
GLPaaaaaaakiaawIcacaGLPaaaaaGaaGOlaiaaywW7caaMf8UaaGzb
VlaaywW7caGGOaGaaG4maiaac6cacaaI1aGaaiykaaaa@84C4@
Notons
qu’il ne s’agit pas d’expressions analytiques, puisque
α
^
k
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHXoqyga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaaaa@3D94@
et
σ
^
k
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHdpWCga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaaaa@3DB8@
dépendent de
τ
^
k
(
U
)
, k = 1 , 2.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaOGaaiilaiaabccacaWGRbGaeyypa0JaaGymaiaacYcacaaIYa
GaaiOlaaaa@43E6@
Néanmoins, ces
expressions sont utiles pour obtenir les formules des variances asymptotiques de
τ
^
1
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIXaaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaaaa@3D85@
et
τ
^
2
(
U
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIYaaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaOGaaiOlaaaa@3E42@
3.4 Estimateurs
du maximum de vraisemblance conditionnel
Un autre moyen d’obtenir les EMV de
τ
k
,
α
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaam4AaaqabaGccaGGSaGaeqySde2aaSbaaSqaaiaadUga
aeqaaaaa@3EBB@
et
σ
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHdpWCda
WgaaWcbaGaam4Aaaqabaaaaa@3B44@
consiste à suivre l’approche
de Sanathanan (1972), qui donne des estimateurs du maximum de vraisemblance conditionnel (EMVC). Ces estimateurs sont
numériquement plus simples à calculer que les EMVI . En outre, si l’on utilise
des covariables dans le modèle pour la probabilité de lien
p
i
j
(
k
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGWbWaa0
baaSqaaiaadMgacaWGQbaabaWaaeWaaeaacaWGRbaacaGLOaGaayzk
aaaaaOGaaiilaaaa@3E97@
cette approche pourrait
encore être utilisée pour obtenir des estimateurs de
τ
k
,
α
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaam4AaaqabaGccaGGSaGaeqySde2aaSbaaSqaaiaadUga
aeqaaaaa@3EBB@
et
σ
k
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHdpWCda
WgaaWcbaGaam4AaaqabaGccaGGSaaaaa@3BFE@
tandis que l’approche de la
vraisemblance inconditionnelle ne pourrait pas l’être, puisque les valeurs des covariables
associées aux éléments non échantillonnés seraient inconnues.
L’idée qui sous-tend l’approche de Sanathanan est de factoriser la fmp des lois multinomiales des fréquences
R
x
(
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaa0
baaSqaaiaahIhaaeaadaqadaqaaiaadUgaaiaawIcacaGLPaaaaaaa
aa@3CE3@
des différentes
configurations des liens comme il suit :
L
1
(
τ
1
,
α
1
,
σ
1
)
∝
f (
{
r
x
(
1
)
}
x ∈ Ω − { 0 }
,
τ
1
− m −
r
1
| {
m
i
} ,
τ
1
,
α
1
,
σ
1
)
=
f (
{
r
x
(
1
)
}
x ∈ Ω − { 0 }
| {
m
i
} ,
τ
1
,
r
1
,
α
1
,
σ
1
) f (
r
1
|
{
m
i
} ,
τ
1
,
α
1
,
σ
1
)
∝
∏
x ∈ Ω − { 0 }
[
π
˜
x
(
1
)
(
α
1
,
σ
1
)
1 −
π
˜
0
(
1
)
(
α
1
,
σ
1
)
]
r
x
(
1
)
×
(
τ
1
− m
) !
(
τ
1
− m −
r
1
) !
[
1 −
π
˜
0
(
1
)
(
α
1
,
σ
1
) ]
r
1
[
π
˜
0
(
1
)
(
α
1
,
σ
1
) ]
τ
1
− m −
r
1
=
L
11
(
α
1
,
σ
1
)
L
12
(
τ
1
,
α
1
,
σ
1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrFfpC0xe9LqFf0xe9
vqaqFeFr0xbba9Fa0P0RWFb9fq0FXxbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaqbaeaabsWaaa
aabaGaamitamaaBaaaleaacaaIXaaabeaakmaabmaabaGaeqiXdq3a
aSbaaSqaaiaaigdaaeqaaOGaaGilaiaahg7adaWgaaWcbaGaaGymaa
qabaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOaGa
ayzkaaaabaGaeyyhIulabaGaamOzamaabmaabaWaaqGaaeaadaGada
qaaiaadkhadaqhaaWcbaGaaCiEaaqaamaabmaabaGaaGymaaGaayjk
aiaawMcaaaaaaOGaay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4
SaeuyQdCLaeyOeI0YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeaa
kiaaiYcacqaHepaDdaWgaaWcbaGaaGymaaqabaGccqGHsislcaWGTb
GaeyOeI0IaamOCamaaBaaaleaacaaIXaaabeaaaOGaayjcSdWaaiWa
aeaacaWGTbWaaSbaaSqaaiaadMgaaeqaaaGccaGL7bGaayzFaaGaaG
ilaiabes8a0naaBaaaleaacaaIXaaabeaakiaaiYcacaWHXoWaaSba
aSqaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBaaaleaacaaIXaaabe
aaaOGaayjkaiaawMcaaaqaaaqaaiabg2da9aqaaiaadAgadaqadaqa
amaaeiaabaWaaiWaaeaacaWGYbWaa0baaSqaaiaahIhaaeaadaqada
qaaiaaigdaaiaawIcacaGLPaaaaaaakiaawUhacaGL9baadaWgaaWc
baGaaCiEaiabgIGiolabfM6axjabgkHiTmaacmaabaGaaCimaaGaay
5Eaiaaw2haaaqabaaakiaawIa7amaacmaabaGaamyBamaaBaaaleaa
caWGPbaabeaaaOGaay5Eaiaaw2haaiaaiYcacqaHepaDdaWgaaWcba
GaaGymaaqabaGccaaISaGaamOCamaaBaaaleaacaaIXaaabeaakiaa
iYcacaWHXoWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBa
aaleaacaaIXaaabeaaaOGaayjkaiaawMcaaiaadAgadaqadaqaaiaa
dkhadaWgaaWcbaGaaGymaaqabaGcdaabbaqaamaacmaabaGaamyBam
aaBaaaleaacaWGPbaabeaaaOGaay5Eaiaaw2haaiaaiYcacqaHepaD
daWgaaWcbaGaaGymaaqabaGccaaISaGaaCySdmaaBaaaleaacaaIXa
aabeaakiaaiYcacqaHdpWCdaWgaaWcbaGaaGymaaqabaaakiaawEa7
aaGaayjkaiaawMcaaaqaaaqaaiabg2Hi1cqaamaarafabeWcbaGaaC
iEaiabgIGiolabfM6axjabgkHiTmaacmaabaGaaCimaaGaay5Eaiaa
w2haaaqab0Gaey4dIunakmaadmaabaWaaSaaaeaacuaHapaCgaacam
aaDaaaleaacaWH4baabaWaaeWaaeaacaaIXaaacaGLOaGaayzkaaaa
aOWaaeWaaeaacaWHXoWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiabeo
8aZnaaBaaaleaacaaIXaaabeaaaOGaayjkaiaawMcaaaqaaiaaigda
cqGHsislcuaHapaCgaacamaaDaaaleaacaWHWaaabaWaaeWaaeaaca
aIXaaacaGLOaGaayzkaaaaaOWaaeWaaeaacaWHXoWaaSbaaSqaaiaa
igdaaeqaaOGaaGilaiabeo8aZnaaBaaaleaacaaIXaaabeaaaOGaay
jkaiaawMcaaaaaaiaawUfacaGLDbaadaahaaWcbeqaaiaadkhadaqh
aaadbaGaaCiEaaqaamaabmaabaGaaGymaaGaayjkaiaawMcaaaaaaa
GccqGHxdaTdaWcaaqaamaabmaabaGaeqiXdq3aaSbaaSqaaiaaigda
aeqaaOGaeyOeI0IaamyBaaGaayjkaiaawMcaaiaacgcaaeaadaqada
qaaiabes8a0naaBaaaleaacaaIXaaabeaakiabgkHiTiaad2gacqGH
sislcaWGYbWaaSbaaSqaaiaaigdaaeqaaaGccaGLOaGaayzkaaGaai
yiaaaadaWadaqaaiaaigdacqGHsislcuaHapaCgaacamaaDaaaleaa
caWHWaaabaWaaeWaaeaacaaIXaaacaGLOaGaayzkaaaaaOWaaeWaae
aacaWHXoWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBaaa
leaacaaIXaaabeaaaOGaayjkaiaawMcaaaGaay5waiaaw2faamaaCa
aaleqabaGaamOCamaaBaaameaacaaIXaaabeaaaaGcdaWadaqaaiqb
ec8aWzaaiaWaa0baaSqaaiaahcdaaeaadaqadaqaaiaaigdaaiaawI
cacaGLPaaaaaGcdaqadaqaaiaahg7adaWgaaWcbaGaaGymaaqabaGc
caaISaGaeq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOaGaayzkaa
aacaGLBbGaayzxaaWaaWbaaSqabeaacqaHepaDdaWgaaadbaGaaGym
aaqabaWccqGHsislcaWGTbGaeyOeI0IaamOCamaaBaaameaacaaIXa
aabeaaaaaakeaaaeaacqGH9aqpaeaacaWGmbWaaSbaaSqaaiaaigda
caaIXaaabeaakmaabmaabaGaaCySdmaaBaaaleaacaaIXaaabeaaki
aaiYcacqaHdpWCdaWgaaWcbaGaaGymaaqabaaakiaawIcacaGLPaaa
caWGmbWaaSbaaSqaaiaaigdacaaIYaaabeaakmaabmaabaGaeqiXdq
3aaSbaaSqaaiaaigdaaeqaaOGaaGilaiaahg7adaWgaaWcbaGaaGym
aaqabaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOa
Gaayzkaaaaaaaa@23A6@
et
L
2
(
τ
2
,
α
2
,
σ
2
)
∝
f (
{
r
x
(
2
)
}
x ∈ Ω − { 0 }
,
τ
2
−
r
2
| {
m
i
} ,
τ
2
,
α
2
,
σ
2
)
=
f (
{
r
x
(
2
)
}
x ∈ Ω − { 0 }
| {
m
i
} ,
τ
2
,
r
2
,
α
2
,
σ
2
) f (
r
2
|
{
m
i
} ,
τ
2
,
α
2
,
σ
2
)
∝
∏
x ∈ Ω − { 0 }
[
π
˜
x
(
2
)
(
α
2
,
σ
2
)
1 −
π
˜
0
(
2
)
(
α
2
,
σ
2
)
]
r
x
(
2
)
×
τ
2
!
(
τ
2
−
r
2
) !
[
1 −
π
˜
0
(
2
)
(
α
2
,
σ
2
) ]
r
2
[
π
˜
0
(
2
)
(
α
2
,
σ
2
) ]
τ
2
−
r
2
=
L
21
(
α
2
,
σ
2
)
L
22
(
τ
2
,
α
2
,
σ
2
) .
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xe9LqFf0xe9
vqaqFeFr0xbba9Fa0P0RWFb9fq0FXxbbf9=e0dfrpm0dXdirVu0=vr
0=vr0=fdbaqaaeGacaGaaiaabeqaamaabaabaaGcbaqbaeaabsWaaa
aabaGaamitamaaBaaaleaacaaIYaaabeaakmaabmaabaGaeqiXdq3a
aSbaaSqaaiaaikdaaeqaaOGaaGilaiaahg7adaWgaaWcbaGaaGOmaa
qabaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaikdaaeqaaaGccaGLOaGa
ayzkaaaabaGaeyyhIulabaGaamOzamaabmaabaWaaqGaaeaadaGada
qaaiaadkhadaqhaaWcbaGaaCiEaaqaamaabmaabaGaaGOmaaGaayjk
aiaawMcaaaaaaOGaay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4
SaeuyQdCLaeyOeI0Iaae4EaiaahcdacaqG9baabeaakiaaiYcacqaH
epaDdaWgaaWcbaGaaGOmaaqabaGccqGHsislcaWGYbWaaSbaaSqaai
aaikdaaeqaaaGccaGLiWoadaGadaqaaiaad2gadaWgaaWcbaGaamyA
aaqabaaakiaawUhacaGL9baacaaISaGaeqiXdq3aaSbaaSqaaiaaik
daaeqaaOGaaGilaiaahg7adaWgaaWcbaGaaGOmaaqabaGccaaISaGa
eq4Wdm3aaSbaaSqaaiaaikdaaeqaaaGccaGLOaGaayzkaaaabaaaba
Gaeyypa0dabaGaamOzamaabmaabaWaaqGaaeaadaGadaqaaiaadkha
daqhaaWcbaGaaCiEaaqaamaabmaabaGaaGOmaaGaayjkaiaawMcaaa
aaaOGaay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4SaeuyQdCLa
eyOeI0YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeaaaOGaayjcSd
WaaiWaaeaacaWGTbWaaSbaaSqaaiaadMgaaeqaaaGccaGL7bGaayzF
aaGaaGilaiabes8a0naaBaaaleaacaaIYaaabeaakiaaiYcacaWGYb
WaaSbaaSqaaiaaikdaaeqaaOGaaGilaiaahg7adaWgaaWcbaGaaGOm
aaqabaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaikdaaeqaaaGccaGLOa
GaayzkaaGaamOzamaabmaabaGaamOCamaaBaaaleaacaaIYaaabeaa
kmaaeeaabaWaaiWaaeaacaWGTbWaaSbaaSqaaiaadMgaaeqaaaGcca
GL7bGaayzFaaGaaGilaiabes8a0naaBaaaleaacaaIYaaabeaakiaa
iYcacaWHXoWaaSbaaSqaaiaaikdaaeqaaOGaaGilaiabeo8aZnaaBa
aaleaacaaIYaaabeaaaOGaay5bSdaacaGLOaGaayzkaaaabaaabaGa
eyyhIulabaWaaebuaeqaleaacaWH4bGaeyicI4SaeuyQdCLaeyOeI0
YaaiWaaeaacaWHWaaacaGL7bGaayzFaaaabeqdcqGHpis1aOWaamWa
aeaadaWcaaqaaiqbec8aWzaaiaWaa0baaSqaaiaahIhaaeaadaqada
qaaiaaikdaaiaawIcacaGLPaaaaaGcdaqadaqaaiaahg7adaWgaaWc
baGaaGOmaaqabaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaikdaaeqaaa
GccaGLOaGaayzkaaaabaGaaGymaiabgkHiTiqbec8aWzaaiaWaa0ba
aSqaaiaahcdaaeaadaqadaqaaiaaikdaaiaawIcacaGLPaaaaaGcda
qadaqaaiaahg7adaWgaaWcbaGaaGOmaaqabaGccaaISaGaeq4Wdm3a
aSbaaSqaaiaaikdaaeqaaaGccaGLOaGaayzkaaaaaaGaay5waiaaw2
faamaaCaaaleqabaGaamOCamaaDaaameaacaWH4baabaWaaeWaaeaa
caaIYaaacaGLOaGaayzkaaaaaaaakiabgEna0oaalaaabaGaeqiXdq
3aaSbaaSqaaiaaikdaaeqaaOGaaiyiaaqaamaabmaabaGaeqiXdq3a
aSbaaSqaaiaaikdaaeqaaOGaeyOeI0IaamOCamaaBaaaleaacaaIYa
aabeaaaOGaayjkaiaawMcaaiaacgcaaaWaamWaaeaacaaIXaGaeyOe
I0IafqiWdaNbaGaadaqhaaWcbaGaaCimaaqaamaabmaabaGaaGOmaa
GaayjkaiaawMcaaaaakmaabmaabaGaaCySdmaaBaaaleaacaaIYaaa
beaakiaaiYcacqaHdpWCdaWgaaWcbaGaaGOmaaqabaaakiaawIcaca
GLPaaaaiaawUfacaGLDbaadaahaaWcbeqaaiaadkhadaWgaaadbaGa
aGOmaaqabaaaaOWaamWaaeaacuaHapaCgaacamaaDaaaleaacaWHWa
aabaWaaeWaaeaacaaIYaaacaGLOaGaayzkaaaaaOWaaeWaaeaacaWH
XoWaaSbaaSqaaiaaikdaaeqaaOGaaGilaiabeo8aZnaaBaaaleaaca
aIYaaabeaaaOGaayjkaiaawMcaaaGaay5waiaaw2faamaaCaaaleqa
baGaeqiXdq3aaSbaaWqaaiaaikdaaeqaaSGaeyOeI0IaamOCamaaBa
aameaacaaIYaaabeaaaaaakeaaaeaacqGH9aqpaeaacaWGmbWaaSba
aSqaaiaaikdacaaIXaaabeaakmaabmaabaGaaCySdmaaBaaaleaaca
aIYaaabeaakiaaiYcacqaHdpWCdaWgaaWcbaGaaGOmaaqabaaakiaa
wIcacaGLPaaacaWGmbWaaSbaaSqaaiaaikdacaaIYaaabeaakmaabm
aabaGaeqiXdq3aaSbaaSqaaiaaikdaaeqaaOGaaGilaiaahg7adaWg
aaWcbaGaaGOmaaqabaGccaaISaGaeq4Wdm3aaSbaaSqaaiaaikdaae
qaaaGccaGLOaGaayzkaaGaaGOlaaaaaaa@1B4E@
Observons
que, dans chaque cas, le premier facteur
L
k
1
(
α
k
,
σ
k
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaadUgacaaIXaaabeaakmaabmaabaGaaCySdmaaBaaaleaa
caWGRbaabeaakiaaiYcacqaHdpWCdaWgaaWcbaGaam4Aaaqabaaaki
aawIcacaGLPaaaaaa@42A2@
est proportionnel
à la fmp conjointe conditionnelle des
{
R
x
(
k
)
}
x
∈
Ω
−
0
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aadkfadaqhaaWcbaGaaCiEaaqaamaabmaabaGaam4AaaGaayjkaiaa
wMcaaaaaaOGaay5Eaiaaw2haamaaBaaaleaacaWH4bGaeyicI4Saeu
yQdCLaeyOeI0IaaCimaaqabaGccaGGSaaaaa@45BD@
sachant que
{
M
i
=
m
i
}
1
n
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aad2eadaWgaaWcbaGaamyAaaqabaGccqGH9aqpcaWGTbWaaSbaaSqa
aiaadMgaaeqaaaGccaGL7bGaayzFaaWaa0baaSqaaiaaigdaaeaaca
WGUbaaaaaa@4183@
et
R
k
=
r
k
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaaS
baaSqaaiaadUgaaeqaaOGaeyypa0JaamOCamaaBaaaleaacaWGRbaa
beaakiaacYcaaaa@3E35@
ce qui est la loi
multinomiale de paramètre de taille
r
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGYbWaaS
baaSqaaiaadUgaaeqaaaaa@3A78@
et de
probabilités
{
π
˜
x
(
k
)
/
[
1
−
π
˜
0
(
k
)
]
}
x
∈
Ω
−
0
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaam
aalyaabaGafqiWdaNbaGaadaqhaaWcbaGaaCiEaaqaamaabmaabaGa
am4AaaGaayjkaiaawMcaaaaaaOqaamaadmaabaGaaGymaiabgkHiTi
qbec8aWzaaiaWaa0baaSqaaiaahcdaaeaadaqadaqaaiaadUgaaiaa
wIcacaGLPaaaaaaakiaawUfacaGLDbaaaaaacaGL7bGaayzFaaWaaS
baaSqaaiaahIhacqGHiiIZcqqHPoWvcqGHsislcaWHWaaabeaakiaa
cYcaaaa@4F97@
et que cette loi
ne dépend pas de
τ
k
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaam4AaaqabaGccaGGUaaaaa@3C02@
Notons
aussi que les seconds facteurs
L
12
(
τ
1
,
α
1
,
σ
1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaigdacaaIYaaabeaakmaabmaabaGaeqiXdq3aaSbaaSqa
aiaaigdaaeqaaOGaaGilaiaahg7adaWgaaWcbaGaaGymaaqabaGcca
aISaGaeq4Wdm3aaSbaaSqaaiaaigdaaeqaaaGccaGLOaGaayzkaaaa
aa@4570@
et
L
22
(
τ
2
,
α
2
,
σ
2
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaikdacaaIYaaabeaakmaabmaabaGaeqiXdq3aaSbaaSqa
aiaaikdaaeqaaOGaaGilaiaahg7adaWgaaWcbaGaaGOmaaqabaGcca
aISaGaeq4Wdm3aaSbaaSqaaiaaikdaaeqaaaGccaGLOaGaayzkaaaa
aa@4574@
sont proportionnels
aux fmp conditionnelles de
R
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaaS
baaSqaaiaaigdaaeqaaaaa@3A23@
et
R
2
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGsbWaaS
baaSqaaiaaikdaaeqaaOGaaiilaaaa@3ADE@
sachant que
{
M
i
=
m
i
}
1
n
,
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaGadaqaai
aad2eadaWgaaWcbaGaamyAaaqabaGccqGH9aqpcaWGTbWaaSbaaSqa
aiaadMgaaeqaaaGccaGL7bGaayzFaaWaa0baaSqaaiaaigdaaeaaca
WGUbaaaOGaaiilaaaa@423C@
qui sont les
lois
Bin
(
τ
1
−
m
,1
−
π
˜
0
(
1
)
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGcbGaae
yAaiaab6gadaqadaqaaiabes8a0naaBaaaleaacaaIXaaabeaakiab
gkHiTiaad2gacaaISaGaaGymaiabgkHiTiqbec8aWzaaiaWaa0baaS
qaaiaahcdaaeaadaqadaqaaiaaigdaaiaawIcacaGLPaaaaaaakiaa
wIcacaGLPaaaaaa@4883@
et
Bin
(
τ
2
,1
−
π
˜
0
(
2
)
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGcbGaae
yAaiaab6gadaqadaqaaiabes8a0naaBaaaleaacaaIYaaabeaakiaa
iYcacaaIXaGaeyOeI0IafqiWdaNbaGaadaqhaaWcbaGaaCimaaqaam
aabmaabaGaaGOmaaGaayjkaiaawMcaaaaaaOGaayjkaiaawMcaaiaa
cYcaaaa@4756@
respectivement, où
Bin
(
τ
,
θ
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaqGcbGaae
yAaiaab6gadaqadaqaaiabes8a0jaaiYcacqaH4oqCaiaawIcacaGL
Paaaaaa@40C1@
désigne la loi binomiale
de paramètre de taille
τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDaa
a@3A2A@
et de
probabilité
θ
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaH4oqCca
GGUaaaaa@3ACD@
Les EMVC
α
^
k
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWHXoGbaK
aadaqhaaWcbaGaam4AaaqaamaabmaabaGaam4qaaGaayjkaiaawMca
aaaaaaa@3D20@
et
σ
^
k
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHdpWCga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGdbaacaGLOaGaayzk
aaaaaaaa@3DA6@
de
α
k
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWHXoWaaS
baaSqaaiaadUgaaeqaaaaa@3ABE@
et
σ
k
, k = 1 , 2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHdpWCda
WgaaWcbaGaam4AaaqabaGccaGGSaGaam4Aaiabg2da9iaaigdacaGG
SaGaaGOmaaaa@401B@
s’obtiennent en maximisant
numériquement
L
11
(
α
1
,
σ
1
)
L
0
(
α
1
,
σ
1
)
et
L
21
(
α
2
,
σ
2
)
(
3.6
)
MathType@MTEF@5@5@+=
feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaigdacaaIXaaabeaakmaabmaabaGaaCySdmaaBaaaleaa
caaIXaaabeaakiaaiYcacqaHdpWCdaWgaaWcbaGaaGymaaqabaaaki
aawIcacaGLPaaacaWGmbWaaSbaaSqaaiaaicdaaeqaaOWaaeWaaeaa
caWHXoWaaSbaaSqaaiaaigdaaeqaaOGaaGilaiabeo8aZnaaBaaale
aacaaIXaaabeaaaOGaayjkaiaawMcaaiaabccacaqGGaGaaeiiaiaa
bwgacaqG0bGaaeiiaiaabccacaqGGaGaamitamaaBaaaleaacaaIYa
GaaGymaaqabaGcdaqadaqaaiaahg7adaWgaaWcbaGaaGOmaaqabaGc
caaISaGaeq4Wdm3aaSbaaSqaaiaaikdaaeqaaaGccaGLOaGaayzkaa
GaaGzbVlaaywW7caaMf8UaaGzbVlaaywW7caGGOaGaaG4maiaac6ca
caaI2aGaaiykaaaa@6584@
par rapport
à
(
α
1
,
σ
1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aahg7adaWgaaWcbaGaaGymaaqabaGccaaISaGaeq4Wdm3aaSbaaSqa
aiaaigdaaeqaaaGccaGLOaGaayzkaaaaaa@3F86@
et
(
α
2
,
σ
2
)
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaadaqadaqaai
aahg7adaWgaaWcbaGaaGOmaaqabaGccaaISaGaeq4Wdm3aaSbaaSqa
aiaaikdaaeqaaaGccaGLOaGaayzkaaGaaiilaaaa@4038@
respectivement.
Notons que, dans (3.6), les facteurs ne dépendent pas de
τ
k
, k = 1 , 2.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaam4AaaqabaGccaGGSaGaam4Aaiabg2da9iaaigdacaGG
SaGaaGOmaiaac6caaaa@40CF@
Enfin, en introduisant les estimations
α
^
k
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWHXoGbaK
aadaqhaaWcbaGaam4AaaqaamaabmaabaGaam4qaaGaayjkaiaawMca
aaaaaaa@3D20@
et
σ
^
k
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHdpWCga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGdbaacaGLOaGaayzk
aaaaaaaa@3DA6@
dans les facteurs de la
fonction de vraisemblance qui dépendent de
τ
k
, k = 1 , 2 ,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaam4AaaqabaGccaGGSaGaam4Aaiabg2da9iaaigdacaGG
SaGaaGOmaiaacYcaaaa@40CD@
et en maximisant ces facteurs,
c’est-à-dire en maximisant
L
12
(
τ
1
,
α
^
1
(
C
)
,
σ
^
1
(
C
)
)
L
MULT
(
τ
1
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaigdacaaIYaaabeaakmaabmaabaGaeqiXdq3aaSbaaSqa
aiaaigdaaeqaaOGaaGilaiqahg7agaqcamaaDaaaleaacaaIXaaaba
WaaeWaaeaacaWGdbaacaGLOaGaayzkaaaaaOGaaGilaiqbeo8aZzaa
jaWaa0baaSqaaiaaigdaaeaadaqadaqaaiaadoeaaiaawIcacaGLPa
aaaaaakiaawIcacaGLPaaacaWGmbWaaSbaaSqaaiaab2eacaqGvbGa
aeitaiaabsfaaeqaaOWaaeWaaeaacqaHepaDdaWgaaWcbaGaaGymaa
qabaaakiaawIcacaGLPaaaaaa@52C8@
et
L
22
(
τ
2
,
α
^
2
(
C
)
,
σ
^
2
(
C
)
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacaWGmbWaaS
baaSqaaiaaikdacaaIYaaabeaakmaabmaabaGaeqiXdq3aaSbaaSqa
aiaaikdaaeqaaOGaaGilaiqahg7agaqcamaaDaaaleaacaaIYaaaba
WaaeWaaeaacaWGdbaacaGLOaGaayzkaaaaaOGaaGilaiqbeo8aZzaa
jaWaa0baaSqaaiaaikdaaeaadaqadaqaaiaadoeaaiaawIcacaGLPa
aaaaaakiaawIcacaGLPaaaaaa@4A38@
par rapport à
τ
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGymaaqabaaaaa@3B11@
et
τ
2
,
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGOmaaqabaGccaGGSaaaaa@3BCC@
respectivement, nous obtenons
que les EMVC
τ
^
1
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIXaaabaWaaeWaaeaacaWGdbaacaGLOaGaayzk
aaaaaaaa@3D73@
et
τ
^
2
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIYaaabaWaaeWaaeaacaWGdbaacaGLOaGaayzk
aaaaaaaa@3D74@
de
τ
1
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGymaaqabaaaaa@3B11@
et
τ
2
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDda
WgaaWcbaGaaGOmaaqabaaaaa@3B12@
sont donnés par (3.5) en
remplaçant
α
^
k
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWHXoGbaK
aadaqhaaWcbaGaam4AaaqaamaabmaabaGaamyvaaGaayjkaiaawMca
aaaaaaa@3D32@
et
σ
^
k
(
U
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHdpWCga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGvbaacaGLOaGaayzk
aaaaaaaa@3DB8@
par
α
^
k
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaaceWHXoGbaK
aadaqhaaWcbaGaam4AaaqaamaabmaabaGaam4qaaGaayjkaiaawMca
aaaaaaa@3D20@
et
σ
^
k
(
C
)
, k = 1 , 2.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHdpWCga
qcamaaDaaaleaacaWGRbaabaWaaeWaaeaacaWGdbaacaGLOaGaayzk
aaaaaOGaaiilaiaadUgacqGH9aqpcaaIXaGaaiilaiaaikdacaGGUa
aaaa@432F@
Observons que ces expressions pour
τ
^
1
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIXaaabaWaaeWaaeaacaWGdbaacaGLOaGaayzk
aaaaaaaa@3D73@
et
τ
^
2
(
C
)
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaDaaaleaacaaIYaaabaWaaeWaaeaacaWGdbaacaGLOaGaayzk
aaaaaaaa@3D74@
sont des expressions
analytiques. L’EMVC de
τ
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacqaHepaDaa
a@3A2A@
est
τ
^
(
C
)
=
τ
^
1
(
C
)
+
τ
^
2
(
C
)
.
MathType@MTEF@5@5@+=
feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn
hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr
4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpipeea0xe9LqFf0x
e9q8qqvqFr0dXdbrVc=b0P0xb9peuD0xXddrpe0=1qpeea0=yrVue9
Fve9Fve8meaabaqaciaacaGaaeqabaWaaeaaeaaakeaacuaHepaDga
qcamaaCaaaleqabaWaaeWaaeaacaWGdbaacaGLOaGaayzkaaaaaOGa
eyypa0JafqiXdqNbaKaadaqhaaWcbaGaaGymaaqaamaabmaabaGaam
4qaaGaayjkaiaawMcaaaaakiabgUcaRiqbes8a0zaajaWaa0baaSqa
aiaaikdaaeaadaqadaqaaiaadoeaaiaawIcacaGLPaaaaaGccaGGUa
aaaa@498D@
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