Est-ce que la réduction du déséquilibre de la réponse accroît l’exactitude des estimations de l’enquête ? Section 5. Estimation du total de population en présence de non-réponse

L’équation (4.2), quand elle est multipliée par N ^ = s d k , MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmOtayaaja Gaeyypa0ZaaabeaeaacaWGKbWaaSbaaSqaaiaadUgaaeqaaaqaaiaa dohaaeqaniabggHiLdGccaGGSaaaaa@3B8F@ peut être exprimée en fonction de trois estimateurs courants du total de population Y = U y k . MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamywaiabg2 da9maaqababaGaamyEamaaBaaaleaacaWGRbaabeaaaeaacaWGvbaa beqdcqGHris5aOGaaiOlaaaa@3B83@ Deux d’entre eux sont possibles en présence de non-réponse,

Y ^ E X P = N ^ r d k y k r d k , Y ^ C A L = N ^ r d k g k y k r d k , ( 5.1 ) MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadweacaWGybGaamiuaaqabaGccqGH9aqpceWGobGb aKaadaWcaaqaamaaqababaGaamizamaaBaaaleaacaWGRbaabeaaae aacaWGYbaabeqdcqGHris5aOGaamyEamaaBaaaleaacaWGRbaabeaa aOqaamaaqababaGaamizamaaBaaaleaacaWGRbaabeaaaeaacaWGYb aabeqdcqGHris5aaaakiaacYcacaaMf8UabmywayaajaWaaSbaaSqa aiaadoeacaWGbbGaamitaaqabaGccqGH9aqpceWGobGbaKaadaWcaa qaamaaqababaGaamizamaaBaaaleaacaWGRbaabeaaaeaacaWGYbaa beqdcqGHris5aOGaam4zamaaBaaaleaacaWGRbaabeaakiaadMhada WgaaWcbaGaam4AaaqabaaakeaadaaeqaqaaiaadsgadaWgaaWcbaGa am4AaaqabaaabaGaamOCaaqab0GaeyyeIuoaaaGccaGGSaGaaGzbVl aaywW7caaMf8UaaGzbVlaaywW7caGGOaGaaGynaiaac6cacaaIXaGa aiykaaaa@6736@

avec g k = x ¯ s Σ r 1 x k . MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaam4zamaaBa aaleaacaWGRbaabeaakiabg2da9iqahIhagaqegaqbamaaBaaaleaa caWGZbaabeaakiaaho6adaqhaaWcbaGaamOCaaqaaiabgkHiTiaaig daaaGccaWH4bWaaSbaaSqaaiaadUgaaeqaaOGaaiOlaaaa@405F@ De ces estimateurs, Y ^ E X P MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadweacaWGybGaamiuaaqabaaaaa@37AD@ est une simple extension de la moyenne de y MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyEaaaa@3515@ dans la réponse et peut être fortement biaisé. L’estimateur par calage Y ^ C A L MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadoeacaWGbbGaamitaaqabaaaaa@3790@ donne à y k MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyEamaaBa aaleaacaWGRbaabeaaaaa@3631@ le poids d k g k / P . MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaSGbaeaaca WGKbWaaSbaaSqaaiaadUgaaeqaaOGaam4zamaaBaaaleaacaWGRbaa beaaaOqaaiaadcfaaaGaaiOlaaaa@39D5@ La propriété de calage est r ( d k g k / P ) x k = s d k x k , MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaabeaeaada qadaqaamaalyaabaGaamizamaaBaaaleaacaWGRbaabeaakiaadEga daWgaaWcbaGaam4AaaqabaaakeaacaWGqbaaaaGaayjkaiaawMcaai aahIhadaWgaaWcbaGaam4AaaqabaaabaGaamOCaaqab0GaeyyeIuoa kiabg2da9maaqababaGaamizamaaBaaaleaacaWGRbaabeaaaeaaca WGZbaabeqdcqGHris5aOGaaCiEamaaBaaaleaacaWGRbaabeaakiaa cYcaaaa@485E@ où le deuxième membre de l’équation est sans biais pour le total x MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaCiEaaaa@3518@ de population U x k , MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaabeaeaaca WH4bWaaSbaaSqaaiaadUgaaeqaaaqaaiaadwfaaeqaniabggHiLdGc caGGSaaaaa@39A0@ ce qui explique pourquoi Y ^ C A L MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadoeacaWGbbGaamitaaqabaaaaa@3790@ peut être considérablement moins biaisé que Y ^ E X P , MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadweacaWGybGaamiuaaqabaGccaGGSaaaaa@3867@ quand x MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaCiEaaaa@3518@ et y MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyEaaaa@3515@ sont bien reliées. Si les valeurs y k MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyEamaaBa aaleaacaWGRbaabeaaaaa@3631@ sont enregistrées pour l’échantillon complet s , MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaam4CaiaacY caaaa@35BF@ une estimation sans biais sera obtenue au moyen de l’estimateur de Horvitz-Thompson

Y ^ F U L = s d k y k . MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadAeacaWGvbGaamitaaqabaGccqGH9aqpdaaeqaqa aiaadsgadaWgaaWcbaGaam4AaaqabaaabaGaam4Caaqab0GaeyyeIu oakiaadMhadaWgaaWcbaGaam4AaaqabaGccaGGUaaaaa@406B@

Nous désignerons ces trois types d’estimateur par EXP, CAL et FUL (pour full sample). Maintenant, l’expression (4.2) multipliée par N ^ = s d k MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmOtayaaja Gaeyypa0ZaaabeaeaacaWGKbWaaSbaaSqaaiaadUgaaeqaaaqaaiaa dohaaeqaniabggHiLdaaaa@3AD5@ se lit

Y ^ E X P Y ^ F U L = ( Y ^ E X P Y ^ C A L ) + ( Y ^ C A L Y ^ F U L ) . ( 5.2 ) MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadweacaWGybGaamiuaaqabaGccqGHsislceWGzbGb aKaadaWgaaWcbaGaamOraiaadwfacaWGmbaabeaakiabg2da9maabm aabaGabmywayaajaWaaSbaaSqaaiaadweacaWGybGaamiuaaqabaGc cqGHsislceWGzbGbaKaadaWgaaWcbaGaam4qaiaadgeacaWGmbaabe aaaOGaayjkaiaawMcaaiabgUcaRmaabmaabaGabmywayaajaWaaSba aSqaaiaadoeacaWGbbGaamitaaqabaGccqGHsislceWGzbGbaKaada WgaaWcbaGaamOraiaadwfacaWGmbaabeaaaOGaayjkaiaawMcaaiaa c6cacaaMf8UaaGzbVlaaywW7caaMf8UaaGzbVlaacIcacaaI1aGaai OlaiaaikdacaGGPaaaaa@5D4F@

Exprimée en mots, Écart de E X P = terme d'ajustement du biais + écart de C A L . MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaeyYaiaabo gacaqGHbGaaeOCaiaabshacaqGGaGaaeizaiaabwgacaaMe8Uaamyr aiaadIfacaWGqbGaeyypa0JaaeiDaiaabwgacaqGYbGaaeyBaiaabw gacaaMe8UaaeizaiaabEcacaqGHbGaaeOAaiaabwhacaqGZbGaaeiD aiaabwgacaqGTbGaaeyzaiaab6gacaqG0bGaaeiiaiaabsgacaqG1b GaaeiiaiaabkgacaqGPbGaaeyyaiaabMgacaqGZbGaey4kaSIaaey6 aiaabogacaqGHbGaaeOCaiaabshacaaMe8UaaeizaiaabwgacaaMe8 Uaam4qaiaadgeacaWGmbGaaiOlaaaa@677C@  L’ajustement calculable est Y ^ E X P Y ^ C A L = N ^ ( x ¯ r x ¯ s ) b r . MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadweacaWGybGaamiuaaqabaGccqGHsislceWGzbGb aKaadaWgaaWcbaGaam4qaiaadgeacaWGmbaabeaakiabg2da9iqad6 eagaqcamaabmaabaGabCiEayaaraWaaSbaaSqaaiaadkhaaeqaaOGa eyOeI0IabCiEayaaraWaaSbaaSqaaiaadohaaeqaaaGccaGLOaGaay zkaaWaaWbaaSqabeaakiadaITHYaIOaaGaaCOyamaaBaaaleaacaWG Ybaabeaakiaac6caaaa@4AF4@ Les deux écarts par rapport à l’estimateur sans biais, Y ^ C A L Y ^ F U L = N ^ ( b r b s ) x ¯ s MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadoeacaWGbbGaamitaaqabaGccqGHsislceWGzbGb aKaadaWgaaWcbaGaamOraiaadwfacaWGmbaabeaakiabg2da9iqad6 eagaqcamaabmaabaGaaCOyamaaBaaaleaacaWGYbaabeaakiabgkHi TiaahkgadaWgaaWcbaGaam4CaaqabaaakiaawIcacaGLPaaadaahaa WcbeqaaOGamai2gkdiIcaaceWH4bGbaebadaWgaaWcbaGaam4Caaqa baaaaa@4A05@ pour CAL et Y ^ E X P Y ^ F U L = N ^ ( y ¯ r y ¯ s ) MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadweacaWGybGaamiuaaqabaGccqGHsislceWGzbGb aKaadaWgaaWcbaGaamOraiaadwfacaWGmbaabeaakiabg2da9iqad6 eagaqcamaabmaabaGabmyEayaaraWaaSbaaSqaaiaadkhaaeqaaOGa eyOeI0IabmyEayaaraWaaSbaaSqaaiaadohaaeqaaaGccaGLOaGaay zkaaaaaa@4524@ pour EXP, ne sont pas calculables en présence de non-réponse, parce qu’ils requièrent les valeurs de y MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyEaaaa@3515@ pour l’échantillon complet.

Comme nous l’avons mentionné, il existe des méthodes pour réduire le déséquilibre I M B MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamysaiaad2 eacaWGcbaaaa@367E@ durant la collecte des données. Un faible déséquilibre est intuitivement intéressant, mais aboutit-il à une plus grande exactitude des estimations ? Ou bien, cela suffit-il de faire intervenir les variables auxiliaires à l’étape de l’estimation, par un ajustement de la pondération par calage comme dans l’estimateur CAL ? Le terme d’ajustement Y ^ E X P Y ^ C A L = N ^ ( x ¯ r x ¯ s ) b r MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadweacaWGybGaamiuaaqabaGccqGHsislceWGzbGb aKaadaWgaaWcbaGaam4qaiaadgeacaWGmbaabeaakiabg2da9iqad6 eagaqcamaabmaabaGabCiEayaaraWaaSbaaSqaaiaadkhaaeqaaOGa eyOeI0IabCiEayaaraWaaSbaaSqaaiaadohaaeqaaaGccaGLOaGaay zkaaWaaWbaaSqabeaakiadaITHYaIOaaGaaCOyamaaBaaaleaacaWG Ybaabeaaaaa@4A38@ peut clairement être réduit en construisant r MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamOCaaaa@350E@ de manière que le déséquilibre soit faible; il est égal à zéro pour l’équilibre parfait x ¯ r = x ¯ s . MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabCiEayaara WaaSbaaSqaaiaadkhaaeqaaOGaeyypa0JabCiEayaaraWaaSbaaSqa aiaadohaaeqaaOGaaiOlaaaa@3A5C@ En pratique, l’estimateur CAL est préféré à l’estimateur EXP, le premier étant habituellement plus précis, en raison de l’information auxiliaire. Mais l’écart Y ^ C A L Y ^ F U L = N ^ ( b r b s ) x ¯ s MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabmywayaaja WaaSbaaSqaaiaadoeacaWGbbGaamitaaqabaGccqGHsislceWGzbGb aKaadaWgaaWcbaGaamOraiaadwfacaWGmbaabeaakiabg2da9iqad6 eagaqcamaabmaabaGaaCOyamaaBaaaleaacaWGYbaabeaakiabgkHi TiaahkgadaWgaaWcbaGaam4CaaqabaaakiaawIcacaGLPaaadaahaa WcbeqaaOGamai2gkdiIcaaceWH4bGbaebadaWgaaWcbaGaam4Caaqa baaaaa@4A05@ est-il plus petit si la réponse r MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamOCaaaa@350E@ a été construite de manière à ce que I M B MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamysaiaad2 eacaWGcbaaaa@367E@ soit faible ? Autrement dit, cela vaut-t-il l’effort (peut-être coûteux) de gérer la collecte des données de manière que x ¯ r MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabCiEayaara WaaSbaaSqaaiaadkhaaeqaaaaa@3653@ soit plus proche de x ¯ s MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGabCiEayaara WaaSbaaSqaaiaadohaaeqaaaaa@3654@ et, par conséquent, que I M B MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamysaiaad2 eacaWGcbaaaa@367E@ soit réduit ? La question est essentiellement celle de savoir si cela ferait aussi se rapprocher b r MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaCOyamaaBa aaleaacaWGYbaabeaaaaa@3625@ et b s . MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiFu0Je9sqqrpepC0xbbL8F4rqqrpepC0xd9Wqpe0dd9 qqaqFeFr0xbbG8FaYPYRWFb9fi0lXxbvc9Ff0dfrpm0dXdHqps0=vr 0=vr0=fdbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaCOyamaaBa aaleaacaWGZbaabeaakiaac6caaaa@36E2@

Date de modification :