3.4.3 Tsis-yuav qhov muaj tseeb kuaj: qauv piv

Tsis yog txhua txhua uas tsis yog-yuav qhov muaj tseeb kuaj yog tib yam. Peb yuav ntxiv tau ntau tswj nyob rau hauv pem hauv ntej kawg.

Tus mus kom ze Wang thiab lug txhawb cov miv siv los laij cov nqi lub sij hawm ntawm lub 2012 US thawj tswj kev xaiv tsa depended nkaus rau kev txhim kho nyob rau hauv cov ntaub ntawv tsom xam. Uas yog, lawv sau raws li muaj ntau yam lus teb raws li lawv tau ces sim mus rov yuag lawv. Ib tug complementary zoo rau ua hauj lwm nrog cov-yuav qhov muaj tseeb piv txwv yog muaj ntau tswj cov ntaub ntawv sau txheej txheem.

Qhov yooj yim piv txwv ntawm ib tug cov tswj uas tsis yog-yuav qhov muaj tseeb piv txwv txheej txheem yog quota zauv, ib tug txheej txheem uas mus rov qab mus rau hauv thaum ntxov hnub ntawm daim ntawv ntsuam xyuas kev tshawb fawb. Nyob rau hauv quota zauv, kev soj ntsuam faib rau cov pejxeem mus rau txawv pab pawg (xws li, cov tub hluas, ntxhais hluas, thiab lwm yam) thiab ces set phiaj rau tus xov tooj ntawm cov neeg yuav tsum tau xaiv nyob rau hauv txhua pab pawg neeg. Cov neeg raug xaiv nyob rau hauv ib tug haphazard yam kom txog thaum tus kws tshawb fawb tau tau ntsib lawv quota nyob rau hauv txhua pab pawg neeg. Vim hais tias ntawm cov quotas, lub resulting qauv zoo ntau tshaj li lub hom phiaj pejxeem tshaj yuav muaj tseeb los yog, tab sis vim hais tias cov probabilities Cov menyuamuas yog tsis paub hais tias muaj ntau yam kev soj ntsuam ntawm yog skeptical ntawm quota zauv. Nyob rau hauv qhov tseeb, quota zauv yog ib tug ua kom muaj tus "Dewey xwb Truman" kev ua yuam kev nyob rau hauv lub 1948 US Medal xaiv tsa. Vim hais tias nws muab ib co tswj cov zauv txheej txheem, txawm li cas los, ib tug yuav saib yuav ua li cas quota zauv yuav muaj ib co zoo dua ib tug kiag li uncontrolled cov ntaub ntawv sau.

Tsiv tshaj quota zauv, ntau niaj hnub le caag los tswj cov uas tsis yog-yuav qhov muaj tseeb piv txwv txheej txheem yog tam sim no tau. Ib tug xws li mus kom ze yog hu ua qauv piv, thiab nws yog siv los ntawm ib co coj mus muag online vaj huam sib luag chaw muab kev pab. Nyob rau hauv nws nyuaj daim ntawv no, cov qauv piv yuav tsum tau ob cov ntaub ntawv qhov chaw: 1) ib tug ua tiav register ntawm cov pej xeem thiab 2) ib tug loj vaj huam sib luag tuaj pab dawb. Nws yog ib qho tseem ceeb tias cov tuaj pab dawb tsis yuav tsum tau ib tug ntau yam qauv los ntawm tej pejxeem; mus sab laj txog hais tias muaj tsis muaj uas yuav tsum tau rau kev xaiv mus rau hauv lub vaj huam sib luag, kuv mam li hu nws lub moj vaj huam sib luag. Tsis tas li ntawd, ob leeg hauv cov pej xeem register thiab cov qias neeg vaj huam sib luag yuav tsum muaj xws li ib co koom haum pab cov lus qhia txog txhua tus neeg, nyob rau hauv qhov kev piv txwv, kuv mam li xav txog lub hnub nyoog thiab pw ua ke, tab sis nyob rau hauv muaj tiag li no koom haum pab ntaub ntawv yuav tsum tau ntau npaum li cas ncauj lus kom ntxaws. Qhov ua kom yuam ntawm cov qauv piv yog yuav xaiv kuaj los ntawm lub moj vaj huam sib luag nyob rau hauv ib txoj kev uas ua tau kuaj uas zoo xws li ntau yam qauv.

Qauv txuam pib thaum ib tug simulated probability qauv yog muab los ntawm cov pejxeem register; no simulated qauv ua ib tug lub hom phiaj qauv. Ces, raws li nyob rau hauv lub koom haum pab ntaub ntawv, tus neeg mob nyob rau hauv lub hom phiaj qauv yog matched rau cov neeg nyob rau hauv lub moj vaj huam sib luag tsim tau ib matched qauv. Piv txwv li, yog hais tias muaj yog ib tug 25 xyoo poj niam nyob rau hauv lub hom phiaj qauv, ces tus soj ntsuam pom ib tug 25 xyoo poj niam los ntawm lub moj vaj huam sib luag yuav tsum nyob rau hauv lub sib phim qauv. Thaum kawg, cov mej zeej ntawm lub matched qauv xam phaj los ua tus kawg set ntawm cov neeg teb.

Txawm tias cov matched qauv zoo li lub hom phiaj qauv, nws yog ib qho tseem ceeb yuav tau nco ntsoov hais tias tus sib phim qauv tsis yog ib tug ntau yam qauv. Matched kuaj tsuas phim lub hom phiaj qauv rau cov paub pab ntaub ntawv (xws li, muaj hnub nyoog thiab pw ua ke), tab sis tsis nyob rau unmeasured yam ntxwv. Piv txwv li, yog tias cov neeg nyob rau hauv lub moj vaj huam sib luag yuav tau poorer-tom qab tag nrho cov, ib tug yog vim li cas mus koom ib tug daim ntawv ntsuam xyuas vaj huam sib luag yog mus khwv tau nyiaj-ces txawm yog hais tias tus sib phim qauv zoo li lub hom phiaj qauv nyob rau hauv cov nqe lus ntawm lub hnub nyoog thiab pw ua ke nws tseem yuav tau ib tug tsis ncaj ncees rau cov neeg pluag neeg. Cov khawv koob ntawm yeej muaj tseeb yuav tshwm sim yog txoj cai tawm muaj teeb meem rau ob leeg ntsuas thiab unmeasured yam ntxwv (ib tug taw tes hais tias yog zoo ib yam nrog peb sib tham uas txuam rau causal inference los ntawm observational kev tshawb fawb nyob rau hauv Tshooj 2).

Nyob rau hauv kev xyaum, cov qauv piv nyob rau muaj ib tug loj thiab muaj ntau haiv neeg vaj huam sib luag xav mus ua kom tiav cov ntawv tshawb fawb, thiab yog li nws yog tsuas yog ua los ntawm cov tuam txhab uas yuav them taus los tsim thiab tswj xws li ib tug vaj huam sib luag. Tsis tas li ntawd, nyob rau hauv kev xyaum, muaj peev xwm yuav muaj teeb meem nrog txuam (tej zaum ib tug zoo phim rau ib tug neeg nyob rau hauv lub hom phiaj qauv tsis nyob rau hauv lub vaj huam sib luag) thiab uas tsis yog-teb (tej zaum cov neeg nyob rau hauv lub sib phim qauv tsis kam mus koom nyob rau hauv lub daim ntawv ntsuam xyuas). Yog li ntawd, nyob rau hauv kev xyaum, kev soj ntsuam ua qauv piv kuj ua tau ib co zoo ntawm post-stratification hloov mus ua kev kwv yees.

Nws yog ib qho nyuaj rau muab pab theoretical guarantees txog cov qauv piv, tab sis nyob rau hauv kev xyaum nws yuav ua tau zoo. Piv txwv li, Stephen Ansolabehere thiab Brian Schaffner (2014) piv peb thaum uas tig mus tshawb fawb txog 1,000 tus neeg ua nyob rau hauv 2010 siv peb sib txawv piv txwv thiab nug kev: mail, xov tooj, thiab ib tug Internet vaj huam sib luag siv cov qauv piv thiab post-stratification haum. Cov kev kwv yees los ntawm peb lub le caag twb zoo heev rau kev kwv yees los ntawm high-quality nrim xws li cov tam sim no Pejxeem Survey (CPS) thiab lub National Health sib tham Survey (NHIS). Dua, ob lub Internet thiab mail tshawb fawb tau tawm los ntawm ib tug nruab nrab ntawm 3 feem pua ​​cov ntsiab lus thiab cov xov tooj daim ntawv ntsuam xyuas yog tawm los ntawm 4 feem pua ​​cov ntsiab lus. Uas tsis no loj yog kwv yees li cas ib tug yuav tau txais los ntawm kuaj ntawm txog 1,000 tus neeg. Txawm tias, tsis muaj leej twg ntawm cov hom ua tshuam zoo dua cov ntaub ntawv, ob lub Internet thiab xov tooj daim ntawv ntsuam xyuas (uas coj hnub los yog lub lim tiam) twb tshuam ceev teb tshaj lub mail daim ntawv ntsuam xyuas (uas coj yim lub hlis), thiab lub Internet daim ntawv ntsuam xyuas, uas siv cov qauv txuam, yog cheaper dua cov lwm yam ob hom.

Nyob rau hauv xaus, kev sib raug zoo kws tshawb fawb thiab statisticians yog incredibly skeptical ntawm hais los ntawm cov uas tsis yog-yuav qhov muaj tseeb kuaj, nyob rau hauv ib feem vim hais tias lawv yog txuam nrog ib co embarrassing failures ntawm daim ntawv ntsuam xyuas kev tshawb fawb xws li cov Zajlus zom poll. Nyob rau hauv ib feem, kuv pom zoo nrog rau qhov no skepticism: unadjusted uas tsis yog-yuav qhov muaj tseeb kuaj yuav ua phem kev kwv yees. Txawm li cas los, Yog hais tias soj ntsuam yuav tau kho rau tus biases nyob rau hauv cov zauv txheej txheem (xws li, post-stratification) los yog tswj cov zauv txheej txheem dog dig (xws li, qauv piv), lawv yuav ua tau zoo dua kev kwv yees, thiab txawm kev kwv yees ntawm txaus zoo rau feem ntau cov hom phiaj. Ntawm cov hoob kawm, nws yuav zoo dua rau ua paug tua yuav tshwm sim, tab sis hais tias tsis muaj lawm zoo nkaus li yuav tsum tau ib tug muaj tiag kev xaiv.

Ob leeg uas tsis yog-yuav qhov muaj tseeb kuaj thiab kev kwv yees kuaj sib txawv nyob rau hauv lawv cov zoo, thiab tam sim no nws yog tej zaum cov rooj plaub uas feem ntau kwv yees los ntawm qhov muaj tseeb kuaj muaj ntau trustworthy tshaj kev kwv yees los ntawm cov-yuav qhov muaj tseeb kuaj. Tab sis, txawm tam sim no, kev kwv yees los ntawm zoo-ua uas tsis yog-yuav qhov muaj tseeb kuaj yog tej zaum zoo dua kev kwv yees los ntawm tsis zoo-ua ntau yam kuaj. Ntxiv mus, uas tsis yog-yuav qhov muaj tseeb kuaj yog tshuam cheaper. Yog li, nws zoo nkaus li uas yuav tshwm vs uas tsis yog-yuav qhov muaj tseeb piv txwv muaj ib tug nqi zoo trade-tawm (Daim duab 3.6). Nrhiav rau pem hauv ntej, kuv xav hais tias kev kwv yees los ntawm zoo-ua uas tsis yog-yuav qhov muaj tseeb kuaj yuav ua pheej yig dua thiab zoo dua. Ntxiv mus, vim hais tias ntawm lub neej puas nyob rau hauv landline xov tooj ntawv tshawb fawb thiab ua cov nqi ntawm cov uas tsis yog-teb, kuv xav hais tias muaj tseeb kuaj yuav ua kim kim thiab ntawm sab zoo. Vim hais tias ntawm cov sij hawm ntev tiam sis, kuv xav hais tias tsis yog-yuav qhov muaj tseeb piv txwv yuav ua nce tseem ceeb heev nyob rau hauv lub thib peb era ntawm daim ntawv ntsuam xyuas kev tshawb fawb.

Daim duab 3.6: ntau yam piv txwv nyob rau hauv kev xyaum thiab uas tsis yog-yuav qhov muaj tseeb piv txwv yog ob qho tib si loj, ua kev pawg. Nyob rau hauv kev, muaj ib tug nqi-kev ua yuam kev lag luam-off nrog non-yuav qhov muaj tseeb piv txwv ua txo nqi tab sis ntau yuam kev. Txawm li cas los, zoo-ua uas tsis yog-yuav qhov muaj tseeb piv txwv yuav ua tau zoo dua kev kwv yees tshaj tsis zoo-ua ntau yam piv txwv. Nyob rau hauv lub neej yav tom ntej, kuv xav hais tias tsis yog-yuav qhov muaj tseeb piv txwv yuav tau txais zoo dua thiab pheej yig dua thaum uas yuav tshwm sim yuav mob heev ntxiv thiab kim kim.