ezimbalini ngakumbi

Eli candelo yenzelwe ukuba isetyenziswe njengesingqiniso, kunokuba kufundwe njengokuba ebalisayo.

• Intshayelelo (Section 3.1)

Uninzi imixholo kwesi sahluko sele savakala Iidilesi yakutshanje kaMongameli kwi-American Association of Public Opinion Research (AAPOR), ezifana Dillman (2002) , Newport (2011) , Santos (2014) , kunye Link (2015) .

Ngaphezulu kwimbali malunga nophuhliso uphando uphando, funda Smith (1976) kunye Converse (1987) . Kuba ngakumbi kwingcamango ngeminyaka emithathu yophando uphando, funda Groves (2011) kunye Dillman, Smyth, and Christian (2008) (leyo echitha yokufama ezintathu ngokwahlukileyo kancinci).

Incopho ngaphakathi inguqu ukusuka kweyokuqala ukuya kwixesha yesibini kuphando saveyi Groves and Kahn (1979) , nto leyo eyenza intloko-to-intloko uthelekiso oluneenkcukacha phakathi ubuso-ngobuso kunye uphando yomnxeba. Brick and Tucker (2007) ejonga ngasemva kuphuhliso zembali digit oluzenzekelayo iindlela ucofo isampulu.

Kuba ngakumbi indlela uphando uphando sitshintshile ngaphambili ukuziqhelanisa notshintsho kuluntu, bona Tourangeau (2004) , Mitofsky (1989) , kunye Couper (2011) .

• Ukubuza vs. ngokuphawula (Section 3.2)

Ukufunda malunga States zangaphakathi ngokubuza imibuzo unokuba yingxaki kuba ngamanye amaxesha abaphenduli ngokwabo azisilindelanga States zazo zangaphakathi. Umzekelo, Nisbett and Wilson (1977) ube iphepha emangalisayo kunye nesihloko evocative: "Ukubalisa ngaphezu sazi:. Iingxelo zomlomo iinkqubo kwengqondo" Xa iphepha ababhali bathi: "izifundo amaxesha (a) abakuboni ubukho khuthazo kubalulekileyo mbelela impendulo, (b) abakuboni ubukho impendulo, kunye (c) abazi ukuba isivuseleli olubachaphazela impendulo. "

Kuba iimpikiswano ukuba abaphandi kukhetha ukuziphatha bokugcina ukuziphatha ingxelo okanye isimo sengqondo, bona Baumeister, Vohs, and Funder (2007) (ngezengqondo) kwaye Jerolmack and Khan (2014) kunye neempendulo (Maynard 2014; Cerulo 2014; Vaisey 2014; Jerolmack and Khan 2014) (wenzululwazi). Umahluko phakathi sicela yaye ngokujonga kwakhona kuvela nakwezoqoqosho, apho abaphandi bethetha ezikhethwa bezibekiwe kunye kutyhilwa. Ngokomzekelo, umphandi unokubuza abaphenduli ukuba akhetha ukutya iayisi khrimu okanye ukuya gym (ukhetho obekiweyo) okanye uphando unokubona kangaphi abantu badle iayisi khrimu kunye uye gym (ukhetho kutyhilwa). Kukho amathandabuzo enzulu kweentlobo ezithile kuchazwe data ukhetho kwi Economics (Hausman 2012) .

Umxholo eziphambili kwezi ngxoxo kukuba ukuziphatha omchaze alusoloko luchanekile. Kodwa ke, ukuziphatha ebhalwe ngokuzenzekelayo isenokungabi oluchanileyo, ukuze kuqokelelwa kwisampula umdla, kwaye ayizukusebenziseka ukuba abaphandi. Ngoko ke, kwezinye iimeko, ndicinga ukuba ukuziphatha lezenzo kunokuba luncedo. Ngaphezu koko, umxholo yesibini eziphambili kwezi ngxoxo kukuba iingxelo malunga neemvakalelo, ulwazi, abakulindeleyo, kunye nezimvo alusoloko luchanekile. Kodwa ke, ukuba ingcaciso malunga nezi States yangaphakathi ziyimfuneko abaphandi-enoba ukunceda ukucacisa ezinye zokuziphatha okanye into ukukucacisa-ngoko sicela zingaba zezifanelekileyo.

• Iyonke imposiso uphando (Section 3.3)

Kuba unyango ubude incwadi iyonke imposiso uphando, funda Groves et al. (2009) okanye Weisberg (2005) . Kuba imbali uphuhliso iyonke imposiso uphando, funda Groves and Lyberg (2010) .

Ngokwemiqathango ukumelwa, njengentshayelelo enkulu kwimiba ephathelene non-impendulo ne non-impendulo icala yingxelo National Research Council on Nonresponse in Science Social Surveys: A Agenda Research (2013) . Enye isishwankathelo luncedo ibonelelwa (Groves 2006) . Kwakhona, imiba ezikhethekileyo yonke kwiJournal of Statistics Official, Public Opinion zekota, kunye Imbali American Academy of Political neNzululwazi ngezeNtlalo ziye zipapashwe kwi isihloko non-impendulo. Okokugqibela, kukho iindlela ezahlukahlukeneyo yokubala umyinge impendulo; ezi ndlela ichazwe ngokwenkcukacha kwicandelo ingxelo yi-American Association of Public Opinion Abaphandi (AAPOR) (Public Opinion Researchers} 2015) .

I-1936 zoncwadi Digest uhlolisiso sele kufundwa ngokweenkcukacha (Bryson 1976; Squire 1988; Cahalan 1989; Lusinchi 2012) . Kuye kwakhona isetyenziswe umzekeliso zisilumkisa ukuqokelelwa kwedatha engalawulekiyo (Gayo-Avello 2011) . Ngowe-1936, uGeorge Gallup wasebenzisa uhlobo nobunzima kakhulu zokusampula, kwaye bakwazi ukuvelisa uqikelelo oluchanileyo ngakumbi isampuli elincinci kakhulu. Impumelelo lukaGallup kwithuba yokuFunda Digest lulibaleke uphuhliso uphando uphando (Converse 1987, Ch 3; Ohmer 2006, Ch 4; Igo 2008, Ch 3) .

Ngokwemiqathango yokulinganisa, ziluncedo olukhulu kuqala ngokuyila emibuzo na Bradburn, Sudman, and Wansink (2004) . Kuba unyango phambili ingqalelo ngokukodwa imibuzo sengqondo, bona Schuman and Presser (1996) . Engakumbi imibuzo pre-uvavanyo ifumaneka Presser and Blair (1994) , Presser et al. (2004) , kunye neSahluko 8 Groves et al. (2009) .

Unyango ukhethekile, incwadi-ubude urhwebo-off phakathi iindleko uphando neempazamo saveyi Groves (2004) .

• Ngubani ukubuza (Section 3.4)

Classic unyango incwadi-ubude zokusampula ezinokwenzeka umgangatho ngokulinganisela ke Lohr (2009) (intshayelelo ngaphezulu) kwaye Särndal, Swensson, and Wretman (2003) (ngaphezulu nesiphambili). A unyango classic Incwadi-ubude post-abahlulwe enxulumene iindlela na Särndal and Lundström (2005) . Kwezinye izicwangciso ubudala yesuntswana, abaphandi nazi kancinci malunga nabangeyo-abaphenduli, nto leyo engasoloko kunjalo ngaphambili. Iintlobo ezahlukeneyo ulungelelwaniso non-impendulo zinokwenzeka xa abaphandi babe nolwazi malunga nabangeyo-abaphenduli (Kalton and Flores-Cervantes 2003; Smith 2011) .

I-Xbox ukufunda Wang et al. (2015) usebenzisa ubuchule ebizwa parameter multilevel nasemva-abahlulwe (MRP, maxa wambi ebizwa ngokuba "Mister P") evumela abaphandi ukuqikelela iseli kuthetha naxa kukho ezininzi, iiseli ezininzi. Nangona kukho ingxoxo malunga nomgangatho uqikelelo kule ndlela, kubonakala ngathi indawo ethembisayo wokuphonononga. Ubuchule lalisetyenziswa yokuqala Park, Gelman, and Bafumi (2004) , kwaye kubekho ukusetyenziswa kwee-mpikiswano (Gelman 2007; Lax and Phillips 2009; Pacheco 2011; Buttice and Highton 2013; Toshkov 2015) . Kuba vetshe uqhagamshelwano phakathi yamatye ngamnye kunye namatye cell-based ukubona Gelman (2007) .

Kuba ezinye indlela uphando web weighting, bona Schonlau et al. (2009) , Valliant and Dever (2011) , kunye Bethlehem (2010) .

Sample lugqibeko naphakanyiswe Rivers (2007) . Bethlehem (2015) uxoxa ukuba intsebenzo matching isampulu okunene ziya kuba efanayo nezinye iindlela isampuli (umzekelo, isampuli nehlelwe) kunye nezinye iindlela hlengahlengiso (umzekelo, iposi-abahlulwe). Vetshe nasezipaneleni intanethi, bona Callegaro et al. (2014) .

Ngamanye amaxesha abaphandi bafumanise ukuba iisampulu linokuba kunye neesampuli non-linokuba lingayivelisi luqikelela ezikumgangatho ofanayo (Ansolabehere and Schaffner 2014) , kodwa ezinye uthelekiso baye bafumanisa ukuba iisampuli non-linokuba benze okubi (Malhotra and Krosnick 2007; Yeager et al. 2011) . Esinye isizathu lula ukuba umahluko kukuba iisampuli non-linokuba kuphuculwe ekuhambeni kwexesha. Ngokuba nembono angabathembi ngakumbi iindlela isampulu non-linokuba ukubona le AAPOR Task Force on Non-linokuba isampuli (Baker et al. 2013) , kwaye ndinicebisa ukufunda ezimbalini ukuba kulandela ingxelo isishwankathelo.

Kuba meta-ngohlalutyo kwesiphumo weighting ukunciphisa ngomkhethe iisampuli non-ngokuqinisekileyo, funda Itheyibhile 2.4 in Tourangeau, Conrad, and Couper (2013) , nto leyo ekhokelela ababhali bagqibe "uhlengahlengiso babonakala izilungiso luncedo kodwa iimpazamo. . . "

• Indlela ukubuza (Section 3.5)

Conrad and Schober (2008) inika umthamo ohleliweyo ebizwa ngqondweni Survey Udliwano Ngekamva, kwaye uzithetha emininzi imixholo kweli candelo. Couper (2011) sijongene imixholo efanayo, kwaye Schober et al. (2015) inika umzekelo omhle wendlela ukuqokelelwa kweenkcukacha iindlela eziyichaphazela isicwangciso esitsha kungabangela iinkcukacha ezisemgangathweni ophezulu.

Omnye umzekelo umdla zokusebenzisa apps Facebook uphando kwinzululwazi zentlalo, funda Bail (2015) .

Xa ufuna ingcebiso engaphezulu ekwenzeni uphando ngamava unentw esibalulekileyo nxaxheba, funda umsebenzi phezu sanelisa Design Indlela (Dillman, Smyth, and Christian 2014) .

Stone et al. (2007) inika unyango ubude incwadi yovavanyo eziphilayo kokomzuzwana neendlela ezinxulumene.

• Surveys kunxulumene nezinye iinkcukacha (Section 3.6)

Judson (2007) uchaza inkqubo yokudibanisa yeesaveyi data yolawulo njengoko "ulwazi umanyano," esithetha ezinye iingenelo le ndlela, kwaye unika eminye imizekelo.

Enye indlela, ukuba abaphandi basebenzise imizila lwedijithali kunye data yolawulo sisakhelo novavanyo ukuze abantu ezineempawu ezithile. Noko ke, ukufikelela kwezi rekhodi kusetyenziswe ifreyim kwakhona ukudala imibuzo enxulumene labucala (Beskow, Sandler, and Weinberger 2006) .

Ngokuphathelele bezibuza gumba, le nkqubo a intsha njengoko bebonakala kwindlela ndiye wachaza oko. Le ndlela uxhulumaniso enzulu iinkalo ezintathu ezinkulu zamanani-imodeli-based post-abahlulwe (Little 1993) , imputation (Rubin 2004) , yaye ezincinane indawo ngokulinganisela (Rao and Molina 2015) . Kwakhona enxulumene nokusetyenziswa variables ngebambela kuphando lwezonyango (Pepe 1992) .

Ukongeza kwimiba yokuziphatha ngokuphathelele kunikezelo idata umkhondo yesuntswana, ukucela ziyanyuswa nazo kusetyenziswa ukuqonda iimpawu ezinobuzaza ukuba abantu ukuze ukhethe akasichazi uphando (Kosinski, Stillwell, and Graepel 2013) .

Uqikelelo iindleko nexesha Blumenstock, Cadamuro, and On (2015) jonga ukuba iindleko-tshintshayo iindleko iindleko olunye uphando-yaye eyongezelelweyo musa zibandakanya esisigxina ezifana neendleko ukucoca ize isebenze ifowuni yedata. Ngokubanzi, ukucela ziyanyuswa iya kuba iindleko ezinzileyo eziphezulu neendleko eziphantsi eziguquguqukayo efanayo imifuniselo digital (jonga iSahluko 4). Iinkcukacha ezingaphezulu malunga data ezisetyenziswa Blumenstock, Cadamuro, and On (2015) Iphepha zikwi Blumenstock and Eagle (2010) kunye Blumenstock and Eagle (2012) . Beendlela ukusuka imputuation ezininzi (Rubin 2004) ukuze ukunceda capture intandabuzo uqikelelo ukubuza ziyanyuswa. Ukuba abaphandi ukwenza ziyanyuswa ngokubuza nje ukunyamekela malunga ebalulekileyo aggregate, kunokuba imikhwa ngamnye-level, ngoko iindlela in King and Lu (2008) kunye Hopkins and King (2010) ukuze abe luncedo. Kuba ngakumbi iindlela umatshini zokufunda Blumenstock, Cadamuro, and On (2015) , bona James et al. (2013) (intshayelelo ngaphezulu) okanye Hastie, Tibshirani, and Friedman (2009) (ngaphezulu nesiphambili). Omnye umatshini yokufunda textbook ethandwayo Murphy (2012) .

Ngokuphathelele bezibuza natyetyiswa, iziphumo Ansolabehere and Hersh (2012) enze amanyathelo ezimbini eziphambili: 1) isakhono Catalist ukudibanisa imithombo yedatha enani emininzi ukuvelisa inkosi datafile echanileyo 2) isakhono Catalist ukunxulumanisa data zesaveyi inkosi datafile yayo. Ngoko ke, Ansolabehere kunye Hersh bayitshekishe ngocoselelo ngalinye kula manyathelo.

Ukudala inkosi datafile, Catalist idibanisa yaye oko ulwazi kwimithombo emininzi eyahlukeneyo kuqukwa: iirekhodi yokuvota ezininzi ezikhawulezayo kwimeko nganye, data evela Ejongene nezokuxhumana i-National Ukutshintsha Registry Idilesi, kwaye idata kwezinye ababoneleli yorhwebo esingazaziyo. Iinkcukacha emasikizi ngendlela yonke le zokucoca kudityaniswe ntoni na ngaphaya umda kule ncwadi, kodwa le nkqubo, enoba ngenyameko, uya ukusasaza iimpazamo imithombo yedatha original yaye iya kusebenza njengentshayelelo iimpazamo. Nangona Catalist wayekulungele ukuxoxa processing yayo data kwaye anike ezinye iinkcukacha zayo ekrwada, oko nje akunakwenzeka kubaphandi ukuba lihlaziye yonke Catalist data umbhobho. Kunoko, abaphandi be kwimeko apho Catalist ifayile yeenkcukacha kwafuneka ezingaziwayo, kwaye mhlawumbi unokumazi, isixa esiphoso. Le yeyona nto ixhalabisayo kuba umhlalutyi ukuze basenokufunisela ukuba umahluko omkhulu phakathi iingxelo zophando phezu CCES kunye nokuziphatha kwi Catalist ifayile inkosi data zinto zazibangelwa iimpazamo kwifayile inkosi data, kungekhona misreporting abaphendulayo.

Ansolabehere kunye Hersh wathabatha iindlela ezimbini ezahlukeneyo ekusombululeni inkxalabo quality data. iqela lokuqala, ukongeza ngokuthelekisa okuvota self-ingxelo ukuba uvote Catalist inkosi ifayile, naye ke abaphandi kuthelekiswa self-ingxelo, uhlanga, ubume yobhaliso lwabavoti (umzekelo, obhalisiweyo okanye abangabhaliswanga) kunye indlela yokuvota (umzekelo, xa umntu, abangekhoyo lokuvota, njl) abo amaxabiso efumaneka oovimba Catalist. Kuba ezi variables ezine ngokwedemografi, abaphandi bafumanisa kumanqanaba aphezulu kakhulu yesivumelwano phakathi isaveyi kunye data kwi Catalist inkosi ifayile kunokuba yokuvota. Ngoko ke, le Catalist ifayile inkosi data kubonakala ukuba iinkcukacha ezisemgangathweni ophezulu iimpawu ngaphandle lokuvota, ebonisa ukuba ayikho mgangathweni ngokubanzi. Okwesibini, ngokuyinxenye usebenzisa idata Catalist, Ansolabehere kunye Hersh liyenzile imilinganiselo ezintathu ezahlukeneyo umgangatho iirekhodi kwisiphaluka lokuvota, kwaye bafumana ukuba izinga eliqikelelwayo kwama-ingxelo yokuvota eneni ezingahambelaniyo naliphi na kula manyathelo ezisemgangathweni data, imfumanise ukuba bacebisa ukuba amazinga aphezulu kwama-ingxelo akalandelwa aqhutywa kwiikhawunti nomgangatho data libya ngokungaqhelekanga.

Ngenxa indalo le fayili yokuvota inkosi, umthombo wesibini iimpazamo enokwenzeka edibanisa iirekhodi uphando kuyo. Umzekelo, ukuba lo unxibelelano kwenziwa ngokungachanekanga oko kukhokelela kwindlela-Luqikelelo umahluko phakathi ukuziphatha yokuvota ingxelo asemthethweni (Neter, Maynes, and Ramanathan 1965) . Ukuba wonke umntu ibe nozinzo, isazisi esisodwa ebe zombini imvelaphi data, ngoko yePSO kuba engenamsebenzi. Noko ke, kwi-US kunye namanye amazwe, akukho isazisi jikelele. Ngokubhekele phaya, nokuba kukho onjalo abantu njengomqwalaseli mhlawumbi kuba mathidala ukunika ukuba ahlole abaphandi! Ngoko, Catalist Bazakwenza unxibelelano usebenzisa zokufanisa abangafezekanga, kule meko pieces ezine ulwazi ummangalelwa ngamnye: igama, isini, iminyaka wokuzalwa, idilesi yasekhaya. Umzekelo, Catalist kwafuneka isigqibo ukuba ceilingceilingceilingceilingceilingceilingceilingceiling J Simpson kwi CCES yaba ngumntu omnye njengoko kaHomer Jay Simpson kwi zabo lwefayile inkosi data. Ukuziqhelisa, ukutshatisa yinkqubo enzima yaye mdaka yaye, ukuze enze izinto mbi ngenxa abaphandi, Catalist ingqalelo ubuchule yayo zingqameneyo kuba apho.

Ukwenzela ukuba kungqinisiswe algorithms ethelekisa, baxhomekeke nemingeni emibini. Okokuqala, Catalist bathatha inxaxheba kukhuphiswano ethelekisa into aqhutywa ezimeleyo, zeqela lesithathu: le Mitre Corporation. Mitre zonke nxaxheba iifayili zedatha ingxolo ezimbini batshatileyo, kwaye amaqela ezahlukeneyo lwakhuphisana ukubuyela Mitre zingqameneyo ilungileyo. Ngenxa Mitre ngokwayo wayeyazi matching oluchanekileyo bakwazi amanqaku amaqela. Iinkampani 40 zakhuphisana khona, Catalist wafika kwindawo yesibini. Olu hlobo ozimeleyo, uvavanyo wesithathu-iqela kobugcisa azivavanywanga yinto enqabileyo kakhulu ezixabisekileyo; sifanele sibe nentembelo yokuba iinkqubo ethelekisa Catalist zezi ncakasana apho urhulumente-of-the-art. Kodwa ke isimo-of-the-art kakuhle ngokwaneleyo? Ukongeza koku khuphiswano ethelekisa, Ansolabehere kunye Hersh wadala ezabo mngeni yabo ethelekisa for Catalist. Ukusuka iprojekthi ngaphambili, Ansolabehere kunye Hersh baqokelelene iirekhodi lwabavoti e Florida. Zaba ezinye zezi ngxelo kunye amanye amasimi abo redacted ukuba Catalist uze kuthelekiswa neengxelo Catalist ngayo la macandelo kumaxabiso zabo ncakasana. Ngethamsanqa, iingxelo Catalist yabe kufutshane amaxabiso kubanjwa, ebonisa ukuba Catalist nako ukungqinelana iirekhodi umvoti buso azo ifayile inkosi data. Le mingeni ababini, omnye wesithathu-iqela omnye Ansolabehere and Hersh, kusinceda sibe nentembelo ngakumbi Catalist algorithms ethelekisa, nangona asikwazi kwakhona ukuphunyezwa ngqo ngokwethu.

Kukho imizamo emininzi edlulileyo ukugunyazisa okuvota. Kuba umbono ukuba uncwadi, bona Belli et al. (1999) , Berent, Krosnick, and Lupia (2011) , Ansolabehere and Hersh (2012) , kunye Hanmer, Banks, and White (2014) .

Kubalulekile ukuba uqaphele ukuba nangona kule meko abaphandi Bakhuthazwe umgangatho data evela Catalist, ezinye iimvavanyo lwabathengisi yorhwebo ziye ngaphantsi ngomdla. Abaphandi baye bafumana mgangathweni xa idata uphando kumsebenzisi-ifayile ukusuka Marketing Systems Group (leyo ngokwayo lahlangana ndawonye data kubanikezeli ezintathu: Acxiom, Experian, kwaye InfoUSA) (Pasek et al. 2014) . Oko kukuthi, ifayile yeenkcukacha akazange lifane iimpendulo kuphando ukuba abaphandi ekulindeleke ukuba ziyinyani, le datafile ukuba iinkcukacha ezilahlekileyo ngenxa inani elikhulu imibuzo, isilinganiso data ungaziwayo namazinga ixabiso uphando ingxelo (ngamanye amazwi iinkcukacha nto ngendlela , hayi random).

Vetshe irekhodi zonxibelelwano phakathi yeesaveyi data zolawulo, bona Sakshaug and Kreuter (2012) kunye Schnell (2013) . Vetshe irekhodi yePSO ngokubanzi, bona Dunn (1946) kunye Fellegi and Sunter (1969) (historical) kwaye Larsen and Winkler (2014) (namhlanje). Iindlela ezifanayo sele ziphuhlisiwe kwinzululwazi yekhompyutha phantsi amagama afana deduplication data, ukuchongwa Ngokomzekelo, igama lomntu, ucupho duplicate, yaye ikope irekhodi singabonwa (Elmagarmid, Ipeirotis, and Verykios 2007) . Kukwakho yabucala nokugcina iindlela ukurekhoda yePSO leyo ayifuni usasazeko ngokobuqu nokuchonga ulwazi (Schnell 2013) . Abaphandi kula Facebook iphuhlise inkqubo probabilisticsly ukunxibelelanisa iingxelo zabo ukuziphatha ukuvota (Jones et al. 2013) ; le yePSO kwenziwa ukuvavanya uphando Ndiza kukuxelela ngazo kwiSahluko 4 (Bond et al. 2012) .

Omnye umzekelo ukunxibelelanisa enkulu-isikali isaveyi yoluntu kwiirekhodi zolawulo urhulumente livela Health and Retirement Survey kunye Administration yoKhuseleko lweNtlalo. Kuba engakumbi kweso sifundo, kubandakanya ulwazi malunga nenkqubo imvume, bona Olson (1996) kunye Olson (1999) .

Inkqubo yokudibanisa imithombo emininzi kweerekhodi zolawulo ibe inkosi datafile-inkqubo Catalist abaqeshwa-eqhelekileyo kwi ofisi manani abanye oorhulumente yesizwe. Abaphandi ababini Statistics eSweden ndinibhalele incwadi eneenkcukacha nesihloko (Wallgren and Wallgren 2007) . Kuba ndininike umzekelo le ndlela kwindawo okwilizwe enye eUnited States (Olmstead County, Minnesota; kwikhaya lekliniki Mayo), funda Sauver et al. (2011) . Ngaphezulu kwi mposiso ukuba uvele iirekhodi zolawulo, bona Groen (2012) .