[ , ] Esahlukweni, ngangizimisele kakhulu ngokulandelana kwamagama. Noma kunjalo, lokhu akuhlali njalo kuthuthukisa ikhwalithi yezilinganiso. Yakha isimo lapho ukuhlelwa kwe-post-stratification kunganciphisa izinga lokulinganisa. (Ukuze uthole ukusikisela, bheka Thomsen (1973) .)
[ , , ] Ukuklama nokuqhuba inhlolovo engekho emathubeni e-Amazon Mechanical Turk ukubuza mayelana nobunikazi bebhamu kanye nesimo sengqondo ngokulawulwa kwezibhamu. Ukuze ukwazi ukuqhathanisa izilinganiso zakho kulabo abathathwe kwisampula esingenzeka, sicela ukopishe umbuzo wombuzo kanye nokukhethwa kokuphendula ngokuqondile ngokuhlola okuphezulu kakhulu okufana nalabo abagijimela iPew Research Research.
[ , , ] I-Goel kanye nozakwethu (2016) bephethe imibuzo engama-49 yokuzikhethela yokuziphatha okuvela ku-General Social Survey (GSS) bese ukhetha ukuhlola yi-Pew Research Center kuya kwesampula esingenakwenzeka kubantu abaphenduliwe abavela ku-Amazon Mechanical Turk. Bese balungiselela ukungabonakali kwedatha besebenzisa i-post-stratification esekelwe emodeli futhi baqhathanisa izilinganiso zabo ezilungiselelwe nalabo abavela ku-GSS kanye no-Pew. Yenza uphenyo olufanayo ku-Amazon Mechanical Turk bese uzama ukuphinda uphinde ufunde isithombe esingu-2a bese ufaka u-2b ngokuqhathanisa izilinganiso zakho ezilungiselelwe nezilinganiso ezivela emaceleni omfisha kakhulu we-GSS no-Pew uphenyo. (Bheka ithebula le-appendix A2 ohlwini lwamakhasi angu-49.)
[ , , ] Ucwaningo oluningi lusebenzisa izinyathelo ezizimele zokusetshenziswa komakhalekhukhwini. Lokhu kuyisimo esithakazelisayo lapho abacwaningi bangakuqhathanisa ukuziphatha okubikwayo ngokuziphatha okuhle (bheka isib. Boase and Ling (2013) ). Ukuziphatha okuvamile okubili okubuzayo mayelana nokubiza nokuthumelela imiyalezo, futhi amafomu amabili avamile ejwayelekile "yizolo" futhi "ngesonto elidlule."
[ , ] U-Schuman and Presser (1996) uphikisa ukuthi imiyalo yombuzo izobaluleka ngezinhlobo ezimbili zemibuto: imibuzo eyingxenye-ingxenye lapho imibuzo emibili ikhona ezingeni elifanayo lenkampani ethile (isibonelo, izilinganiso ezimbili zokukhethwa koMengameli); kanye nengxenye-yonke imibuzo lapho umbuzo ojwayelekile ulandela umbuzo ocacile (isb., ukubuza ukuthi "Uneliseke kangakanani ngomsebenzi wakho?" kulandelwe "Uneliseke kangakanani ngokuphila kwakho?").
Baqhubeka beveza izinhlobo ezimbili zomphumela we-oda lombuzo: imiphumela ehambisanayo lapho kwenzeka izimpendulo zombuzo kamuva (kunokuba zibe khona) kulabo abanikezwe embuzweni wangaphambili; Ukungafani kwemiphumela kwenzeka uma kukhona umehluko omkhulu phakathi kwezimpendulo zemibuzo emibili.
[ , ] Ukwakha umsebenzi we-Schuman and Presser, u- Moore (2002) uchaza ubukhulu obuhlukile bomphumela womyalelo wombuzo: imiphumela engezayo nokukhupha. Ngenkathi ukungafani nokuphumelela kwemiphumela kukhishwa njengomphumela wabahlolwa kwabaphenduli bezinto ezimbili ngokuphathelene nomunye nomunye, imiphumela ehlanganisayo neyokukhipha iyakhiwa lapho abaphendulayo bezwela ngokwengeziwe kuhlaka olukhudlwana lapho imibuzo ibuzwa khona. Funda Moore (2002) , bese uhlela bese uqhuba ucwaningo olwenziwe ku-MTurk ukuze ubonise imiphumela engeza noma yokukhipha.
[ , ] UChristopher Antoun kanye nozakwethu (2015) baqhube isifundo ngokuqhathanisa amasampuli alula atholakala emithonjeni emine yokuqoqa inthanethi: MTurk, Craigslist, Google AdWords ne-Facebook. Yakha inhlolovo elula futhi uqoqe abahlanganyeli ngokusebenzisa okungenani imithombo emibili yokuqasha inthanethi (le mithombo ingahlukahluka emithonjeni emine esetshenziswa ku- Antoun et al. (2015) ).
[ ] Ngomzamo wokubikezela imiphumela ye-2016 EU Referenndum (okungukuthi, i-Brexit), i-YouGov-i-intanethi esekelwe emakethe yokucwaninga emakethe eqhutshwa e-intanethi yephaneli labantu abangaba ngu-800 000 e-United Kingdom.
Incazelo eningiliziwe yemodeli yezibalo ze-YouGov ingatholakala ku-https: //yougov.co.uk/news/2016/06/21/yougov-referendum-model/. Uma ukhuluma, u-YouGov uhlukanisile abavoti ngezinhlobo ezisuselwe ku-choice jikelele wokuvota okhethweni luka-2015, ubudala, iziqu, ubulili, nosuku lokuxoxisana, kanye nendawo abahlala kuyo. Okokuqala, basebenzise idatha eqoqwe kusukela ku-YouGov panelist ukulinganisa, phakathi kwalabo abavotele, inani labantu abantu abavotelayo abazovota. Balinganisela ukuguqulwa kohlobo ngalunye lwabavoti ngokusebenzisa uhlelo lwe-2015 eBrithani Election Study (BES), inhlolovo ebusweni bombuso obhekene nokulandela, okwakungavumelekile ukuphuma emaqenjini okhetho. Ekugcineni, balinganisela ukuthi bangaki abantu ababekhona kulolu hlobo lwamavoti abakhethiweyo ku-electorate, ngokususelwa ku-Recense Census no-Annual Population Survey (ngolunye ulwazi olwengeziwe kusuka eminye imithombo yedatha).
Izinsuku ezintathu ngaphambi kokuvota, u-YouGov ubonise ukuholwa kwamaphuzu amabili we-Shiya. Ngosuku lokuvota, i-poll ibonise ukuthi umphumela wawusondele kakhulu ukushayela (49/51 Hlalani). Ucwaningo lokugcina olwenziwe ngosuku lubikezela ukuthi u-48/52 uhlale uhlala (https://yougov.co.uk/news/2016/06/23/yougov-day-poll/). Empeleni, lesi sibalo siphuthele imiphumela yokugcina (52/48 Leave) ngamaphuzu amane amaphesenti.
[ , ] Bhala imidwebo ukuze ufanise amaphutha okumelela ngamunye ku-3.2.
[ , ] Ucwaningo lwe-Blumenstock nosebenza nabo (2015) luhilelekile ukwakha imodeli yokufunda imishini engasebenzisa idatha ye-digital trace ukubikezela izimpendulo zezocwaningo. Manje, uzozama into efanayo nge-dataset ehlukile. Kosinski, Stillwell, and Graepel (2013) bathole ukuthi ukuthandwa kwe-Facebook kungakwazi ukubikezela izici nezici ezithile. Okumangalisa ukuthi lezi zibikezelo zingaba nenembile nakakhulu kunabangane nozakwabo (Youyou, Kosinski, and Stillwell 2015) .
[ ] Toole et al. (2015) basebenzisa amarekhodi imininingwane yocingo (CDRs) kusuka emafonini ephathekayo ukubikezela izitayela zokungaqashwa komphakathi.