Me ka ole-probability Eia kekahi laʻana, paona ke wehe ei kena ae la ia e ka ea mea hōʻikeʻikeʻuʻuku kaʻina.
I ka mea ala i noiʻi ke kaupaonaʻana pane mai probability Eia kekahi laʻana, aole lakou e hiki no hoi ke kaupaonaʻana pane ka politika, probability Eia kekahi laʻana. No ka mea, laʻana, e like me kekahi koho ana i na CPS, manao wale ia oe i kau hae hoʻolaha pāʻoihana ma luna o nā tausani o na punaewele e Imi i komo i ka anamanaʻo ke hoʻomaopopoʻia ka nele ana. Maoli ai, ina e-Liloa i ka mea mawaena o oukou hāpana e lilo i ka olelo maikai o ka nele ana. Ko oukou skepticism paha, no ko oukou manao i kekahi poe kanaka he nui e kiʻi mai ia hoʻoholo 'ia kāu ana mamua o na mea e ae. No ka mea, laʻana, ka poe i mai hoomaunauna i ka puu o ka manawa ma ka pūnaewele puni honua, he emi e kiʻi mai ia hoʻoholo 'ia kāu anamanaʻo.
E like me ka makou i ike ai ma ka pauku hope, nae, ina aole makou i ike i ka hāpana ua koho-e like me ka makou e hana aku me ka probability Eia kekahi laʻana-a laila, hiki iā kākou ke wehe ei kena ae la ia e ka mea hōʻikeʻikeʻuʻuku kaʻina. Eia naʻe, e hana me ka ole-probability Eia kekahi laʻana i ka wa, aole makou i ike i ua koho 'o ka hāpana. Aka, ia kakou ke hana assumptions e pili ana i ka mea hōʻikeʻikeʻuʻuku 'ana, a laila, hoi kaumaha ma ke ala hookahi. Ina keia mau assumptions ua pololei, alaila, ke kaumaha, e wehe i nā ei kauoha ia e ka mea hōʻikeʻikeʻuʻuku kaʻina.
No ka mea, kumu, manao wale ia ma ka pane ana i ko oukou hae hoʻolaha pāʻoihana, oe pūʻali koa i mea 100.000 respondents. Eia naʻe, aole oe i manaoio i keia mau 100.000 respondents poe, he mea kaulele hāpana oʻAmelika makua. I mea, i ka wa a oukou e hoohalike ai ko oukou respondents i ka US heluna, oe loaa i na kanaka, mai kekahi Hui Pūʻia (e like me, New York) ua ma luna o ka-poe a me ka poe kanaka, mai kekahi Hui Pūʻia (e like me, Alaska) he lalo-poe. Pela, no ka nele ana o ko oukou hāpana hoʻoneʻe 'ia paha e lilo i ka ino olelo o ka nele ana i loko o ka pale umauma hoʻokahi heluna.
ala hanakimi e wehe i ka ei i hanaia ma ka mea hōʻikeʻikeʻuʻuku kaʻina mea, e hāʻawi aku paona i kela a me keia kanaka; lalo paona i ka poe kanaka mai Huipuia i ka poe luna, na poe ma ka hāpana (e like me, New York), a oi aku paona i ka poe kanaka mai Huipuia i ka poe malalo o-na poe iloko o ka hāpana (e like me, Alaska). More ua hōʻike hewa, ke kaupaona no kela a me keia respondent, ua pili ia i ko lakou prevalence iloko o ko oukou hāpana pili i ko lakou prevalence iloko o ka US heluna. Keia kaumaha ke kaʻina hana no ka mea, ua kapaia kia-stratification, a me ka manaʻo o kēia ipu kālā e paipai aku i ka hana i loko o ka Pauku 3.4.1 kahi respondents mai o Rode Island, ua haawiia emi ke kaupaonaʻana ma mua o respondents mai Kaleponi. Post-stratification pono ia oe ike lawa ai i ko oukou mau respondents iloko o pūʻulu, a me ka ike i ka like o ka pale umauma hoʻokahi lehulehu ma kela a me keia hui.
Oiai ke kaumaha o ka probability hāpana, a no ka non-probability hāpana no ka like mathematically (e nānā oaoieei-i paiiaʻi na), lakou i hana pono iloko oʻokoʻa wahi, i hiki. Ina o ke kanaka noiʻi i ka pono probability hāpana ( 'o ia hoʻi, aoleʻaoʻao pale kikowaena, a aole hoi ole-pane), a laila, kaumaha e paka pāʻewaʻole koho no ka mea, na kāna moʻolelo i loko o nā hihia. Keia ikaika theoretical kumu no ke aha la e Uwao o ka probability Eia kekahi laʻana loaa ia lakou pela uʻi. Ma ka kekahi lima, kaumaha ole probability Eia kekahi laʻana e wale paka pāʻewaʻole koho no ka mea, na kāna moʻolelo, ina ka pane propensities, he hookahi no kela mea keia mea iloko o kela a me keia hui. Ma kekahi olelo, manao hou ia i ko kakou kumu hoohalike, me kia-stratification e paka pāʻewaʻole koho, ina pau i loko o New York i ka hookahi probability o komo a pau i loko o Alaska i ka hookahi probability o ka 'oihana, a pela aku. Keia assumption ua kapaia ka homogeneous-pane-propensities-loko-hui assumption, a he Radio he pelekikena a kī i ka ike ina kia-stratification, e hana maikaʻi me ka 'ole-probability Eia kekahi laʻana.
Eia naʻe, i loko o ko mākou laʻana, ka homogeneous-pane-propensities-loko-pūʻulu assumption mea olelo lealea, e noho nei. Oia hoi, he mea lealea na mea a pau i loko o Alaska i ka hookahi probability o kona noho ana iloko o ko oukou anamanaʻo. Aka, aia he ekolu nui heluʻai e malama i ka manao ana e kia-stratification, o na mea a pau e he mea hou i olelo mua iaʻi.
Mua, homogeneous-pane-propensities-loko-pūʻulu assumption i oi aku plausible e like me ka heluʻana o nā pūʻulu, piʻi pū. A, noiʻi 'aʻole e kaupalena i ka pūʻulu pono ma ka hookahi Geographic Ana. No ka mea, laʻana, ia makou ke hana pūʻulu nānā 'ana i Aupuni, makahiki, ke keka, a me ka nui o ka hoʻonaʻauao' ana. Ka mea, he oi aku plausible ia he mea homogeneous pane propensities iloko o ka hui o 18-29, wahine, ke kulanui puka e noho ana ma Alaska ma mua i loko o ka pae ana o na kanaka a pau e noho ana ma Alaska. Pela, e like me ka heluʻana o nā pūʻulu hoʻohana no ka mea, kia-stratification, piʻi pū, na assumptions e pono ai e kokua ia lilo lakou i poe kupono. Haawiia mai i keia mea, he mea e like me ka noiʻi e makemake ai e hana i ka nui helu o na pūʻulu no ka pou-stratification. Aka, e like me ka heluʻana i nā pūʻulu, piʻi pū, noiʻi holo i loko o kaʻokoʻa pilikia, aeaiiuo sparsity. Ina he ua wale no, he uuku o na kanaka iloko o kela a me keia hui, alaila, o na koho e oi ole lilo, a ma ka lōʻihi hihia a aia no he hui ia i aole respondents, alaila kia-stratification loa, mai hemo mai i lalo. Elua aoao mai loko mai o kēia hoʻomaopop 'ana i tension ma waena o ka plausibility o homogeneous- pane-propensity-loko-hui assumption, a me ke koi no ka mea, o ke hāpana nui like' ole ma kela hui. Kekahi kokoke loa i kahi mea e hoʻoneʻe aku i ka oi Ka Uluwehi O Ke ana helu kumu hoʻohālike no ka helu ana paona, a me kekahi mea e ohi i ka nui, oi aku ka nohona hāpana, ka mea e kokua hamama ai kupono hāpana nui like 'ole ma kela hui. A, i kekahi manawa, noiʻi hana elua, e like me kaʻu e kakau i oi aku au mamuli lalo nei.
A lua o ka noonoo i ka wa e hana ana me na kia-stratification ka politika, probability Eia kekahi laʻana o ia ka homogeneous-pane-propensity-loko-pūʻulu assumption ka mea, ua pinepine i ka wa a ke kālailai probability Eia kekahi laʻana hana. Ke kumu i keia assumption ua pono ai no ka probability Eia kekahi laʻana i ka hoʻomaʻamaʻa, ua i probability Eia kekahi laʻana i ole-pane, a me nā mea i na hana no ka hoʻomaʻa i ole pane he kia-stratification e like me ia i ho'ākāka 'ia ma luna. O ka holo ana, pono no ka mea, he nui noiʻi i kekahi assumption aʻole i manaʻo 'ia oukou e oi aku ka hana ia mea. Aka, aole aha i ka wa a ke kapakai ole-probability Eia kekahi laʻana e probability Eia kekahi laʻana ma ka hoʻomaʻamaʻa, kakou e malama i ka manao o ka mea e hilinai ana ma assumptions a auxiliary ike i ka mea e paka koho. Ma ka hapanui o realistic napoo ana no hoi, aole he wale, aohe assumption-ole kokoke i kuhi.
Eia ke oki, ina oe mālama i kekahi olelo i like ko-i loko o ko kākou kumu hoʻohālike ka nele i uku-a lailaʻoukou pono i ke ano nawaliwali mamua o homogeneous-pane-propensity-loko-pūʻulu assumption. Ua hōʻike hewa, aole oe i pono e kuhi na mea a pau i ka ia pane propensity, oe wale no pono, e kuhi ia ka mea, aole correlation mawaena pane propensity, a me ka nele i ana i loko o nä hui. O ka papa, a hiki i keia nawaliwali Penei e, aole noho i loko o kekahi mau wahi, i hiki. No ka mea, laʻana, i noonoo ai i aayoaeuiinoe i ka like ana o ka Americans e hana i manawaleʻai hana. Ina ua oi ao ao e ae e noho ma loko o ke anamanaʻo poe kanaka e hana manawaleʻai hana, alaila, noiʻi e systematically ma luna-kiÿeleo i ke dala o ka hana manawaleʻa, a ina e hana lakou kia-stratification ka hoʻoponopono ', he hopena ia i hōʻike empirically e Abraham, Helms, and Presser (2009) .
E like me kaʻu i olelo mua aku, ua nana aku la au ole-probability Eia kekahi laʻana me ka nui skepticism ma ka lawelaweʻepekema, ma ka hapa no ko lākou kūlana ma loko o kekahi o nā mea hilahila ihola hoʻolako 'ole i nā lā kakahiaka o ka anamanaʻo noiʻi. A ole mamuli o ka mamao makou i hele me ka ole-probability Eia kekahi laʻana o ka noiʻi a Wei Wang, David Rothschild, Sharad Goel, a me Andrew Gelman ka mea pono hou ai i ka hope o ka 2012 US koho me ka pili-probability hāpana o American Xbox mea hoʻohana -a decidedly ole-kaulele hāpana o Americans (Wang et al. 2015) . I ka noiʻi pūʻali koa i mea respondents mai ka XBox gaming nenoaiu, a me oukou ke kali, ke Xbox hāpana skewed kane, a skewed opiopio: 18 - 29 makahiki Nalani Olds e ahu iho ai 19% o ka electorate aka, 65% o ka Xbox hāpana, a me na kanaka e ae 47% o ka electorate a me ka 93% o ka Xbox hāpana (Kii 3.4). Ma muli o kēia mau ikaika HI biases, i ka maka Xboxʻikepili he ilihune anakahi o ke koho ana mai. He wānana ka lanakila ikaika no Mitt Romney ma luna o Barack Obama. Hou, o keia no kekahi hana no nā pilikia o ka maka, unadjusted ole probability Eia kekahi laʻana, a ua reminiscent o ka palapala, ai 'hōʻuluʻulu manaʻo fiasco.
Eia naʻe, Wang, a me nā hoapili uaʻike nōʻo ia ia mau pilikia, a ho'āʻo e kaupaonaʻana i nā respondents e hooponopono no ka mea, o ka mea hōʻikeʻikeʻuʻuku kaʻina. I like ko, lakou hoʻohana i oi aku Ka Uluwehi O Keʻano o ka kia-stratification aʻu i hai aku ia oukou no. Pono ke kuai aʻo 'iki hou e pili ana i ko lakou kokoke loa i kahi, no ka mea, kükulu kilo i kia-stratification, a me ka hana mana Wang, a me nā hoapili i hoʻohana oia kekahi o na mea hoʻopīhoihoi, hoʻokokoke mai i kaumaha ole-probability Eia kekahi laʻana.
I ko kakou mea hoʻohālike e pili ana aayoaeuiinoe ka nele i loko o ka Pauku 3.4.1, makou i māhele i ka heluna kanaka i loko o nā pūʻulu nānā 'ana i kahi o ka noho. Akä naÿe, Wang, a me nā hoapili māhele i ka heluna kanaka i loko o 176.256 pūʻulu ho'ākāka 'ia e: hoohanau (2 waeʻano), lāhui (4 waeʻano), makahiki (4 waeʻano), ka hoʻonaʻauao (4 waeʻano), moku'āina (51 waeʻano), aoao ID (3 waeʻano), ideology (3 waeʻano) a me 2008 koho (3 waeʻano). Me ka hou hui, ka manao i ka noiʻi i mea e increasingly e kiʻi mai ia i loko o kela a me keia hui, pane propensity ua uncorrelated me ka kākoʻo no Obama. A malaila, aole kūkulu kanaka-pae lua, e like me makou i hana ai i loko o ko makou laʻana, Wang, a me nā hoapili hoʻohana i ka luna 'kükohu ke hoʻomaopopoʻia ka nui o na kanaka ma kela hui ana e koho balota no ka mea, Obama. Eia ke oki, lakou i huiʻia kēia hui koho o kākoʻo, me ka ike ka nui o kela a me keia hui e paka i ka manaʻo nui o ka nohona ma kākoʻo. Ma na olelo e, kua iho la lakou i ka heluna kanaka i okoa hui, aayoaeuiinoe i ka kākoʻo no Obama, ma kela a me keia hui, a laila lawe aʻela ia i ke kaupaonaʻana awelika o ka hui koho e paka i ka pono olelo.
Pela, nui ka hana nui loa ia i loko o ko lākou kokoke loa i kahi o ke hoʻomaopopoʻia ka kākoʻo no Obama, ma kela a me keia o keia mau 176.256 pūʻulu. Eia nae ko lakou panel komo 345.858 kū hoʻokahi i komo, he nui helu ma ka hae o ke koho Wahi koho, he ua nui, he nui no ka hui a Wang, a me nā hoapili he aneane aohe respondents. Nolaila, ke hoʻomaopopoʻia ka kokua o kela a me keia hui lakou hoʻohana heʻano hana kapaia multilevel regression me ka lāʻau kū-stratification, a noiʻi aloha pumehana, kahea Mr. P.ʻano nui, ke hoʻomaopopoʻia ka kokua no Obama i loko o kekahi hui, Mr. P. loko 'ike mai ka nui pili nā pūʻulu. No ka mea, laʻana, noonoo i ka hana nui loa ia o aayoaeuiinoe i ka kākoʻo no Obama mawaena wahine, Hispanics, ma waena o 18-29 makahiki, ka poe kulanui puka, i kakau Democrats, i ka hoʻoponopono, e hōʻike ana e like me ka 'oʻoleʻa lōpū, a me ka poe i hooholoia no ka mea, Obama i ka 2008. keia he loa, loa kekahi hui, a he mea hiki iaʻaʻohe mea nāna i loko o ka hāpana me keia ano. Nolaila, e hana i koho a puni keia pae, Mr. P. loko pu koho mai nā kānaka i loko o ua aneane like hui.
E ho ohana i kēia Ka Ikepili ka hoʻolālā, he Wang, a me nā hoapili hiki ke hoʻohana i ka XBox ole-probability hāpana e ikaika mai no ua manaʻo i ka pono kākoʻo a Obama i loaa iloko o ka 2012 koho (Kii 3.5). Ma ka mea o ko lakou koho ua oi aku ka pololei ma mua o ka 'Ohi o ka lehulehu manao mau poo. Pela, ma keia hihia, kaumaha, ua hōʻike hewa Mr. P.-mea e hana i ka hana maikai? Ee i na biases i ole-probabilityʻikepili; biases ka poe i nanaia ka wa a oukou e nana i ka poe koho, mai ka unadjusted Xbox ikepili.
He nui no ka papa haʻawinaʻelua, mai ka hoʻopaʻa 'ana o Wang, a me nā hoapili. Mua, unadjusted ole probability Eia kekahi laʻana ke alakai i kaʻino koho; ka mea, he haʻawina i, ua lohe mua na noiʻi. Eia naʻe,ʻo ka lua o ka haʻawina, he mea ole-probability Eia kekahi laʻana, i ka wa kaumaha pono, e hiki nae paka lā maikaʻi koho. I mea, i ko lakou koho ua oi aku ka pololei mamua o na koho mai pollster.com, he aggregation o ka oi kuʻuna koho balota.
Eia ke oki, he ua nui hoʻokau i ka mea a makou e aʻo mai i keia mea i ho'ākāka 'mahele. Wale no kia-stratification hana maikai ma keia ua neo, aole he kumu ia mea, e hana maikai iloko o na hihia. In mea, ke koho nei paha kekahi o na easiest napoo ana no hoi no ka pollsters i këia koho ana i aneane 100 makahiki, he mau manaʻo (kakou ke ike i Amelika Huipû i na koho ana), a me ka 'aoʻao' ike a me ka HI ano he kŘpa a predictive o ke koho balota. Ma keia wahi, makou nele paa kumumanaʻo o ka empirical ka hoao ana i ka wa kaumaha hoʻololi 'ana i' ole-probability Eia kekahi laʻana e paka pono pololei koho e ike. Kekahi mea i mea maopopo, naʻe, ua inā e hāʻawi pio e hana me ka 'ole-probability Eia kekahi laʻana, alaila, ua ikaika kumu e manaoio i hoʻoponopono iho koho, e aho mamua o ka politika-hoʻoponopono iho koho.