Bugu da ari, sharhin

Wannan sashe da aka tsara za a yi amfani a matsayin tunani, maimakon a karanta a matsayin labari.

• Gabatarwa (Sashe 3.1)

Mutane da yawa daga cikin jigogi a cikin wannan sura sun kuma an echoed a cikin 'yan Presidential imel a Amirka Association of Public Nazarin labarai Research (AAPOR), kamar Dillman (2002) , Newport (2011) , Santos (2014) , da kuma Link (2015) .

Don ƙarin tarihi baya game da ci gaban da binciken bincike, ganin Smith (1976) da kuma Converse (1987) . Don ƙarin a kan ra'ayin uku yanayi dazamani na binciken bincike, ganin Groves (2011) da kuma Dillman, Smyth, and Christian (2008) (wanda ya karya up uku yanayi dazamani dan kadan daban).

A ganiya cikin miƙa mulki daga na farko zuwa na biyu zamanin a duba bincike ne Groves and Kahn (1979) , wanda ya aikata wani cikakken kai-da-kai kwatanta tsakanin fuska-da-fuska da tarho binciken. Brick and Tucker (2007) dubi baya a tarihi ci gaba da bazuwar lambar bugun kira daukan samfur hanyoyin.

Don ƙarin yadda binciken bincike ya canza a baya a mayar da martani ga canje-canje a cikin al'umma, ga Tourangeau (2004) , Mitofsky (1989) , da kuma Couper (2011) .

• Tambayar vs. lura (Sashe 3.2)

Koyo game da ciki jihohi da tambayoyi zai iya zama matsala, domin wani lokacin da weights da kansu ba su sani su na ciki jihohi. Alal misali, Nisbett and Wilson (1977) da ban mamaki takarda da evocative title: "Bayyana fiye da za mu iya sani: fi'ili rahotanni a kan shafi tunanin mutum matakai." A cikin takarda marubuta gama: "batutuwa ne, wani lokacin (a) m na kasancewar a kara kuzari da muhimmanci rinjayi wata amsa, (b) m na zama na mayar da martani, kuma (c) m cewa mai kara kuzari ya shafi mayar da martani. "

Domin muhawara cewa masu bincike ya kamata fi son lura hali to ya ruwaito hali ko halaye, gani Baumeister, Vohs, and Funder (2007) (tunani) da kuma Jerolmack and Khan (2014) da martani (Maynard 2014; Cerulo 2014; Vaisey 2014; Jerolmack and Khan 2014) (ilimin halayyar zaman jama'a). Bambanci tsakanin tambayar da lura kuma taso a tattalin arziki, inda masu bincike magana game da bayyana da kuma saukar da zaɓin. Alal misali, wani mai bincike zai iya tambayar weights ko sun fi son cin ice cream ko je dakin motsa jiki (ya bayyana fifiko) ko bincike zai iya lura da yadda sau da yawa mutane ci ice cream da kuma je dakin motsa jiki (saukar da zaɓin). Akwai zurfi shakka daga wani iri bayyana fifiko data tattalin arziki (Hausman 2012) .

A main theme daga wadannan muhawara shi ne cewa ya ruwaito hali ba ko da yaushe m. Amma, ta atomatik rubuta hali bazai m, mai yiwuwa ba za a tattara a kan wani samfurin ban sha'awa, kuma mai yiwuwa ba a damar zuwa masu bincike. Saboda haka, a wasu yanayi, ina ganin cewa ya ruwaito hali zai iya zama da amfani. Bugu da ari, a karo na biyu main theme daga wadannan muhawara shi ne cewa rahotanni game da motsin zuciyarmu, da ilmi ba, tsammanin, da kuma ra'ayoyin da aka ba su ko da yaushe m. Amma, idan bayani game da wadannan ciki jihohin da ake bukata da bincike-dai ya taimake bayyana wasu hali ko kuma a matsayin abu da za a bayyana-to tambayar iya zama ya dace.

• Total binciken kuskure (Sashe 3.3)

Domin littafin tsawon jiyya a total binciken ɓata, ga Groves et al. (2009) ko Weisberg (2005) . Ga wani tarihin ci gaban total binciken ɓata, ga Groves and Lyberg (2010) .

A cikin sharuddan misali, mai girma, gabatarwa ne ga al'amurran da suka shafi wadanda ba mayar da martani da kuma wadanda ba amsa nuna bambanci ne National Research Council rahoto kan Nonresponse a Social Science Safiyo: A Research Tsari (2013) . Wani amfani Siffar da aka bayar da (Groves 2006) . Har ila yau, dukan musamman al'amurran da suka shafi na Journal of Official Statistics, Public Nazarin labarai kwata, da kuma The Annals na Amirka, Academy of siyasa da kuma Social Science an buga a topic wadanda ba amsa. A karshe, akwai zahiri da yawa hanyoyi daban-daban na kirga mayar da martani rate. wadannan hanyoyin an aka bayyana a cikin daki-daki, a cikin wani rahoto da Amurka Association of Public Nazarin labarai Masu bincike (AAPOR) (Public Opinion Researchers} 2015) .

The 1936 Literary Digest zabe da aka yi nazari a cikin daki-daki (Bryson 1976; Squire 1988; Cahalan 1989; Lusinchi 2012) . An kuma an yi amfani da matsayin misali ka yi gargaɗi da taragutsan data tarin (Gayo-Avello 2011) . A 1936, George Gallup amfani da more sophisticated nau'i na daukan samfur, kuma ya iya samar da karin m kimomi da yawa karami samfurin. Gallup ta nasara a kan Literary Digest wani milestone ci gaban binciken bincike (Converse 1987, Ch 3; Ohmer 2006, Ch 4; Igo 2008, Ch 3) .

A cikin sharuddan ji, mai girma na farko hanya domin zayyana questionnaires ne Bradburn, Sudman, and Wansink (2004) . Domin a more m jiyya mayar da hankali musamman a kan ra'ayi tambayoyi, ganin Schuman and Presser (1996) . More on pre-gwaji tambayoyi yana samuwa a cikin Presser and Blair (1994) , Presser et al. (2004) , da kuma Babi na 8 na Groves et al. (2009) .

The classic, littafi-tsawon lura da harkokin ciniki-kashe tsakanin binciken halin kaka da kuma duba kurakurai ne Groves (2004) .

• Wanda ya nẽmi (Sashe 3.4)

Classic littafin-tsawon lura da misali yiwuwa daukan samfur da hakkin su Lohr (2009) (more gabatarwa) da kuma Särndal, Swensson, and Wretman (2003) (more m). A classic littafin-tsawon jiyya na post-stratification da related hanyoyin ne Särndal and Lundström (2005) . A wasu digital shekaru saituna, masu bincike san quite a bit game da wadanda ba weights, wanda shi ne sau da yawa ba gaskiya a baya. Daban-daban siffofin wadanda ba amsa gyara mai yiwuwa ne a lokacin da masu bincike da bayani game da wadanda ba weights (Kalton and Flores-Cervantes 2003; Smith 2011) .

The Xbox nazarin Wang et al. (2015) yana amfani da wata dabara da ake kira multilevel komawa da baya, kuma post-stratification (MRP, wani lokacin ake kira "Mister P") da damar bincike don kimanta cell nufin ko da yaushe akwai mutane da yawa, mutane da yawa Kwayoyin. Ko da yake akwai wasu shawarwari game da ingancin da kimomi daga wannan dabara, ga alama kamar alamar yankin don gano. The dabara aka fara amfani da Park, Gelman, and Bafumi (2004) , da kuma a can ya kasance m amfani da muhawara (Gelman 2007; Lax and Phillips 2009; Pacheco 2011; Buttice and Highton 2013; Toshkov 2015) . Don ƙarin a kan alaka tsakanin mutum nauyi da cell tushen nauyi ga Gelman (2007) .

Don wasu hanyoyin weighting web safiyo, gani Schonlau et al. (2009) , Valliant and Dever (2011) , da kuma Bethlehem (2010) .

Sample matching aka samarwa da Rivers (2007) . Bethlehem (2015) ya bayar da hujjar cewa, wasan kwaikwayon na sample matching za a zahiri su yi kama da sauran daukan samfur fuskanci (misali, rabe daukan samfur) da kuma sauran gyara hanyoyin (misali, post-stratification). Don ƙarin on online bangarori, gani Callegaro et al. (2014) .

Wani lokaci masu bincike sun gano cewa, yiwuwar samun samfurori da kuma wadanda ba yiwuwa samfurori samar kimomi da irin wannan quality (Ansolabehere and Schaffner 2014) , amma sauran kwatancen sun gano cewa, ba yiwuwa samfurori yi muni (Malhotra and Krosnick 2007; Yeager et al. 2011) . Daya zai yiwu dalilin wadannan bambance-bambance ne cewa ba yiwuwa samfurori sun inganta a kan lokaci. Domin a more pessimistic view of maras yiwuwa daukan samfur hanyoyin ganin da AAPOR Task Force a Non-Yiwuwar Samfur (Baker et al. 2013) , da kuma na kuma bayar da shawarar karanta sharhin da cewa ya bi summary rahoton.

Ga wani meta-bincike a kan sakamako daga weighting don rage nuna bambanci a ba-Yiwuwar samfurori, ga Table 2.4 a Tourangeau, Conrad, and Couper (2013) , wadda take kaiwa marubuta mala "sabawa ze zama da amfani, amma fallible gyare-gyare. . . "

• Yadda za a tambaye (Sashe 3.5)

Conrad and Schober (2008) na samar da wani edited girma mai taken Hasashenmu na Survey Interview na Future, kuma bayani da yawa daga cikin jigogi a wannan sashe. Couper (2011) bayani m jigogi, kuma Schober et al. (2015) yayi wani m misali na yadda data tarin hanyoyin da aka kera wani sabon wuri zai iya haifar da mafi girma quality data.

Domin wani m misali na yin amfani da Facebook apps ga zamantakewa kimiyya safiyo, gani Bail (2015) .

Don ƙarin shawara a kan yin safiyo wani m da kuma muhimmanci kwarewa ga mahalarta, ganin aiki a kan wanda aka kera Design Hanyar (Dillman, Smyth, and Christian 2014) .

Stone et al. (2007) yayi wani littafi tsawon lura da muhalli na yanzu-yanzu kima da kuma alaka da hanyoyin.

• Safiyo nasaba da sauran bayanai (Sashe 3.6)

Judson (2007) ya bayyana aiwatar da hada safiyo da administrative data matsayin "bayanai hadewa," tattauna wasu abũbuwan amfãni daga wannan hanya, da kuma yayi da wasu misalai.

Wata hanya kuma da masu bincike za su iya yin amfani da digital burbushi da administrative data ne mai daukan samfur frame ga mutanen da tare da takamaiman halaye. Duk da haka, samun damar wadannan records da za a yi amfani da daukan samfur frame kuma iya ƙirƙirar tambayoyi da suka shafi bayanin tsare (Beskow, Sandler, and Weinberger 2006) .

Game da Littafi roƙa, wannan dabarar ne ba kamar yadda sabon matsayin shi zai bayyana daga yadda na taba bayyana shi. Wannan m na m sadarwa zuwa uku da manyan yankunan a statistics-model na tushen post-stratification (Little 1993) , imputation (Rubin 2004) , da kuma karamin yanki hakkin (Rao and Molina 2015) . Haka kuma an alaka da amfani da surrogate canji a likita bincike (Pepe 1992) .

Baya ga da'a al'amurran da suka shafi game da samun dama da digital alama data, Littafi roƙa za a iya amfani da su infer m halaye da cewa mutane su zabi ya bayyana a cikin wani binciken (Kosinski, Stillwell, and Graepel 2013) .

A kudin da lokaci kimomi a Blumenstock, Cadamuro, and On (2015) mai da more m cost-kudin daya ƙarin binciken-, kuma kada ku hada gyarawa halin kaka kamar cost domin tsabtace da kuma aiwatar da kira data. A general, Littafi roƙa zai yiwuwa da high gyarawa halin kaka da kuma low m halin kaka kama digital gwajen (duba Chapter 4). More cikakken bayani a kan data kasance a cikin Blumenstock, Cadamuro, and On (2015) takarda ne a Blumenstock and Eagle (2010) da kuma Blumenstock and Eagle (2012) . Halarci daga mahara imputuation (Rubin 2004) zai taimaka kama rashin tabbas a kimomi daga Littafi roƙa. Idan masu bincike aikatãwa Littafi tambayar kawai damu da tara kirga, maimakon mutum-matakin halaye, sa'an nan kuma yanki a King and Lu (2008) da kuma Hopkins and King (2010) na iya zama da amfani. Don ƙarin game da na'ura koyo fuskanci a Blumenstock, Cadamuro, and On (2015) , gani James et al. (2013) (more gabatarwa) ko Hastie, Tibshirani, and Friedman (2009) (more m). Wani rare na'ura koyo littafi ne Murphy (2012) .

Game wadãtar da roƙa, sakamakon a Ansolabehere da Hersh (2012) hinjis biyu key matakai: 1) da ikon na Catalist hada yawa disparate data kafofin samar da wani cikakken master datafile na 2 kuma) ikon da Catalist danganta da binciken bayanai zuwa da master datafile. Saboda haka, Ansolabehere da Hersh duba kowane daga cikin wadannan matakai a hankali.

Don ƙirƙirar master datafile, Catalist hadawa da jitu bayanai daga mutane da yawa daban-daban kafofin ciki har da: mahara zabe records tilas daga kowace jiha, data daga Post Office ta National Change na Address Registry, kuma bayanai daga wasu unspecified kasuwanci azurtawa. The gory cikakkun bayanai game da yadda wannan duka tsaftacewa da tattara abubuwa masu kyau da ya faru ne bayan da ikon yinsa, na wannan littafin, amma wannan tsari, komai da hankali, za propagate kurakurai a cikin ainihin data kafofin da zai gabatar kurakurai. Ko da yake Catalist ya shirye su tattauna da data aiki da kuma samar da wasu da raw data, shi ne kawai zai yiwu ba ga masu bincike da za a duba dukan Catalist data bututun. Maimakon haka, masu bincike sun kasance a halin da ake ciki inda Catalist data fayil da wasu unknown, kuma watakila unknowable, adadin ɓata. Wannan shi ne mai tsanani damuwa saboda wani mai zargi iya yayata cewa manyan bambance-bambance a tsakanin binciken rahotanni a kan CCES da hali a Catalist master data fayil aka lalacewa ta hanyar kurakurai a cikin master data fayil, ba ta misreporting da weights.

Ansolabehere da Hersh ɗauki biyu daban-daban hanyoyin magance data quality damuwa. Na farko, ban da gwada kai ruwaito zabe to kada kuri'a a Catalist master file, da masu bincike kuma idan aka kwatanta da kai ruwaito jam'iyyar, tseren, masu jefa} uri'a rajista status (misali, rajista ko ba rajista) da kuma zabe hanya (misali, a cikin mutum, absentee takardar za ~ en, da dai sauransu) zuwa ga waɗanda dabi'u samu a cikin Catalist databases. Don wadannan hudu alƙaluma canji, da masu bincike gano da yawa mafi girma matakan yarjejeniya tsakanin binciken rahoton da data a Catalist master fayil fiye da zabe. Saboda haka, Catalist master data fayil bayyana a yi high quality bayanai ga halaye wanin zabe, bayar da shawara cewa shi ne, ba matalauta overall quality. Na biyu, a sashi amfani da bayanai daga Catalist, Ansolabehere da Hersh raya uku daban-daban matakan da ingancin County zabe records, kuma suka gano cewa, da kiyasta kudi na kan-rahoto na zabe shi ne da gaske da alaqa ga wani daga cikin wadannan bayanai quality matakan, a binciken da bayar da shawarar cewa babban rates na kan-rahoto ba su anã kõra da kananan hukumomi da unusually low data quality.

Ganin halittar wannan master zabe fayil, na biyu tushen m kurakurai da aka cudanya da binciken records da shi. Alal misali, idan wannan hada huldodi ne yake aikata kuskure shi zai iya kai wa ga wani a kan-kimanta na bambanci tsakanin ruwaito da inganta zabe hali (Neter, Maynes, and Ramanathan 1965) . Idan kowane mutum yana da barga, musamman ganowa da yake a duka data kafofin, to hada huldodi zai zama maras muhimmanci. A Amurka da mafi yawan sauran kasashen, duk da haka, babu wani a duniya ganowa. Bugu da ari, ko da akwai irin wannan ganowa mutane za yiwuwa ya zama komo acikin don samar da shi a nazarin bincike! Saboda haka, Catalist ya yi da hada huldodi amfani ajizai identifiers, a cikin wannan harka hudu guda da bayanai game da kowace wanda ake kara: name, jinsi, haihuwa shekara, da kuma gida address. Alal misali, Catalist ya shirya idan Homie J Simpson a CCES ne guda mutum a matsayin Homer Jay Simpson a master data fayil. A yi, matching ne mai wuya da kuma m tsari, da kuma, a yi al'amura muni ga masu bincike, Catalist dauke da daidai da dabara ya zama mallakar tajirai.

Domin inganta da matching Algorithms, sun dogara a kan biyu kalubale. Na farko, Catalist halarci wani matching gasar da aka gudanar da wani zaman kanta, na uku-jam'iyyar: da MITRE Corporation. MITRE tanadar da dukkan mahalarta biyu m data files da za a dace, kuma daban-daban teams gasar komawa MITRE mafi kyau iri daya. Saboda MITRE kanta san daidai daidai da sun kasance iya score da teams. Daga cikin 40 kamfanoni da cewa gasar, ya zo na biyu Catalist wuri. Wannan irin m, na uku-jam'iyyar kimantawa mallakar tajirai fasaha ne quite rare da wuce yarda m. shi ya kamata ba mu tabbaci cewa Catalist ta matching hanyoyin ne da gaske a jihar-of-da-art. Amma shi ne jihar-of-da-art kyau isa? Baya ga wannan matching gasar, Ansolabehere da Hersh halitta nasu iri daya kalubale ga Catalist. Daga baya aikin, Ansolabehere da Hersh ya tattara masu jefa} uri'a records daga Florida. Suka bayar da wasu daga cikin wadannan records tare da wasu daga gonakinsu redacted to Catalist sa'an nan idan aka kwatanta Catalist ta rahotanni daga cikin wadannan filaye don su ainihin dabi'u. Abin farin, Catalist ta rahotanni sun kusa da kange dabi'u, na nuna cewa Catalist iya daidaita m zabe records uwa su master data fayil. Wadannan biyu kalubale, daya da na uku-jam'iyyar da kuma daya daga Ansolabehere da Hersh, ba mu more amincewa a Catalist matching Algorithms, ko da yake ba za mu iya duba su ainihin aiwatar da kanmu.

Akwai mutane da yawa sun kasance m yunkurin inganta zabe. Domin wani bayyani na cewa wallafe-wallafe, gani Belli et al. (1999) , Berent, Krosnick, and Lupia (2011) , Ansolabehere and Hersh (2012) , da kuma Hanmer, Banks, and White (2014) .

Yana da muhimmanci a lura da cewa ko da yake a wannan harka masu bincike da aka karfafa da ingancin bayanai daga Catalist, wasu kimantawa na sayar da dillalai sun kasa m. Masu bincike sun gano matalauta quality lõkacin data daga wani binciken da wani mabukaci-fayil daga Marketing Systems Group (wanda kanta merged tare da bayanai daga uku samar: Acxiom, Experian, kuma InfoUSA) (Pasek et al. 2014) . Wato, da data fayil bai dace da binciken martani da cewa masu bincike za su zama daidai, da datafile ya bace data ga wani babban yawan tambayoyi, da kuma m data juna da aka dangantaka zuwa ruwaito binciken darajar (a cikin wasu kalmomi da m data kasance din , ba bazuwar).

Don ƙarin a kan rikodin hada huldodi tsakanin safiyo da administrative data, ga Sakshaug and Kreuter (2012) da kuma Schnell (2013) . Don ƙarin a kan rikodin hada huldodi a general, ga Dunn (1946) da kuma Fellegi and Sunter (1969) (tarihi) da kuma Larsen and Winkler (2014) (zamani). Similar fuskanci sun kuma an ci gaba a kimiyyar kwamfuta karkashin sunaye kamar data deduplication, misali ganewa, suna iri daya, Kwafin ganewa, da kwafi rikodin ganewa (Elmagarmid, Ipeirotis, and Verykios 2007) . Akwai kuma bayanin tsare tsare hanyoyin rikodin hada huldodi da ba su bukatar da watsa daga kaina gano bayanai (Schnell 2013) . Masu bincike a Facebook ɓullo da wani hanya zuwa probabilisticsly danganta su records to zabe hali (Jones et al. 2013) . wannan hada huldodi da aka yi kimanta wani gwaji da zan gaya muku game da a Babi na 4 (Bond et al. 2012) .

Wani misali na cudanya a manyan-sikelin zamantakewa binciken gwamnati administrative records zo daga Lafiya da ritaya Survey da Social Security Administration. Don ƙarin a binciken, ciki har da bayanai game da amsa hanya, ga Olson (1996) da kuma Olson (1999) .

The tsari na hada mutane da yawa samo administrative records a cikin wani master datafile-da tsari da Catalist ma'aikata-ne na kowa a cikin ilimin kididdiga ofisoshin wasu gwamnatocin} asashensu. Biyu masu bincike daga Statistics Sweden sun rubuta cikakken littafi a kan topic (Wallgren and Wallgren 2007) . Domin misalin wannan dabarar a cikin wani aure County a Amurka (Olmstead County, Minnesota, gida na Mayo Clinic), gani Sauver et al. (2011) . Don ƙarin a kan kurakurai da za su iya bayyana a administrative records, gani Groen (2012) .