Ayyukan

key:

  • mataki na wahala: sauki sauki , matsakaici matsakaici , m wuya , Da wuya wuya
  • bukatar ilimin lissafi ( bukatar ilimin lissafi )
  • bukatar coding ( bukatar coding )
  • data tarin ( data tarin )
  • ta favorites ( ta fi so )

  1. [ wuya , bukatar ilimin lissafi ] A cikin babi na, na sosai tabbatacce game post-stratification. Duk da haka, ba ko da yaushe inganta ingancin kimomi. Gina wani halin da ake ciki inda za a iya post-stratification iya rage ingancin kimomi. (Ga wani ambato, gani Thomsen (1973) ).

  2. [ wuya , data tarin , bukatar coding ] Design da kuma gudanar da maras yiwuwa binciken a kan Amazon MTurk su tambaye game da gun dukiya ( "Shin, kai, ko aikata wani a cikin iyali, mallaka a gun, bindiga ko bindiga? Shin da ka, ko kuma wani a cikin iyali?") Da halaye ga gun iko ( "Me kuke zaton shi ne mafi muhimmanci-kare hakkin Amirkawa ya mallaki bindigogi, ko don sarrafa gun dukiya?").

    1. Har yaushe ka duba sama? Nawa ne kudin? Ta yaya demographics na sample kwatanta da demographics na US yawan?
    2. Mene ne raw kimanta na gun dukiya ta amfani da samfurin?
    3. Correct ga wadanda ba representativeness na sample amfani post-stratification ko wasu m. Yanzu abin da ke cikin kimanta na gun dukiya?
    4. Ta yaya ka kimomi kwatanta da latest kimanta daga benci Research Center? Me kuke tunani bayyana discrepancies, idan akwai wani?
    5. Maimaita motsa jiki 2-5 ga halaye wajen gun iko. Ta yaya your binciken bambanta?

  3. [ wuya , data tarin , bukatar coding ] Goel da kuma abokan aiki (2016) an gudanar da wani maras yiwuwa na tushen duba kunshi 49 mahara-zabi halaye tambayoyi kõma daga Gaba Social Survey (GSS) kuma zaži safiyo da benci Research Center a Amazon MTurk. Sai suka daidaita ga wadanda ba representativeness bayanai ta amfani da model na tushen post-stratification (Mr. P), da kuma kwatanta gyara kimomi da waɗanda kiyasta amfani yiwuwa tushen GSS / benci safiyo. Gudanar da wannan binciken a kan MTurk da kuma kokarin rubanya Figure 2A da Figure 2b by gwada your gyara kimomi da kimomi daga mafi yawan 'yan harsasai GSS / benci (Dubi Shafi Table A2 ga jerin 49 tambayoyi).

    1. Kwatanta da bambanci your sakamakon da sakamakon daga benci kuma GSS.
    2. Kwatanta da bambanci your sakamakon da sakamakon daga MTurk binciken a Goel, Obeng, and Rothschild (2016) .

  4. [ matsakaici , data tarin , bukatar coding ] Kuma da yawa karatu amfani kai rahoton da matakan wayar hannu aiki data. Wannan shi ne mai ban sha'awa wuri inda masu bincike za a iya kwatanta kai ruwaito hali da Logged hali (duba misali, Boase and Ling (2013) ). Biyu na kowa halayyar su tambaye game da kiran da texting, biyu kuma na kowa lokaci Frames ne "jiya" da kuma "a makon da ya wuce."

    1. Kafin tattara wani data, abin da na kai rahoto matakan da kake tunani ne mafi m? Me ya sa?
    2. Kurtu 5 of your friends to ya kasance a cikin binciken. Don Allah a takaicce takaita yadda wadannan 5 abokai aka dauka samfur. Iya wannan Samfur hanya sa takamaiman biases a cikin kimomi?
    3. Don Allah ka tambaye su wadannan micro-binciken:
    • "Sau nawa kuka yi amfani da wayar hannu a kira wasu jiya?"
    • "Da yawa saƙonnin rubutu ba ka aika jiya?"
    • "Sau nawa kuka yi amfani da wayarka ta hannu don kira wasu a karshe kwana bakwai?"
    • "Sau nawa yi ka yi amfani da wayarka ta hannu zuwa aika ko karba saƙonnin rubutu / SMS a karshe kwana bakwai?" Da zarar binciken ne complete, tambayar su duba su amfani data kamar shigad da su wayar ko mai bada sabis.
    1. Ta yaya kai rahoton amfani kwatanta shiga data? Wanne ne mafi m, wanda shi ne m m?
    2. Yanzu hada da bayanai da ka tattara, da tare da bayanai daga wasu mutane a cikin aji (idan kana yin wannan aiki ga wani aji). Da wannan ya fi girma dataset, maimaita part (d).

  5. [ matsakaici , data tarin ] Schuman da Presser (1996) bayar da hujjar cewa tambaya umarni zai kome for iri biyu tsakanin tambayoyi: part-part tambayoyi inda tambayoyi biyu ne a wannan matakin na bayani dalla-dalla (eg ratings biyu shugaban 'yan takara); kuma part-duka tambayoyi inda a general tambaya ya bi wani more musamman tambaya (eg tambayar "Ta yaya gamsu ne ka tare da aiki?" bi da "Ta yaya gamsu ne ka tare da rai?").

    Suka kara faye hali iri biyu tambaya domin sakamako: daidaito effects faruwa lokacin da martani ga wani daga baya tambaya aka kawo kusa (daga gare su zai zama in ba haka ba) zuwa ga waɗanda aka bai wa wani a baya tambaya. bambanci effects faruwa lokacin da akwai mafi girma bambance-bambance tsakanin martani ga tambayoyi biyu.

    1. Create a biyu daga kashi-kashi tambayoyi da ka yi tunanin zai yi babban tambaya domin sakamako, a biyu daga kashi-duka tambayoyi da ka yi tunanin zai yi babban tsari sakamako, kuma wani biyu daga tambayoyi wanda tsari, kun yi zaton ba zai kome. Run wani binciken gwaji a kan MTurk to jarraba ku tambayoyi.
    2. Ta yaya babba ne part-part sakamako kuka kasance iya halitta? Was shi a daidaito ko bambanci sakamako?
    3. Ta yaya babba ne part-dukan sakamako kuka kasance iya halitta? Was shi a daidaito ko bambanci sakamako?
    4. Was akwai wata tambaya domin sakamako a cikin biyu, inda za ka ba tsammani domin zai kome?

  6. [ matsakaici , data tarin ] Gina a kan aikin Schuman da Presser, Moore (2002) ya bayyana a raba girma Tambaya domin sakamako: ƙari kuma subtractive. Duk da yake da bambanci da daidaito effects ake samar a matsayin sakamako na weights 'kimantawa na biyu abubuwa dangane da juna, ƙari kuma subtractive effects ake samar a lõkacin da weights an sanya mafi m ga ya fi girma tsarin cikin abin da tambayoyi shirya kai. Read Moore (2002) , to, tsara da kuma gudanar da wani binciken da gwaji a kan MTurk don nuna ƙari ko subtractive effects.

  7. [ wuya , data tarin ] Christopher Antoun da kuma abokan aiki (2015) gudanar da wani binciken gwada saukaka samfurori samu daga hudu daban-daban online] aukar kafofin: MTurk, Craigslist, Google AdWords da Facebook. Design mai sauki binciken da kuma kurtu mahalarta ta akalla biyu daban-daban online] aukar kafofin (za su iya zama daban-daban kafofin daga hudu kafofin amfani da Antoun et al. (2015) ).

    1. Kwatanta cost per kurtu, cikin sharuddan kudi da kuma lokaci, tsakanin daban-daban kafofin.
    2. Kwatanta abun da ke ciki na samfurori samu daga kafofi da dama.
    3. Kwatanta ingancin data tsakanin samfurori. Domin ra'ayi game da yadda za a auna data quality daga weights, gani Schober et al. (2015) .
    4. Mene ne ka fi so source? Me ya sa?

  8. [ matsakaici ] YouGov, an internet na tushen kasuwa bincike m, gudanar online zabe mai panel game da 800,000 weights a Birtaniya da kuma amfani da Mr. P. hango ko hasashen sakamakon EU raba gardama (ie, Brexit) inda Birtaniya jefa ƙuri'a zabe ko dai ya kasance a ko barin Tarayyar Turai.

    A cikakken bayanin YouGov ta ilimin kididdiga model ne a nan (https://yougov.co.uk/news/2016/06/21/yougov-referendum-model/). Wajen magana, YouGov partitions masu jefa} uri'a cikin iri bisa 2015 general zaben kuri'a zabi, shekaru, cancantar, jinsi, ranar hira, kazalika da zasu suka zauna a. Na farko, da suka kasance sunã data tattara daga YouGov panelists to kimanta, daga waɗanda suka kada kuri'a, da rabo daga mutane na kowane zabe type suka yi nufin su zabe iznin. Suka kimanta fito kowane zabe type ta amfani da 2015 Birtaniya Zaben Nazarin (BES) post-zaben fuska-da-fuska binciken, wanda inganta fito daga za ~ e Rolls. A karshe, suka kimanta yadda mutane da yawa akwai kowane zabe type a za ~ e bisa latest Census da Annual Population Survey (tare da wasu kari bayani daga BES, YouGov binciken bayanai daga a kusa da babban zabe, da kuma bayani a kan yadda mutane da yawa zabe kowace qungiya a kowace mazabarta).

    Kwana uku kafin zaben, YouGov nuna wani biyu aya gubar for iznin. A Hauwa'u na zabe, da zabe ya nuna ma kusa kira (49-51 zauna). A karshe on-da-rana binciken annabta 48/52 a cikin ni'imar zauna (https://yougov.co.uk/news/2016/06/23/yougov-day-poll/). A gaskiya, wannan kimanta rasa karshe sakamakon (52-48 iznin) by hudu yawan maki.

    1. Yi amfani da total binciken kuskure tsarin tattauna a wannan babi tantance abin da zai iya yi tafi ba daidai ba.
    2. YouGov ta mayar da martani bayan zaben (https://yougov.co.uk/news/2016/06/24/brexit-follows-close-run-campaign/) ya bayyana cewa: "Wannan alama a cikin wani babban ɓangare saboda fito - wani abu da mun ce duk tare zai zama da muhimmanci ga sakamako na irin wannan finely daidaita tseren. Our fito model aka tushen, a bangare, a kan ko weights ya zabe a karshe general zaben da fito matakin sama na babban zabe tada model, musamman a Arewa. "Shin, wannan canza amsar part (a)?

  9. [ matsakaici , bukatar coding ] Rubuta kwaikwaiyo misalta kowane daga cikin misali kurakurai a Figure 3.1.

    1. Create a halin da ake ciki inda wadannan kurakurai zahiri sake fita.
    2. Create a halin da ake ciki inda kurakurai fili juna.

  10. [ wuya , bukatar coding ] A binciken da na Blumenstock da kuma abokan aiki (2015) da hannu gina na'ura ilmantarwa model da zai iya amfani da digital alama data hango ko hasashen binciken martani. Yanzu, za ka yi kokarin wannan abu da daban-daban dataset. Kosinski, Stillwell, and Graepel (2013) gano cewa Facebook likes iya hango ko hasashen mutum halaye da kuma halaye. Abin mamaki, wadannan tsinkaya zai iya zama ma fi m fiye da wadanda na abokai da abokan aiki (Youyou, Kosinski, and Stillwell 2015) .

    1. Read Kosinski, Stillwell, and Graepel (2013) , da kuma rubanya Figure 2. Su bayanan suna samuwa a nan: http://mypersonality.org/
    2. Yanzu, rubanya Figure 3.
    3. A karshe, kokarin da model a kan kansa Facebook data: http://applymagicsauce.com/. Ta yaya da kyau yake aiki a gare ku?

  11. [ matsakaici ] Toole et al. (2015) amfani kira daki-daki records (CDRs) daga mobile phones hango ko hasashen tara rashin aikin yi trends.

    1. Kwatanta da bambanci da zane na Toole et al. (2015) da Blumenstock, Cadamuro, and On (2015) .
    2. Kada ka yi tunanin CDRs ya kamata ya maye gurbin na gargajiya safiyo, gaba gare su, ko ba za a yi amfani a duk gwamnati policymakers zuwa waƙa rashin aikin yi? Me ya sa?
    3. Abin da shaida zai shawo ku cewa CDRs iya gaba daya maye gurbin gargajiya matakan da rashin aikin yi kudi?

Powered by Open Review Toolkit