3.3.1 wakilci
Misali ne game da yin inferences daga weights to your manufa yawan.
Domin ya fahimci irin kurakurai da cewa zai iya faruwa a lokacin da inferring daga weights ga ya fi girma yawan, bari mu yi la'akari da Literary Digest bambaro zabe cewa kokarin hango ko hasashen sakamakon da 1936 US zaben shugaban kasa. Ko da yake shi ne fiye da 75 da suka wuce, wannan debacle har yanzu yana da wani muhimmin darasi don ya koya masu bincike a yau.
Literary Digest wani m general-sha'awa mujallar, da kuma lokacin da na fara a 1920 suka fara gudu bambaro Polls to hango ko hasashen sakamakon zaben shugaban kasa. Don yin wadannan tsinkaya za su aika kuri'un to kuri'a na mutane, sa'an nan kuma kawai Tally sama da kuri'un da aka mayar da. Literary Digest alfahari ruwaito cewa kuri'un da suka samu da aka ba "mai nauyi, gyara, ko fassara." Wannan hanya daidai annabta da lashe na zaben a 1920, 1924, 1928 da kuma 1932. a 1936, a tsakiyar Great mawuyacin, Literary Digest aika kuri'un zuwa 10 da mutane miliyan, waɗanda sunayensu predominately zo daga tarho kundayen da mota rajista records. Ga yadda suke bayyana su hanya:
"THE Digest ta m-guje na'ura motsa tare da gaggãwar daidaici na shekara talatin 'kwarewa don rage guesswork wuya facts. . . .Wannan Mako 500 alkalama tõno fitar da fiye da kwata na miliyan adiresoshin a rana. Kowane rana, a cikin wani babban dakin high sama motor-ribboned Fourth Avenue, a New York, 400 ma'aikata deftly slide miliyan guda na buga al'amari-isa ya sa a kan arba'in birnin tubalan-cikin jawabi ƙunsa [sic]. Kowane hour, a cikin Digest'S kansa Post Office Substation, uku chattering tambur metering inji shãfe haske da kuma tattake farin oblongs. gwani akwatin gidan ma'aikata flipped su cikin bulging mailsacks. rundunar nike motoci sped su bayyana mail-jiragen kasa. . . Next mako, na farko da amsoshi daga wadannan miliyan goma zai fara da mai shigowa tide na alama kuri'un, ya zama sau uku-bari, tabbatar, biyar-sau giciye-tallace da kuma totaled. A lokacin da na karshe adadi da aka totted kuma bari, idan da kwarewa mai rarrabẽwa, kasar za su san su a cikin ƙunshi juzu'i na 1 bisa dari da ainihin m kuri'a arba'in miliyan [masu jefa} uri'a. "(Agusta 22, 1936)
The Digest ta fetishization na size ne nan take recognizable ga wani "babban data" bincike a yau. Daga cikin miliyan 10 kuri'un rarraba, mai ban mamaki da miliyan 2.4 kuri'un da aka mayar da-ke wajen 1,000 sau ya fi girma fiye zamani siyasa zaben. Daga wadannan miliyan 2.4 weights da hukunci ya bayyana: Literary Digest annabta cewa abokin hamayyarsa Alf Landon aka je kayar da tabbatacce Franklin Roosevelt. Amma, a gaskiya, da kishiyar ya faru. Roosevelt ya ci Landon a gangarar ƙasa. Ta yaya za Literary Digest tafi daidai ba tare da sosai data? Our zamani fahimtar daukan samfur sa Literary Digest ta kurakurai bayyanannu, kuma yana taimaka mana mu guji yin irin wannan kurakurai a nan gaba.
Tuna a fili game da daukan samfur na bukatar da mu yi la'akari hudu daban-daban kungiyoyin mutane (Figure 3.1). Na farko rukuni na mutane ne manufa yawan. wannan shi ne rukuni cewa bincike ya kira a matsayin yawan amfani. A cikin hali na Literary Digest manufa yawan ya jefa ƙuri'a a 1936 Zaben Shugaban Kasa. Bayan yankan shawara a kan wani manufa yawan, wani mai bincike na gaba bukatar ci gaba da jerin mutanen da za a iya amfani da shi don daukan samfur. Wannan jerin ake kira a daukan samfur frame da yawan a kan daukan samfur frame ake kira frame yawan. A cikin hali na Literary Digest da firam yawan shi ne 10 da mutane miliyan waɗanda sunayensu zo predominately daga tarho kundayen da mota rajista records. Fi dacewa da manufa yawan jama'a da kuma firam yawan zai zama daidai da wannan, amma a yi wannan shi ne sau da yawa ba haka al'amarin. Bambance-bambance tsakanin manufa yawan jama'a da kuma yawan frame an kira ɗaukar hoto ɓata. Ɗaukar hoto kuskure ba, da kanta da tabbacin matsaloli. Amma, idan mutãne, a cikin frame yawan masu tsare daban-daban daga mutane ba a cikin frame yawan za a yi ɗaukar hoto nuna bambanci. Ɗaukar hoto kuskure ne na farko daga cikin manyan flaws da Literary Digest zabe. Suna so su koya game da masu jefa} uri'a-wannan shi ne manufa yawan-sai suka gina a daukan samfur frame predominately daga tarho kundayen da mota registries, kafofin cewa a kan-wakilta arziki Amirkawa suke mafi kusantar su goyi bayan Alf Landon (yi tunani cewa biyu daga cikin wadannan fasahar, waxanda suke da na kowa a yau, sun gwada sabon a lokaci da cewa Amurka da ke a cikin tsakiyar Great mawuyacin).
Adadi 3.1: wakilci kurakurai.
Bayan fassara da firam yawan, mataki na gaba shi ne don a bincike don zaɓar sample yawan. wadannan su ne mutanen da mai bincike zai ƙoƙari ya yi tambayoyi. Idan sample yana halaye daban-daban fiye da frame yawan, sa'an nan kuma za mu iya gabatar da daukan samfur ɓata. Wannan shi ne irin kuskure quantified a gefe na kuskure da yakan accompanies kimomi. A cikin yanayin da Literary Digest fiasco, akwai zahiri ba sample. suka yi yunkurin tuntube kowa da kowa a cikin frame yawan. Ko da yake babu daukan samfur ɓata, akwai fili har yanzu kuskure. Wannan ya bayyana cewa ribace-ribace na kurakurai da cewa suna yawanci ruwaito da kimomi daga safiyo yawanci misleadingly kananan. ba su hada da dukan samo ɓata.
A karshe, wani mai bincike ƙoƙarin tambayoyi kowa da kowa a cikin samfurin yawan. Wadanda mutane cewa an samu nasarar yi hira an kira weights. Fi dacewa, da samfurin yawan da weights zai zama daidai da wannan, amma a yi a can ne ba amsa. Wancan ne, mutanen da suke zaba a sample ƙi shiga. Idan mutane suka karɓa ne daban-daban, daga waɗanda ba su amsa, to, akwai iya zama wadanda ba amsa nuna bambanci. Non-amsa nuna bambanci na biyu main matsala tare da Literary Digest zabe. Only 24% na mutanen da suka karbi takardar za ~ en amsa, kuma ya juya waje cewa mutane da suka taimake Landon kasance mafi kusantar su amsa.
Beyond kawai kasancewa wani misali gabatar da ra'ayoyi da misali, da Literary Digest zabe ne ake maimaita misãli, jawo hankalin masu bincike game da hatsarori da taragutsan daukan samfur. Abin baƙin ciki, na yi tunanin cewa da darasi da cewa mutane da yawa zana daga wannan labarin ne da ba daidai ba daya. Mafi na kowa halin kirki na labarin shi ne cewa masu bincike ba zai iya koyi wani abu daga wadanda ba yiwuwa samfurori (ie, samfurori ba tare da m yiwuwa na tushen sharudda zabi mahalarta). Amma, kamar yadda zan nuna daga baya a wannan babi, shi ke ba quite dama. A maimakon haka, ina ganin akwai gaske biyu dabi'u ga wannan labarin. dabi'u da suke a matsayin gaskiya a yau suna nan yadda suke a 1936. Da farko, babban adadin haphazardly tattara bayanai za ta ba da tabbacin mai kyau kimanta. Na biyu, masu bincike bukatar zuwa asusu na yadda za su data aka tattara a lokacin da suke yin kimomi daga gare ta. A wasu kalmomin, saboda data tarin tsari a cikin Literary Digest zabe da aka tsare Ƙirgar zuwa wasu weights, masu bincike bukatar ka yi amfani da wani karin hadaddun hakkin tsari da nauyi wasu weights fiye da wasu. Daga baya a wannan sura, zan nuna muku wata irin weighting hanya-post-stratification-da za su iya taimaka maka ka yi m kimomi da wadanda ba yiwuwa samfurori.
