3.6.1 Amplified ngokubuza
Ukunxulumanisa uphando yakho imizila yedijithal kungafana njengokucela wonke imibuzo akho ngamaxesha onke.
Ukubuza efika ngokubanzi iindidi ezimbini eziphambili: uphando lwesampulu nezivumelano. nophando Sample, apho ukufikelela inani elincinane labantu, unokuba bhetyebhetye, ngexesha, kwaye ayafikeleleka. Nangona kunjalo, uphando isampulu, kuba zisekelwe kwisampulu, badla kulinganiselwe ukusonjululwa kwazo; kunye isampula yesaveyi, kunzima rhoqo ukwenza uqikelelo malunga kwimimandla wejografi okanye amaqela athile abakhoyo. Ukubalwa, kwelinye icala, azame udliwano bonke abemi. Baye isisombululo kakhulu, kodwa ke abiza imali eninzi ngokubanzi, kuxineneyo ingqwalasela (kuquka kuphela inani elincinane imibuzo), kwaye ngexesha elifanelekileyo (bona kwenzeka ucwangciso esisigxina, ezifana qho kwiminyaka 10) (Kish 1979) . Ngoku thelekela ukuba abaphandi bakwazi ukudibanisa iimpawu eziphambili kuphando lwesampulu nezivumelano; Ucinga ukuba abaphandi unokubuza zonke umbuzo ukuba wonke yonke imihla.
Ngokucacileyo, oku lamaxesha, Abazobi, usoloko-survey luhlobo njengephupha nenzululwazi yoluntu. Kodwa ke, kubonakala ukuba unokuqala olu beko ngokudibanisa imibuzo uphando evela inani elincinane labantu imizila digital kubantu abaninzi. Ndiya kubiza olu hlobo indibanisela ezikhala ngokubuza. Ukuba kwenziwe kakuhle, oko kusinceda libonelela ziqikelela ukuba bengingqi (imimandla ezincinane kuyo), hlalutye ngaphezulu (amaqela athile ngokwedemografi), kwaye ngexesha kakhulu.
Omnye umzekelo ndikubuza ziyanyuswa iphuma umsebenzi uYoshuwa Blumenstock, owayefuna ukuqokelela idata ezaziza kumnceda uphuhliso sikhokelo kumazwe ahluphekayo. Ngokungakumbi, Blumenstock wayefuna ukudala inkqubo ukulinganisa ubutyebi kunye nentlalo-ntle edibanisa ukuphelela ubalo kunye bhetyebhetye kunye nezihlandlo uphando (Blumenstock 2014; Blumenstock, Cadamuro, and On 2015) . Enyanisweni, ndiye esele ichaziwe umsebenzi Blumenstock ngayo ngokufutshane kwiSahluko 1.
Ukuqalisa, Blumenstock yabambisana nomboneleli inkulu ifowuni ephathwayo in Rwanda. Le nkampani Wazenzela iirekhodi transaction zingachazwa ukuba zezikabani ukusuka malunga 1.5 lezigidi yangaphezulu ukuziphatha-2005 nowama-2009 iinkuni ziqulathe ulwazi malunga ngalinye umyalezo umnxeba nokubhaliweyo ezifana ixesha lokuqala, ubude, kunye nendawo omalunga wejografi umfowuni nesamkeli. Ngaphambi kokuba uqalise ukuthetha malunga nemiba manani, kubalulekile wayebonisa ukuba eli nyathelo lokuqala kusenokuba yenye inzima. Njengoko kuchaziwe kwiSahluko 2, inkoliso yedijithali data wokulanda ayinakufikelelwa ukuba abaphandi. Kwaye, iinkampani ezininzi sizathu bayathandabuza ukwabelana data zabo kuba yinto yabucala; ukuba abathengi zabo mhlawumbi akazange silindele ukuba iirekhodi zabo ziya kwabelwana-kwenzaliso-nabaphandi. Kulo mzekelo, abaphandi wathabatha amanyathelo ngenyameko anonymize idata kunye nomsebenzi wabo lwalonganyelwe ngumntu wesithathu-iqela (oko kukuthi, IRB yabo). Kodwa ke, nangona ezi nzame, ezi data mhlawumbi nangoku achongeka kwaye kusenokwenzeka ziqulathe ulwazi ontununtunu (Mayer, Mutchler, and Mitchell 2016; Landau 2016) . Ndiza kubuyela ezi umbuzo yokuziphatha kwiSahluko 6.
Khumbula ukuba Blumenstock wayenomdla yokulinganisa ubutyebi kunye nentlalo-ntle. Kodwa ke, ezi mpawu a ngqo kwiirekhodi umnxeba. Ngamanye amazwi, ezi ngxelo ikhwelo aziphelelanga solu phando, bekuyinto eqhelekileyo imizila yedijithali ukuba saxoxwa kwiSahluko 2 Kodwa ke, kubonakala ngathi ukuba iirekhodi ubizo mhlawumbi benalo ulwazi malunga nobutyebi kunye nentlalo-ntle. Ngoko, enye indlela yokubuza umbuzo Blumenstock ingadodobaliseka: ngaba kunokwenzeka ukuqikelela ukuba umntu uya kusabela uphando olusekelwe data zabo wokulanda yedijithali? Ukuba kunjalo, ngoko ngokubuza abantu abambalwa sikwazi ukuthelekelela iimpendulo wonke umntu.
Ukuhlola oku empirically, abancedisi Blumenstock nophando ukusuka Kigali Institute of Science and Technology wabiza isampula malunga namawaka abathengi ifowuni ephathwayo. Abaphandi bachaza iinjongo zeprojekthi ukuba abathathi-nxaxheba, wacela imvume yabo ukunxulumanisa iimpendulo zesaveyi iirekhodi umnxeba, waza wabuza kubo uthotho lwemibuzo ukulinganisa ubutyebi babo kunye nentlalo-ntle, njengoko enjalo 'Ngaba unayo kunomathotholo? "kwaye" Ingaba unepropati ibhayisekile? "(jonga uMfanekiso 3.11 kuba uluhlu buso). Bonke abathathi-nxaxheba kuphando zaye babuyekezwe ngemali.
Okulandelayo, Blumenstock wasebenzisa inkqubo inyathelo-ezimbini ezixhaphakileyo kwinzululwazi data: zobunjineli osemqoka ilandelwe yokufunda supervised. Okokuqala, xa inyathelo osemqoka zobunjineli, kuba wonk 'ubani ukuba udliwano, Blumenstock aguqulwa iirekhodi umnxeba ibe iseti iimpawu malunga umntu ngamnye; izazinzulu data ukuze kubiza ezi mpawu "mpawu" kunye ntle ukuzibiza "variables." Ngokomzekelo, umntu ngamnye, Blumenstock ibalwe Inani lilonke leentsuku umsebenzi, inani labantu acacileyo umntu iye yahlangana, isixa imali echithwa airtime, njalo-njalo. Okubalulekileyo, osemqoka zobunjineli okulungileyo ufuna ulwazi yophando. Umzekelo, ukuba kubalulekile ukwahlula phakathi iminxeba yasekhaya neyamazwe ngamazwe (unokulindela thina bantu kubiza ngamazwe ukuba izityebi), oku kufuneka kwenziwe kwi-inyathelo osemqoka zobunjineli. Umphandi ngokuqonda encinane eRwanda ukuze kuquka olu phawu, kwandule ukusebenza oqikelela umfuziselo luyakuchaneka.
Okulandelayo, kwi inyathelo yokufunda kweliso, Blumenstock wakha statistical model ukuqikelela impendulo survey kumntu ngamnye ngokusekelwe iimpawu zabo. Kule meko, Blumenstock wasebenzisa eguqulweyo yobuchule kunye esibayeni-10 cross-yosebenzo, kodwa akaba babesebenzisa iintlobo ezinye iindlela manani okanye umatshini yokufunda.
Ngoko njani ukusebenza kakuhle njani? Ngaba Blumenstock akwazi ukuqikelela iimpendulo kulibhala imibuzo efana "Ngaba nonomathotholo?" Kwaye "Ingaba unepropati ibhayisekile?" Esebenzisa iimpawu zivela iirekhodi umnxeba? Yohlobo lwe. Ukuchaneka neengqikelelo baba eliphezulu kwezinye izinto (Figure 3.11). Kodwa ke, kubalulekile kusoloko ukuthelekisa indlela uqikelelo entsonkothileyo ngokuchasene indlela elula. Kule meko, kukho indlela elula kukuba ukuqikelela ukuba wonke umntu uya kukunika impendulo yeyona ixhaphakileyo. Umzekelo, 97,3% ingxelo bendihlala unomathotholo ukwenzela ukuba Blumenstock elavakaliswa ukuba wonke umntu uza ingxelo bendihlala unomathotholo ange waba elichanileyo 97,3%, nto leyo efana ngokumangalisayo nokusebenza kwenkqubo yakhe olunzima kakhulu (97.6% ukuchaneka). Ngamanye amazwi, yonke idatha ezintle kunye imizekelo kwandise ukuchaneka yingqikelelo ukusuka 97,3% ukuya 97,6%. Noko ke, ezinye imibuzo efana nale: "Ingaba unepropati ibhayisekile?", Neengqikelelo kuphucuke ukusuka 54,4% ukuya 67,6%. Okuninzi ngokubanzi, Umfanekiso 3.12 ibonisa ukuba ezinye izinto Blumenstock akazange ukuphucula kakhulu ngaphaya ukwenza ingqikelelo elula olusisiseko, kodwa ukuba ezinye iimpawu kwabakho ukuphucuka.
Umfanekiso 3.11: ukuchana oluluqikelelo zemodeli statistical abaqeqeshiweyo kunye neerekhodi umnxeba. Iziphumo kuLuhlu 2 Blumenstock (2014) .
Umfanekiso 3.12: Uthelekiso ukuchaneka axelwe zemodeli statistical abaqeqeshiweyo kunye neerekhodi umnxeba wokuxela elula lokwandlalela. Manqaku jittered kancinane ukuphepha uphumelo; khangela Uluhlu 2 Blumenstock (2014) amaxabiso ngqo.
Kweli nqanaba nokuba ucinga ukuba ezi ziphumo kancinane iyadanisa, kodwa ngonyaka nje omnye emva koko, Blumenstock kunye ababini nabalingane-Gabriel Cadamuro noRobert On-lapapasha iphepha in Science neziphumo kakhulu engcono (Blumenstock, Cadamuro, and On 2015) . Kwakukho izizathu ezibini eziphambili bobugcisa ukuze ukuphuculwa: 1) basebenzisa iindlela nobunzima kakhulu (oko kukuthi, indlela entsha zibekhona ubunjineli kunye nemodeli yokufunda umatshini nobunzima kakhulu) kunye 2) kunokuba uzama ukuqonda iimpendulo kwimibuzo yesaveyi ngamnye (umzekelo, "Ngaba nonomathotholo?"), bazama ukuba Ukuthelekelela isalathiso ubutyebi libumbene.
Blumenstock noogxa wabonisa intsebenzo indlela yabo ngeendlela ezimbini. Okokuqala, bafumana ukuba abantu kwisampulu yabo, wayenokwenza umsebenzi omhle ukuxela ubutyebi babo iirekhodi umnxeba (Figure 3.14). Okwesibini, kanaphakade okona kubaluleke kakhulu, Blumenstock noogxa wabonisa ukuba inkqubo yabo onokubangela uqikelelo ophezulu-umgangatho emhlabeni ubutyebi eRwanda. Ngokungakumbi, basebenzisa sokufunda kumatshini wazo, baqeqeshwa kwi sampulu yabo labantu abamalunga 1000, ukuqikelela ubutyebi bonke 1.5 sezigidi zabantu kwiirekhodi umnxeba. Ngaphezu koko, idatha geospatial okulungiswe ifowuni yedata (khumbula ukuba ifowuni yedata iquka indawo kweseli ende ekufutshane ukuze umnxeba nganye), abaphandi baye bakwazi ukuqikelela kwindawo esikuyo yokuhlala lomntu ngamnye. Ukubeka ezi uqikelelo ezimbini, uphando yavelisa Luqikelelo emhlabeni ubutyebi umrhumo xa granularity entle kakhulu lwemihlaba. Umzekelo, bakwazi ukuqikelela ubutyebi avareji ngamnye iiseli Rwanda ka-2148 (iyunithi yolawulo lincinane elizweni). Ezi nqobo ubutyebi kwangaphambili baba hlalutye ngoko kunzima ukukhangela. Ngoko ke, abaphandi oluqinile iziphumo zabo ukuvelisa uqikelelo ubutyebi avareji kwizithili 30 Rwanda ngayo. Le iingqikelelo-kwinqanaba lesithili babezalana ngamandla uqikelelo ukusuka igolide uphando yemveli umgangatho, iRwandan Demographic and Health Survey (Figure 3.14). Nangona uqikelelo evela kwimithombo ababini efanayo, uqikelelo ukusuka Blumenstock kunye nabalingane 50 amaxesha Akubizi kakhulu kwaye amaxesha 10 ngokukhawuleza (xa iindleko phantsi kulinganiswa ngokwemigaqo iindleko eziguquguqukayo). Le kukuhla kakhulu iindleko kuthetha ukuba kunokuba kuqhuba yonke iminyaka embalwa na umgangatho Demographic and Health Surveys-eziingxilimbela zesaveyi amancinci kudityaniswe data ezinkulu wokulanda yedijithali ungagiliswa nyanga nganye.
Umfanekiso 3.13: Umzobo Blumenstock, Cadamuro, and On (2015) . idata esuka kwinkampani yefowuni ke atshintshelwe matrix kunye kumqolo omnye umntu ngamnye lomhlathi enye kuphawu ngamnye (oko kukuthi, umahluko). Okulandelayo, abaphandi wakha sokufunda supervised ukuqikelela iimpendulo kuphando ukusuka umntu osemqoka matrix. Emva koko, le sokufunda supervised yasetyenziselwa ukuba sibone iimpendulo yesaveyi wonke. Enyanisweni, abaphengululi basebenzisa iimpendulo abantu malunga amawaka ukuba sibone ubutyebi abantu abamalunga nesigidi esinye. Kwakhona, abaphandi kuqikelelwa kwindawo esikuyo yokuhlala bonke abantu 1.5 yezigidi ngokusekelwe iindawo iminxeba zabo. Xa ezi Uqikelelo zimbini zidibene-ubutyebi kuqikelelwa kunye nendawo eliqikelelwayo yokuhlala-iziphumo zazifana uqikelelo ukusuka Demographic and Health Survey, uhlolisiso yemveli yegolide-umgangatho (Figure 3.14).
Umfanekiso 3.14: Iziphumo ukusuka Blumenstock, Cadamuro, and On (2015) . Ngelo umntu-inqanaba, abaphandi baye bakwazi ukwenza umsebenzi eyamkelekileyo yokuqikelela ubutyebi umntu ukusuka iirekhodi zabo umnxeba. Uqikelelo ubutyebi-nto-kwinqanaba lesithili zisekelwe kuqikelelo ngamnye-inqanaba ubutyebi kunye nendawo yokuhlala-iziphumo zazifana iziphumo ukusuka Demographic and Health Survey, uhlolisiso yemveli igolide-umgangatho.
Elokuphetha, Blumenstock ngayo Amplified ngokubuza indlela odityanisiweyo zophando yedatha ngedatha wokulanda yedijithali ukuvelisa uqikelelo ezinokuthelekiswa kunye noqikelelo uphando ngegolide-umgangatho. Lo mzekelo ethile kwakhona licacisa ezinye urhwebo-offs phakathi kubuza ziyanyuswa neendlela uphando zemveli. Okokuqala, uqikelelo ecela mkhulu ngakumbi ngexesha, kakhulu ayibizi, kunye nangokwamahlalutye ngaphezulu. Kodwa ke, kwelinye icala, ngeli xesha, akukho isiseko theoretical lo hlobo bawucele ziyanyuswa olomeleleyo. Oko, le umzekelo omnye alubonisi xa uya kusebenza naxa ayiyi. Ngaphezu koko, indlela ayizange ziyanyuswa akanazo iindlela ezilungileyo ubungakanani ukungaqiniseki malunga iingqikelelo zayo. Nangona kunjalo, umbuzo ziyanyuswa linonxibelelwano enzulu iinkalo ezintathu ezinkulu zamanani-imodeli-based post-abahlulwe (Little 1993) , imputation (Rubin 2004) , kunye ukulinganisela encinane-ndawo (Rao and Molina 2015) yaye ngoko ke silindele ukuba inkqubela iya kuba ngokukhawuleza.
ndikubuza kubalilapo lulandelayo iresiphi ezisisiseko na kanye ukuba imeko yakho ethile. Kukho izinto ezimbini kunye namanyathelo ezimbini. Izithako ezimbini 1) dataset wokulanda yedijithali ukuba iphangalele kodwa ibhityile (oko kukuthi, ibe abantu abaninzi, kodwa hayi inkcazelo oyifunayo malunga nabantu ngalinye) kunye 2) ngohlolisiso imxinwa kodwa ubukhulu (oko kukuthi, umntu abantu abambalwa kuphela, kodwa inkcazelo oyifunayo abo bantu). Ngoko ke, kukho amanyathelo ezimbini. Okokuqala, kuba abantu zombini imvelaphi data, ukwakha umfuziselo yokufunda kumatshini eyisebenzisa data wokulanda digital ukuqikelela uphando iimpendulo. Okulandelayo, sebenzisa ukuba imodeli umatshini ukufunda ukuba sibone iimpendulo uphando wonke kwi data umkhondo yesuntswana. Ngoko ke, ukuba kukho umbuzo ufuna ukucela ukuba bazizibonelo, kwaye abantu, jonga data wokulanda digital kwabo bantu anokusetyenzisWa ukuqikelela impendulo yabo.
Ukuthelekisa Blumenstock ezame yokuqala neyesibini le ngxaki nazo zikhanyisa isifundo esibalulekileyo malunga utshintsho ukusuka kwixesha yesibini iindlela era wesithathu kulibhala uphando: kwasekuqaleni hayi ukuphela. Oko kukuthi, amaxesha amaninzi, indlela yokuqala iya kuba eyona, kodwa ukuba abaphandi iyaqhubeka ukusebenza, izinto ziba ngcono. Okuninzi ngokubanzi, xa kuhlolwa iindlela ezintsha zophando zentlalo kweminyaka yesuntswana, kubalulekile ukuba wenze uvavanyo ezibini ezicacileyo: 1) ingaba kakuhle lo msebenzi ngoku 2) ingaba kakuhle Ucinga ukuba oku kusebenza kwixesha elizayo njengoko mhlaba data utshintsho yaye abaphandi ukunikela ingqalelo engakumbi kule ngxaki. Nangona, abaphandi bayaqeqeshwa ukwenza uhlobo yokuqala yovavanyo (indlela elungileyo le othile kophando), owesibini kukholisa ukubaluleka ngakumbi.
