{"id":1897,"date":"2020-11-15T23:23:43","date_gmt":"2020-11-15T20:23:43","guid":{"rendered":"http:\/\/www.gokberkcan.com\/?p=1897"},"modified":"2020-11-18T20:45:40","modified_gmt":"2020-11-18T17:45:40","slug":"rda-temel-veri-analizi","status":"publish","type":"post","link":"https:\/\/www.gokberkcan.com\/?p=1897","title":{"rendered":"R’da Temel Veri Analizi"},"content":{"rendered":"\n

R ve Analiz<\/h1>\n\n\n\n

Bu yaz\u0131y\u0131 yazmaktaki amac\u0131m R kullanarak bilimsel \u00e7al\u0131\u015fma yapmak isteyen fakat bir sebepten ba\u015flayamayanlara asl\u0131nda s\u00fcrecin ne kadar kolay oldu\u011funu g\u00f6stermek. Metin i\u00e7inde kullan\u0131lan bilgileri farkl\u0131 kaynaklardan elde ettim:<\/p>\n\n\n\n

  1. Paketlerin kullan\u0131m\u0131n\u0131, fonksiyonlar\u0131n i\u00e7eri\u011fini paketlerde yer alan vinyetlerden okuyarak \u00f6\u011frendim. <\/li>
  2. Buradaki bilgilerin baz\u0131lar\u0131n\u0131 Datacamp’de ald\u0131\u011f\u0131m R kurslardan edindim. Bug\u00fcne kadar tamamlad\u0131\u011f\u0131m kurslar\u0131n (career track) isimlerini en sonda yazd\u0131m. Derslerin detaylar\u0131na ba\u011flant\u0131lardan ula\u015fabilirsiniz.<\/li>
  3. Burada bahsi ge\u00e7en konular\u0131n \u015fahs\u0131m ad\u0131na \u00e7ok zorla\u015ft\u0131\u011f\u0131 noktalarda Stack Exchange, Stack Overflow ve \u00fccretsiz R sitelerinde yaz\u0131lanlar\u0131 takip ederek sonuca ula\u015ft\u0131m. Hangi sayfa benim i\u00e7in en kolay anlat\u0131m\u0131 yapt\u0131ysa (ben nas\u0131l \u00f6\u011frendiysem), ona at\u0131f yapt\u0131m. Kulland\u0131\u011f\u0131m b\u00fct\u00fcn kaynaklar\u0131 kodlardan \u00f6nce payla\u015ft\u0131m. <\/li><\/ol>\n\n\n\n

    \u00d6nce hangi kodun ne i\u015fe yarad\u0131\u011f\u0131n\u0131 yorumlarla yan\u0131na yazd\u0131m, ancak kopyala-yap\u0131\u015ft\u0131r ile konsola att\u0131\u011f\u0131mda \u00e7ok u\u011fra\u015ft\u0131rd\u0131. O y\u00fczden daha anla\u015f\u0131l\u0131r olmas\u0131 amac\u0131yla kodlar\u0131n alt\u0131na ekledim.<\/p>\n\n\n\n

    Buradaki temel anlat\u0131m\u0131ndan elde edilecek fayday\u0131 b\u00fcy\u00fctmek, farkl\u0131 boyutlara getirmek tamamiyle sizin \u00e7aban\u0131zla ilgili. R kullanarak istatistiki analize giri\u015f yapmak isteyenler i\u00e7in faydal\u0131 olmas\u0131 dile\u011fiyle.<\/p>\n\n\n\n

    Haydi ba\u015flayal\u0131m!<\/p>\n\n\n\n\n\n\n\n

    R Paketleri<\/h2>\n\n\n\n

    Burada yapaca\u011f\u0131m\u0131z i\u015flemlere ba\u015flamadan \u00f6nce a\u015fa\u011f\u0131daki R paketlerinin kurulmas\u0131 ve \u00e7al\u0131\u015ft\u0131r\u0131lmas\u0131 gerekli. RStudio kullanaca\u011f\u0131m\u0131z i\u00e7in readxl paketi zaten veri aktar\u0131m\u0131 yap\u0131l\u0131rken etkinle\u015fece\u011fi i\u00e7in a\u015fa\u011f\u0131daki tabloya eklemedim. Her bir paket i\u00e7in kurulum ve etkinle\u015ftirme komutlar\u0131 a\u015fa\u011f\u0131daki tabloda. Kopyala yap\u0131\u015ft\u0131r yap\u0131p konsola aktarabilece\u011finiz gibi R al\u0131\u015fkanl\u0131\u011f\u0131n\u0131z\u0131n artmas\u0131 i\u00e7in tek tek yazman\u0131z\u0131 tavsiye ederim.<\/p>\n\n\n\n

    \n\n\n\n\n\n\n\n\n\n
    Paket Ad\u0131<\/th>\nNeden?<\/th>\nKurulum<\/th>\nEtkinle\u015ftirme<\/th>\n<\/tr>\n<\/thead>\n
    ggplot2<\/td>\nG\u00fczel grafikler<\/td>\ninstall.packages("ggplot2")<\/td>\nlibrary(ggplot2)<\/td>\n<\/tr>\n
    dplyr<\/td>\nVeriseti manip\u00fclasyonu<\/td>\ninstall.packages("dplyr")<\/td>\nlibrary(dplyr)<\/td>\n<\/tr>\n
    lmtest<\/td>\nD\u00fczeltilmi\u015f standartlar hatalar<\/td>\ninstall.packages("lmtest")<\/td>\nlibrary(lmtest)<\/td>\n<\/tr>\n
    sandwich<\/td>\nD\u00fczeltilmi\u015f standartlar hatalar<\/td>\ninstall.packages("sandwich")<\/td>\nlibrary(sandwich)<\/td>\n<\/tr>\n
    Hmisc<\/td>\nKorelasyon Matrisi<\/td>\ninstall.packages("Hmisc")<\/td>\nlibrary(Hmisc)<\/td>\n<\/tr>\n
    quantreg<\/td>\nKantil Regresyon<\/td>\ninstall.packages("quantreg")<\/td>\nlibrary(quantreg)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n\n\n\n

    Veri seti<\/h2>\n\n\n\n

    MODAVICA 2020’de sundu\u011fum “Effect of Life Cycles on Capital Expenditures: Evidence from Borsa Istanbul<\/em>” bildirisinde kulland\u0131\u011f\u0131m veri setinin buradaki anlat\u0131ma g\u00f6re d\u00fczenlenmi\u015f hali. Setin yap\u0131s\u0131 a\u015fa\u011f\u0131da sunulmu\u015ftur. Veri setimizin ilk \u00fc\u00e7 s\u00fctunu tan\u0131msal de\u011ferler oldu\u011fu i\u00e7in baz\u0131 fonksiyonlarda aral\u0131klar\u0131 kullanaca\u011f\u0131z.<\/p>\n\n\n\n

    \n\n\n\n\n\n\n\n\n\n\n\n\n
    De\u011fi\u015fken Kodu<\/th>\nDe\u011fi\u015fken Ad\u0131<\/th>\nDe\u011fi\u015fken Yap\u0131s\u0131<\/th>\n<\/tr>\n<\/thead>\n
    ID<\/td>\nBirim<\/td>\nTan\u0131msal<\/td>\n<\/tr>\n
    FY<\/td>\nY\u0131l<\/td>\nTan\u0131msal<\/td>\n<\/tr>\n
    SC<\/td>\nSekt\u00f6r<\/td>\nTan\u0131msal<\/td>\n<\/tr>\n
    BD<\/td>\nBa\u011f\u0131ml\u0131 De\u011fi\u015fken<\/td>\nS\u00fcrekli<\/td>\n<\/tr>\n
    X1<\/td>\nBa\u011f\u0131ms\u0131z<\/td>\nS\u00fcrekli<\/td>\n<\/tr>\n
    X2<\/td>\nKontrol<\/td>\nS\u00fcrekli<\/td>\n<\/tr>\n
    X3<\/td>\nKontrol<\/td>\nS\u00fcrekli<\/td>\n<\/tr>\n
    X4<\/td>\nKontrol<\/td>\nKukla<\/td>\n<\/tr>\n
    X5<\/td>\nKontrol<\/td>\nKukla<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n\n\n\n

    Veri setinin aktar\u0131m\u0131<\/h3>\n\n\n\n

    \u00d6ncelikle a\u015fa\u011f\u0131daki veri setini bilgisayar\u0131n\u0131za indirin. <\/p>\n\n\n\n

    (yeni sekmede a\u00e7\u0131lacak)<\/p>\n\n\n\n

    \u015eahsi tercihim RStudio’da bir proje olu\u015fturman\u0131z. Bunu iki sebeple s\u00f6yl\u00fcyorum:<\/p>\n\n\n\n

    1. .RData dosyas\u0131n\u0131z sadece burada yapt\u0131klar\u0131n\u0131zla ilgili olacakt\u0131r. Veri setini bozsan\u0131z bile ba\u015fka \u00e7al\u0131\u015fmalara ait dosyalar\u0131n\u0131z yerli yerinde duracakt\u0131r (deneyim diyelim).<\/li>
    2. En son a\u015famada kulland\u0131\u011f\u0131m\u0131z not defteri de a\u00e7t\u0131\u011f\u0131n\u0131z projenin bir par\u00e7as\u0131 olaca\u011f\u0131 i\u00e7in komutlar\u0131 hat\u0131rlamak daha kolay olacak (bu sitenin hi\u00e7bir kazan\u00e7 amac\u0131 yoktur). <\/li><\/ol>\n\n\n\n

      RStudio’daki “Import Dataset<\/em>” butonu ile dosyay\u0131 indirdi\u011finiz konumdan i\u00e7eri vseti<\/em> ismiyle aktar\u0131n.<\/p>\n\n\n\n

      \u00c7al\u0131\u015fma Plan\u0131m\u0131z<\/h2>\n\n\n\n
      1. Deskriptif istatistik hesaplayaca\u011f\u0131z
        1. \u00d6zet istatistikkleri raporlayaca\u011f\u0131z.<\/li>
        2. BD, X1<\/sub>, KD1<\/sub>,KD2<\/sub>,KD3<\/sub>,KD4<\/sub> ve KD5<\/sub> i\u00e7in korelasyon matrisi haz\u0131rlayaca\u011f\u0131z.<\/li>
        3. X1<\/sub> de\u011fi\u015fkenini baz al\u0131p gruplar i\u00e7in farklar testi yapaca\u011f\u0131z<\/li>
        4. Gruplar\u0131 kullanarak toplam ve y\u0131llara g\u00f6re grafikler haz\u0131rlayaca\u011f\u0131z.<\/li><\/ol><\/li>
        5. BD = X1<\/sub>+KD1<\/sub>+KD2<\/sub>+KD3<\/sub>+KD4<\/sub>+KD5<\/sub> modelini y\u0131l ve birim i\u00e7in kukla de\u011fi\u015fkenler kullanarak sabit etkiler ile test edece\u011fiz.
          1. Modeli standart hatalar\u0131n d\u00fczelterek raporlayaca\u011f\u0131z.<\/li><\/ol><\/li>
          2. Ayn\u0131 modeli kantil regresyon ile test edece\u011fiz.
            1. Kantil regresyon modeline bootstrapping yapaca\u011f\u0131z.<\/li>
            2. Kantil regresyon modelinde R2<\/sup> hesaplayaca\u011f\u0131z.<\/li>
            3. Kantil Regresyon modelini toplam olarak raporlayaca\u011f\u0131z.<\/li><\/ol><\/li>
            4. B\u00fct\u00fcn elde etti\u011fimiz sonu\u00e7lar\u0131 payla\u015f\u0131labilir hale getirece\u011fiz.<\/li><\/ol>\n\n\n\n

              Deskriptif \u0130statistikler<\/h2>\n\n\n\n

              \u00d6zet \u0130statistiki<\/h3>\n\n\n\n

              \u00d6ncelikle veri setimizde ne neymi\u015f bir bakal\u0131m. G\u00f6zlemlerimin tan\u0131mlar\u0131n\u0131n (ilk \u00fc\u00e7 s\u00fctun) herhangi bir istatistiki kar\u015f\u0131l\u0131\u011f\u0131 olmad\u0131\u011f\u0131 i\u00e7in istastiki \u00f6zetimize eklemiyoruz. summary<\/em> fonksiyonu i\u00e7indeki vseti[,4:10]<\/em> tan\u0131m\u0131n\u0131n da amac\u0131 ilk \u00fc\u00e7\u00fc istemedi\u011fimizi belirtmek. Daha kapsaml\u0131 \u00f6zet istatistikler i\u00e7in bu makaleyi<\/a> kullanabilirsiniz. Bana bu kadar s\u0131f\u0131r yetmez daha fazla isterim diyorsan\u0131z da \u00fc\u00e7\u00fcnc\u00fc komutu \u00e7al\u0131\u015ft\u0131r\u0131n.<\/p>\n\n\n\n

              desk<-summary(vseti[,4:10])\ndesk\noptions(digits = 7)<\/code><\/pre>\n\n\n\n
              Korelasyon Matrisi<\/h3>\n\n\n\n
              Korelasyon matrisini iki \u015fekilde elde edebiliriz R’da. Birinci y\u00f6ntem R’\u0131n kendi korelasyon matris fonksiyonu olan “cor” ile yap\u0131labilir ancak bu fonksiyona dair esas sorun ili\u015fkilerin istatistiki anlaml\u0131l\u0131\u011f\u0131 raporlanmaz.<\/p>\n\n\n\n
              kormat<-cor(vseti[,4:10])\nkormat<\/code><\/pre>\n\n\n\n
              Peki nedir bu sorunun \u00e7\u00f6z\u00fcm\u00fc? Hmisc paketi. A\u015fa\u011f\u0131daki komutlar\u0131 \u00e7al\u0131\u015ft\u0131rd\u0131\u011f\u0131n\u0131zda kar\u015f\u0131n\u0131za gelen korelasyon matrisinde iki k\u0131s\u0131m olacak. Birincisi de\u011fi\u015fkenler aras\u0131 korelasyon, ikincisi korelasyonlar\u0131n istatistiki anlaml\u0131l\u0131\u011f\u0131.<\/p>\n\n\n\n
              kormat_pv<-rcorr(as.matrix((vseti[,4:10])))\nkormat_pv<\/code><\/pre>\n\n\n\n
              Farklar\u0131n Anlaml\u0131\u011f\u0131 Testi<\/h3>\n\n\n\n
              Elimizdeki veri setindeki X1 de\u011fi\u015fkeninde 0.00, 0.25, 0.50, 0.75 ve 1.00 de\u011ferlerinden 39’ar adet olmak \u00fczere 195 g\u00f6zlem var. Hi\u00e7bir birim i\u00e7in eksik g\u00f6zlemimiz yok. X1 de\u011fi\u015fkeni i\u00e7indeki gruplar\u0131n Y ortalamalar\u0131 aras\u0131nda bir fark olup olmad\u0131\u011f\u0131n\u0131 inceleyelim \u015fimdi. De\u011ferleri nas\u0131l gruplayaca\u011f\u0131z. <\/p>\n\n\n\n
              \n\n\n\n\n\n\n\n\n
              X1 De\u011feri<\/th>\nGrup<\/th>\n<\/tr>\n<\/thead>\n
              0.00<\/td>\n1<\/td>\n<\/tr>\n
              0.25<\/td>\n2<\/td>\n<\/tr>\n
              0.50<\/td>\n3<\/td>\n<\/tr>\n
              0.75<\/td>\n4<\/td>\n<\/tr>\n
              1.00<\/td>\n5<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n\n\n\n
              Grup12 <- vseti %>%\nfilter(X1 == 0 | X1 == 0.25) %>%\nselect(X1, BD)\n\nGrup13 <- vseti %>%\nfilter(X1 == 0 | X1 == 0.5) %>%\nselect(X1, BD)\n\nGrup14 <- vseti %>%\nfilter(X1 == 0 | X1 == 0.75) %>%\nselect(X1, BD)\n\nGrup15 <- vseti %>%\nfilter(X1 == 0 | X1 == 1) %>%\nselect(X1, BD)\n\nGrup23 <- vseti %>%\nfilter(X1 == 0.25 | X1 == 0.5) %>%\nselect(X1, BD)\n\nGrup24 <- vseti %>%\nfilter(X1 == 0.25 | X1 == 0.75) %>% \nselect(X1, BD)\n\nGrup25 <- vseti %>%\nfilter(X1 == 0.25 | X1 == 1) %>%\nselect(X1, BD)\n\nGrup34 <- vseti %>%\nfilter(X1 == 0.5 | X1 == 0.75) %>%\nselect(X1, BD)\n\nGrup35 <- vseti %>% \nfilter(X1 == 0.5 | X1 == 1) %>%\nselect(X1, BD)\n\nGrup45 <- vseti %>%\nfilter(X1 == 0.75 | X1 == 1) %>%\nselect(X1, BD)<\/code><\/pre>\n\n\n\n
              \u00dcstteki komutlar\u0131 k\u0131saca izah etmek isterim. dplyr paketini etkinle\u015ftirdi\u011fimiz i\u00e7in %>% operat\u00f6r\u00fc ile b\u00fct\u00fcn komutu ile tek bir sat\u0131rda yazmak yerine par\u00e7alara b\u00f6lm\u00fc\u015f oluyoruz.<\/p>\n\n\n\n
              Grup12 <- vseti %>%\nfilter(X1 == 0 | X1 == 0.25) %>%\nselect(X1, BD)\n\n# Grup12 tablosunu olu\u015ftur ve vseti isimli veri setini kullan\n# vseti i\u00e7erisindeki X1 de\u011fi\u015fkeni i\u00e7in 0 ve 0.25 olan g\u00f6zlemleri filtrele\n# vseti i\u00e7indeki 0 ve 0.25 olan X1 g\u00f6zlemlerini i\u00e7in X1 ve BDS de\u011fi\u015fkenlerini Grup12 tablosuna yerle\u015ftir.<\/code><\/pre>\n\n\n\n
              Gruplar\u0131 olu\u015fturdu\u011fumuza g\u00f6re anlaml\u0131l\u0131k testlerimiz yapal\u0131m her bir grup i\u00e7in (UC Business Analytics R Programming Guide<\/a>).<\/p>\n\n\n\n
              t.test(BD ~ X1, data = Grup12)\nt.test(BD ~ X1, data = Grup13)\nt.test(BD ~ X1, data = Grup14)\nt.test(BD ~ X1, data = Grup15)\nt.test(BD ~ X1, data = Grup23)\nt.test(BD ~ X1, data = Grup24)\nt.test(BD ~ X1, data = Grup25)\nt.test(BD ~ X1, data = Grup34)\nt.test(BD ~ X1, data = Grup35)\nt.test(BD ~ X1, data = Grup45)<\/code><\/pre>\n\n\n\n
              Grafiklerin Olu\u015fturulmas\u0131<\/h3>\n\n\n\n
              Grafikleri olu\u015fturmak i\u00e7in \u00f6nce her bir grubun ve gruplar\u0131n y\u0131llara g\u00f6re ortalamas\u0131n\u0131 veri setimize ekleyelim. <\/p>\n\n\n\n
              vseti<- vseti %>% \ngroup_by(X1) %>%\nmutate(BD_X1_ort=mean(BD))\n \n\n# vseti kullanarak vseti'ne i\u015flemleri yap\n# vseti'ni X1 kullanarak grupla \n# vseti'nin sonuna BD'nin X1 gruplar\u0131na g\u00f6re ortalamas\u0131n\u0131 ekle\n\nvseti <- vseti %>%\ngroup_by(X1, FY) %>% \nmutate(BD_X1_FY_ort=mean(BD)) \n\n# vseti kullanarak vseti'ne i\u015flemleri yap \n# vseti'ni X1 ve FY kullanarak grupla \n# vseti'nin sonuna BD'nin X1-FY gruplar\u0131na g\u00f6re ortalamas\u0131n\u0131 ekle<\/code><\/pre>\n\n\n\n
              \u015eimdi grafikleri \u00e7\u0131kartabiliriz. \u00d6nce toplamdan gidelim (ggplot2<\/a> ve ggplot2 cheat sheet<\/a>)<\/p>\n\n\n\n
              ggplot(vseti, aes(x=X1, y=BD_X1_ort, fill=X1))+\ngeom_point(size=3,aes(colour=X1))+\nlabs(title=\"Grafigin Ana Basligi\", x =\"X1 Gruplari\", y = \"Ortalama Bagimli Degisken\")+\ntheme(plot.title = element_text(hjust = 0.5))+\nscale_y_continuous(breaks = seq(0,0.1,0.01))\n\n#vseti'ni kullanarak BD~X1 grafi\u011fini olu\u015ftur.\n#noktalar\u0131 g\u00f6ster, nokta b\u00fcy\u00fckl\u00fc\u011f\u00fc 3 olsun, X1'e g\u00f6re renklendir\n#eksenleri ve grafi\u011fin ad\u0131n\u0131 yaz\n#grafi\u011fin ad\u0131n\u0131 ortala\n#y eksenindeki ara de\u011ferleri ayarla\n<\/code><\/pre>\n\n\n\n
              \u015eimdi y\u0131llara g\u00f6re grafikleri \u00e7\u0131kartal\u0131m<\/p>\n\n\n\n
              ggplot(vseti, aes(x=FY, y=BD_X1_FY_ort, fill=X1))+\ngeom_line(aes(colour=X1, group=X1))+\ngeom_point(size=1,aes(colour=X1))+\nlabs(title=\"Grafigin Ana Basligi\", x =\"X1 Gruplari\", y = \"Yillara Gore Ortalama Bagimli Degisken\")+\ntheme(plot.title = element_text(hjust = 0.5))+\nscale_y_continuous(breaks = seq(0,0.1,0.01))+\nscale_x_continuous(breaks = seq(2015,2017,1))\n\n#vseti'ni kullanarak BD~FY grafi\u011fini olu\u015ftur.\n#\u00e7izgi grafik olu\u015ftur\n#noktalar\u0131 g\u00f6ster, nokta b\u00fcy\u00fckl\u00fc\u011f\u00fc 3 olsun, X1'e g\u00f6re renklendir\n#eksenleri ve grafi\u011fin ad\u0131n\u0131 yaz\n#grafi\u011fin ad\u0131n\u0131 ortala\n#y eksenindeki ara de\u011ferleri ayarla\n<\/code><\/pre>\n\n\n\n
              Regresyon Modeli <\/h2>\n\n\n\n
              \u00c7al\u0131\u015fmam\u0131zdaki temel varsay\u0131m\u0131m\u0131z (hipotez) a\u015fa\u011f\u0131daki gibidir:<\/p>\n\n\n\n
              H0: X1 BD'yi etkilememektedir\nH1: X1 BD'yi etkilemektedir (Beklenen i\u015faret: )<\/pre>\n\n\n\n
              Bu varsay\u0131m\u0131 test etmek modelimizi en k\u00fc\u00e7\u00fck kareler regresyonu olarak kaydedelim. Modele birim ve y\u0131l sayesinde sabit etkileri de test etmi\u015f oluyoruz. A\u00e7\u0131k\u00e7as\u0131 birim ve y\u0131l olmadan daha g\u00fczel sonu\u00e7lar elde edebilirsiniz ancak iyi bir dergide do\u011frudan kabul g\u00f6rmesi m\u00fcmk\u00fcn olmayacak. Modelin sonu\u00e7lar\u0131n\u0131 g\u00f6rmek i\u00e7in ikinci komutu \u00e7al\u0131\u015ft\u0131r\u0131n. <\/p>\n\n\n\n
              model<-lm(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti)\nsummary(model)\n\n#factor() fonksiyonunun amac\u0131 her bir birim koduna ve y\u0131la kukla de\u011fi\u015fken vermek yerine hepsini \u00e7al\u0131\u015ft\u0131rmak. R her bir gruptan bir tanesini analize dahil etmeyecek<\/code><\/pre>\n\n\n\n
              Standart Hatalar D\u00fczeltilmi\u015f Model<\/h3>\n\n\n\n
              Do\u011frusal ile test etti\u011fimiz modelde bir sorun var. Modelin standart hatalar\u0131n\u0131 birim ve y\u0131llara g\u00f6re k\u00fcmeleyerek sonu\u00e7lar\u0131n istatistiki olarak sa\u011flaml\u0131\u011f\u0131n\u0131 yapaca\u011f\u0131z.<\/p>\n\n\n\n
              model_SE<-coeftest(model,vcov=vcovHC(model,type=\"HC1\",cluster=c(PDS$ID,PDS$FY)))<\/code><\/pre>\n\n\n\n
              Modelin kantil regresyon ile test edilmesi<\/h2>\n\n\n\n
              Varsayal\u0131m ki do\u011frusal regresyon i\u00e7in gerekli \u015fartlar\u0131 sa\u011flayamad\u0131k (Baltagi, 2011<\/a>) ve modelimizi parametrik olmayan regresyon ile test etmemiz gerekiyor . Koenker ve  Basset (1978) taraf\u0131ndan geli\u015ftirilen kantil regresyon bu ama\u00e7la kullanabilece\u011finiz y\u00f6ntemlerden biri. Analiz i\u00e7in gereken paket bizzat regresyonu geli\u015ftiren Koenker’e ait. Modeli bir kere \u00e7al\u0131\u015ft\u0131rmak ve kaydetmek yetmeyecek. Ne yapaca\u011f\u0131z peki? Toplamda 7 kere \u00e7al\u0131\u015ft\u0131rm\u0131\u015f olaca\u011f\u0131z. rq<\/em> fonksiyonu i\u00e7indeki tau<\/em> hangi kantil de\u011ferinin test edilece\u011fini g\u00f6stermektedir.<\/p>\n\n\n\n
              model_0.05<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.05)\nmodel_0.10<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.10)\nmodel_0.25<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.25)\nmodel_0.50<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.50)\nmodel_0.75<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.75)\nmodel_0.90<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.90)\nmodel_0.95<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.95)<\/code><\/pre>\n\n\n\n
              Kantil regresyon ve bootstrapping<\/h3>\n\n\n\n
              En k\u00fc\u00e7\u00fck kareler y\u00f6nteminde standart hatalar\u0131 birim\/y\u0131l olarak k\u00fcmeleyerek modelin sa\u011flaml\u0131\u011f\u0131n\u0131 yapm\u0131\u015ft\u0131k. Kantil regresyonda neden b\u00f6yle bir \u015fey yapm\u0131yoruz? Kantil regresyonda standart hatalar\u0131n k\u00fcmelenmesi i\u00e7in sonu\u00e7lar\u0131n yeniden \u00f6ny\u00fcklenmesi gerekiyor (Hao ve Naiman, 2007<\/a>). Bu i\u015flem esas modelde yapt\u0131\u011f\u0131m\u0131za g\u00f6re daha kolay.<\/p>\n\n\n\n
              summary.rq(model_0.05,se=\"boot\")\nsummary.rq(model_0.10,se=\"boot\")\nsummary.rq(model_0.25,se=\"boot\")\nsummary.rq(model_0.50,se=\"boot\")\nsummary.rq(model_0.75,se=\"boot\")\nsummary.rq(model_0.90,se=\"boot\")\nsummary.rq(model_0.95,se=\"boot\")<\/code><\/pre>\n\n\n\n
              Kantil regresyon i\u00e7in R2 hesaplamak<\/h3>\n\n\n\n
              Koenker diyor ki <\/a><\/p>\n\n\n\n
              I don't much like R1, or R2 for that matter, so it isn't likely to be automatically provided in quantreg any time soon.\n(R1 veya R2'den \u00e7ok ho\u015flanm\u0131yorum, bu y\u00fczden de yak\u0131n bir zamanda quantreg'de otomatik olarak yer almas\u0131 olas\u0131 de\u011fil (\u00c7eviri: GC)<\/pre>\n\n\n\n
              Peki siz bu modellerin alt\u0131nda R2<\/sup> eklemek istiyorsan\u0131z ne yapman\u0131z laz\u0131m? (\u00c7\u00f6z\u00fcm i\u00e7in Koenker ve Machado, 1999<\/a>. R-squared in quantile regression<\/a>, Dimitriy V. Masterov (https:\/\/stats.stackexchange.com\/users\/7071\/dimitriy-v-masterov), R-squared in quantile regression, URL (version: 2014-12-17): https:\/\/stats.stackexchange.com\/q\/129251) <\/p>\n\n\n\n
              fit0 <- rq(BD~1,tau=0.95,data=vseti)\nfit1<-rq(BD~X1+KD1+KD2+KD3+KD4+KD5+factor(ID)+factor(FY),data=vseti, tau=0.05)\nrho <- function(u,tau=.5)u*(tau - (u < 0))\nR1 <- 1 - fit1$rho\/fit0$rho\nR1<\/code><\/pre>\n\n\n\n
              Bu form\u00fcl\u00fc her bir kantil noktas\u0131 i\u00e7in tek tek (0.05, 0.10, 0.25, 0.50, 0.75, 0.90 ve 0.95) hesaplaman\u0131z gerekiyor. <\/p>\n\n\n\n
              R Notebook<\/h2>\n\n\n\n
              Sizinle payla\u015ft\u0131\u011f\u0131m kodlar\u0131 \u00e7al\u0131\u015fman\u0131z\u0131n (\u00f6devler de dahil olmak \u00fczere) analizine uygun olarak de\u011fi\u015ftirebilir, sonu\u00e7lar\u0131 tek bir dosyada raporlayabilirsiniz. Kodlar\u0131n ne i\u015fe yarad\u0131\u011f\u0131n\u0131 g\u00f6steremeyen bir not uygulamas\u0131 yerine ger\u00e7ek bir \u00e7al\u0131\u015fma defteri isterseniz R Notebook i\u015finizi g\u00f6recektir. R Markdown<\/a> kullanarak metni istedi\u011finiz gibi  bi\u00e7imlendirebilirsiniz. Kodlar\u0131 R Notebook’ta nas\u0131l kullanacaks\u0131n\u0131z?<\/p>\n\n\n\n
              ```{r}\ndesk<-summary(vseti[,4:10])\ndesk\n```<\/code><\/pre>\n\n\n\n
              Yukar\u0131daki kod y\u0131\u011f\u0131n\u0131 (chunk) hem sizin ald\u0131\u011f\u0131n\u0131z notlar\u0131 saklayacak hem de bu kodun ne anlama geldi\u011fini merak etti\u011finizde y\u0131\u011f\u0131n\u0131n en sa\u011f\u0131ndaki oynat tu\u015funa basarak (“Run Current Chunk”) sonu\u00e7lar\u0131 da g\u00f6rmenizi sa\u011flayacak.<\/p>\n\n\n\n
              Datacamp Career Track<\/h2>\n\n\n\n
              Statistician with R<\/a> (Tamamand\u0131)
              Quantitative Analyst with R<\/a> (Devam ediyor)<\/p>\n","protected":false},"excerpt":{"rendered":"
              R ve Analiz Bu yaz\u0131y\u0131 yazmaktaki amac\u0131m R kullanarak bilimsel \u00e7al\u0131\u015fma yapmak isteyen fakat bir sebepten ba\u015flayamayanlara asl\u0131nda s\u00fcrecin ne kadar kolay oldu\u011funu g\u00f6stermek. Metin i\u00e7inde kullan\u0131lan bilgileri farkl\u0131 kaynaklardan elde ettim: Paketlerin kullan\u0131m\u0131n\u0131, fonksiyonlar\u0131n i\u00e7eri\u011fini paketlerde yer alan vinyetlerden okuyarak \u00f6\u011frendim. Buradaki bilgilerin baz\u0131lar\u0131n\u0131 Datacamp’de ald\u0131\u011f\u0131m R kurslardan edindim. Bug\u00fcne kadar tamamlad\u0131\u011f\u0131m kurslar\u0131n (career […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[9],"tags":[],"class_list":["post-1897","post","type-post","status-publish","format-standard","hentry","category-r"],"_links":{"self":[{"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=\/wp\/v2\/posts\/1897","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1897"}],"version-history":[{"count":19,"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=\/wp\/v2\/posts\/1897\/revisions"}],"predecessor-version":[{"id":2199,"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=\/wp\/v2\/posts\/1897\/revisions\/2199"}],"wp:attachment":[{"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1897"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1897"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.gokberkcan.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1897"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}