{"id":492,"date":"2020-02-05T20:28:19","date_gmt":"2020-02-05T17:28:19","guid":{"rendered":"https:\/\/www.aciltipakademisi.org\/?p=492"},"modified":"2021-11-16T04:10:02","modified_gmt":"2021-11-16T01:10:02","slug":"kategorik-degiskenler-cok-gozlu-tablolar-ve-ki-kare-hesabi","status":"publish","type":"post","link":"https:\/\/tatd.org.tr\/atak\/2020\/02\/05\/kategorik-degiskenler-cok-gozlu-tablolar-ve-ki-kare-hesabi\/","title":{"rendered":"Kategorik de\u011fi\u015fkenler, \u00c7ok g\u00f6zl\u00fc tablolar ve ki-kare hesab\u0131"},"content":{"rendered":"\n<p>\u00c7ok&nbsp;g\u00f6zl\u00fc tablolar iki kalitatif&nbsp;verinin birbiriyle kar\u015f\u0131la\u015ft\u0131r\u0131lmas\u0131nda kulland\u0131\u011f\u0131m\u0131z kar\u015f\u0131la\u015ft\u0131rma metodudur.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Kalitatif veriler s\u0131n\u0131fland\u0131rma&nbsp;belirtir. Bu tip veriyi tutan de\u011fi\u015fkenlere <strong>kategorik\/gruplu\/nominal de\u011fi\u015fkenler <\/strong>ad\u0131 verilir<\/li><li>Bu veri Cinsiyet, meslek, kurum, Well\u2019s risk grubu, HT varl\u0131\u011f\u0131\/yoklu\u011fu vb gibi bir veridir. Verideki say\u0131 bir grubu\/kategoriyi temsil eder.<\/li><li>\u00d6rne\u011fin, Cinsiyet de\u011fi\u015fkeninin de\u011feri olan 0 ya da 1 say\u0131lar\u0131n\u0131n bir anlam\u0131 yoktur, asl\u0131nda bu de\u011fer bir kategorinin kodudur. Cinsiyeti 1&#8217;den k\u00fc\u00e7\u00fck, cinsiyetleri ortalamas\u0131 1,4 gibi bir de\u011ferlendirme anlams\u0131zd\u0131r. Cinsiyet de\u011ferlerinin ortalamas\u0131, medyan\u0131 gibi merkezilik \u00f6l\u00e7\u00fctlerinin de bir anlam\u0131 yoktur.<\/li><li>Her bir grup\/kategori\/fakt\u00f6r i\u00e7indeki&nbsp;miktar say\u0131l\u0131r. \u00d6rne\u011fin, Cinsiyet de\u011fi\u015fkeninin erkek grubundaki ki\u015fi say\u0131s\u0131na o grubun&nbsp;<strong>frekans\u0131&nbsp;<\/strong>denir<\/li><li>Meslek [doktor\/hem\u015fire\/veri giri\u015f] de\u011fi\u015fkeninin 3 grubu grubu\/kategorisi\/fakt\u00f6r\u00fc vard\u0131r<\/li><li>Cinsiyet [erkek\/kad\u0131n], &nbsp;sonlan\u0131m [yat\u0131\u015f\/taburcu], yat\u0131\u015f [servis\/YB\u00dc] de\u011fi\u015fkenlerinin 2 grubu\/kategorisi\/fakt\u00f6r\u00fc vard\u0131r. 2 grubu\/kategorisi\/fakt\u00f6r\u00fc olan de\u011fi\u015fkenlere&nbsp;<strong>Dikotom de\u011fi\u015fken<\/strong> de denilir.<\/li><li>S\u0131ral\u0131 olan&nbsp;kategorik\/gruplu\/nominal de\u011fi\u015fkenlere<strong> Ordinal de\u011fi\u015fken<\/strong>&nbsp;ad\u0131 verilir. \u00d6\u011fretim durumu [ilk\/orta\/lise], kafa travmas\u0131 tipi [hafif\/orta\/a\u011f\u0131r], GKS [3,4,5,&#8230;,14,15] hep ordinal verilerdir.&nbsp;<\/li><\/ul>\n\n\n\n<h4 class=\"wp-block-heading\">Kategorik verilerin SPSS\u2019e girilmesi<\/h4>\n\n\n\n<p>SPSS&#8217;e kategorik de\u011fi\u015fkenler kaydedilirken her kategoriye bir numara verilir [0,1,2,3,4 vb].&nbsp;Her dene\u011fe ait veri ise sat\u0131rlara eklenir. Sadece say\u0131 kullan\u0131l\u0131r, isim&nbsp;ya da harf yaz\u0131lmaz.&nbsp;\u00d6rne\u011fin, cinsiyet de\u011fi\u015fkeninin alt gruplar\u0131\/kategorileri olan erkek ve kad\u0131n verisi yaz\u0131l\u0131rken, her sat\u0131rdaki denek i\u00e7in &#8220;Erkek&#8221; veya &#8220;Kad\u0131n&#8221; metinleri yaz\u0131lmaz, &#8220;E&#8221; veya &#8220;K&#8221; harfleri kullan\u0131lmaz. Bunun yerine Erkek ve Kad\u0131n kategorilerine kar\u015f\u0131l\u0131k gelen, bir kenara not edilmi\u015f, 0 ve 1 gibi say\u0131lar kullan\u0131l\u0131r.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-12.53.20.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-12.53.20-1024x506.png\" alt=\"\" class=\"wp-image-22287\" \/><\/a><\/figure><\/div>\n\n\n\n<p><strong>Variable view <\/strong>ekran\u0131nda, o de\u011fi\u015fkene ait sat\u0131rdaki&nbsp;<strong>Value labels<\/strong> se\u00e7ene\u011fi ile, o de\u011fi\u015fkenin kategorileri i\u00e7in tan\u0131mlanan numaralar\u0131n hangi anlama geldi\u011fini unutmamak i\u00e7in kaydedebilirsiniz. Ancak dosyan\u0131z\u0131 kaydederken (save data) .sav dosyas\u0131 d\u0131\u015f\u0131ndaki kay\u0131t sistemlerinin bu bilgiyi saklamad\u0131\u011f\u0131n\u0131 unutmay\u0131n.<\/p>\n\n\n\n<h4 class=\"wp-block-heading\">Kategorik de\u011fi\u015fkenin bildirilmesi<\/h4>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"alignright\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.17.44.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.17.44-300x128.png\" alt=\"\" class=\"wp-image-22289\" \/><\/a><\/figure><\/div>\n\n\n\n<p>Kategorik bir de\u011fi\u015fken bildirilirken frekans\u0131yla ifade edilir.&nbsp;SPSS\u2019de <strong>Analyze &gt; Descriptive Statistics &gt; Frequencies&nbsp;<\/strong>men\u00fcs\u00fcnden ilgili de\u011fi\u015fkene ait d\u00f6k\u00fcm yap\u0131l\u0131r.<\/p>\n\n\n\n<p>\u0130lgili men\u00fcye girilince a\u015fa\u011f\u0131daki se\u00e7im ekran\u0131 g\u00f6r\u00fcl\u00fcr. \u0130stenen kategorik de\u011fi\u015fkenler de\u011fi\u015fken alan\u0131na al\u0131n\u0131r.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.19.28.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.19.28-300x177.png\" alt=\"\" class=\"wp-image-22291\" \/><\/a><\/figure><\/div>\n\n\n\n<p>OK tu\u015funa bas\u0131ld\u0131\u011f\u0131ndan a\u015fa\u011f\u0131da listelenen d\u00f6k\u00fcmler al\u0131n\u0131r.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.19.42.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.19.42-706x1024.png\" alt=\"\" class=\"wp-image-22290\" \/><\/a><\/figure><\/div>\n\n\n\n<p>Kategorik de\u011fi\u015fken, hasta gruplar\u0131n\u0131n temel \u00f6zelliklerini i\u00e7eriyorsa \u00e7al\u0131\u015fman\u0131n 1. Tablosunda g\u00f6sterilmelidir.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.22.44.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.22.44-864x1024.png\" alt=\"\" class=\"wp-image-22292\" \/><\/a><\/figure><\/div>\n\n\n\n<h3 class=\"wp-block-heading\">El ile Ki-kare hesab\u0131<\/h3>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"alignright\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/1.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/1-300x147.png\" alt=\"\" class=\"wp-image-22278\" \/><\/a><\/figure><\/div>\n\n\n\n<p>\u00c7ok g\u00f6zl\u00fc tablolara Olas\u0131l\u0131k tablolar\u0131 (contingency tables) ya da \u00e7apraz tablolar (crosstabs) da denilir.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>S\u0131f\u0131r hipotezi: iki de\u011fi\u015fken aras\u0131nda ili\u015fki ya da korelasyon yoktur<\/li><li>Alternatif hipotez: iki de\u011fi\u015fken aras\u0131nda ili\u015fki ya da korelasyon vard\u0131r<\/li><\/ul>\n\n\n\n<p>\u00c7apraz tablolar olu\u015fturulurken etken, fakt\u00f6r ya da ba\u011f\u0131ms\u0131z de\u011fi\u015fken sat\u0131ra; etkilenen, sonlan\u0131m ya da ba\u011f\u0131ml\u0131 de\u011fi\u015fken s\u00fctuna konulur.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"alignright\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/2.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/2-300x131.png\" alt=\"\" class=\"wp-image-22279\" \/><\/a><\/figure><\/div>\n\n\n\n<h4 class=\"wp-block-heading\"><em><strong>Pearson\u2019s chi-square (Pearson ki-kare).<\/strong><\/em><\/h4>\n\n\n\n<ul class=\"wp-block-list\"><li>Klasik ki-kare testidir. \u00d6rneklem say\u0131s\u0131ndan son derece etkilenir. \u00c7ok geni\u015f \u00f6rneklerde en ufak sapma bile anlaml\u0131yken k\u00fc\u00e7\u00fck \u00f6rneklemlerde b\u00fcy\u00fck sapmalar bile anlams\u0131zd\u0131r.<\/li><li>\u00d6rneklem boyutu k\u00fc\u00e7\u00fcld\u00fck\u00e7e de\u011feri d\u00fc\u015fer, bu y\u00fczden de \u00f6rneklem boyutuna ba\u011fl\u0131 <strong>\u0130ki \u00f6n \u015fart\u0131<\/strong> vard\u0131r:\n<ol><li>\u00c7ok g\u00f6zl\u00fc tablonun her kutusunda en az\u0131ndan 1 vaka olmal\u0131d\u0131r.<\/li><li>\u00c7ok g\u00f6zl\u00fc tablonun kutular\u0131ndan en fazla %20\u2019sinde (yani 5&#8217;de 1&#8217;inde) 5\u2019den az vaka olmal\u0131d\u0131r.<\/li><\/ol>\n<\/li><li>Yukar\u0131daki \u015fartlar\u0131n sa\u011fland\u0131\u011f\u0131 her durumda, ve 2 x 2 d\u0131\u015f\u0131ndaki her t\u00fcr tabloda <strong>Pearson chi-square <\/strong>hesaplan\u0131r<\/li><\/ul>\n\n\n\n<h4 class=\"wp-block-heading\"><em><strong>Fisher\u2019s exact test (Fisher\u2019in kesin testi)<\/strong><\/em><\/h4>\n\n\n\n<ul class=\"wp-block-list\"><li>Pearson ki-kare\u2019nin \u015f\u00fcpheli sonu\u00e7 verdi\u011fi k\u00fc\u00e7\u00fck \u00f6rneklem boyutlar\u0131nda kullan\u0131l\u0131r.<\/li><li>Kesin (exact) testlerden olup yakla\u015f\u0131k de\u011fil tam de\u011ferleri hesaplar<\/li><li>Eskiden sadece 2 x 2 tablolar i\u00e7in hesaplanabilirken art\u0131k her t\u00fcrl\u00fc tabloda hesaplanabilmektedir. E\u011fer herhangi ba\u015fka bir se\u00e7enek se\u00e7ilmezse SPSS sadece 2 x 2 tablolarda hesaplama yapar. E\u011fer <strong>Exact<\/strong> ayarlar\u0131ndan <strong>Monte Carlo<\/strong> se\u00e7ene\u011fi aktif hale getirilirse m x n say\u0131da kutusu olan herhangi bir \u00e7okg\u00f6zl\u00fc tablo i\u00e7in de hesaplama yap\u0131l\u0131r ve \u00e7\u0131kt\u0131da bildirilir. Bu durumda p de\u011ferinin %95 g\u00fcven aral\u0131\u011f\u0131 da bildirilir.<\/li><\/ul>\n\n\n\n<p>E\u011fer iki fakt\u00f6r\/grup (sat\u0131r) aras\u0131nda fark yoksa, her s\u00fctunda sat\u0131rlara kar\u015f\u0131l\u0131k gelen kutularda ayn\u0131 oranda vaka olmas\u0131 gerekir. \u00d6rne\u011fin, yukar\u0131daki tabloda 112 hastadan 42\u2019si kad\u0131nd\u0131r. O zaman sedasyon i\u00e7in ek doz (s\u00fctun) gereken (yes) 25 hastadan 42\/112 kadar\u0131n\u0131n (9,4) kad\u0131nlar aras\u0131ndan \u00e7\u0131kmas\u0131n\u0131 beklerim.&nbsp; Halbuki bekledi\u011fimizden daha az say\u0131da (4) vaka bu gruptan \u00e7\u0131km\u0131\u015ft\u0131r. Alttaki tabloda bu \u015fekilde hesaplanan beklenen (expected) de\u011ferler ile g\u00f6zlenen (observed) de\u011ferler verilmi\u015ftir.<\/p>\n\n\n\n<p>Ki-kare testinin sonucu son derece mant\u0131kl\u0131 bir hesaba dayan\u0131r. \u00dcstteki tabloda g\u00f6r\u00fclen her bir kutuya d\u00fc\u015fen say\u0131larla, alttaki tabloda ifade edilen beklenen (expected) say\u0131lar, yani beklenen ve g\u00f6zlenen de\u011ferler aras\u0131ndaki fark, ki-kare da\u011f\u0131l\u0131m\u0131 ile kar\u015f\u0131la\u015ft\u0131r\u0131l\u0131r.<\/p>\n\n\n<p>[table id=34 \/]<\/p>\n\n\n\n<p>(G\u00f6zlenen \u2013 Beklenen)<sup>2 <\/sup>\/ Beklenen<\/p>\n\n\n\n<p>T\u00fcm kutular i\u00e7in yukar\u0131daki form\u00fcle g\u00f6re sapma miktarlar\u0131 hesaplan\u0131r ve hepsi toplan\u0131r. Bu toplamda elde edilen say\u0131 ki-kare de\u011feridir.<\/p>\n\n\n<p>[table id=35 \/]<\/p>\n\n\n\n<p>\u03c7(chi)<sup>2<\/sup>= 0,89 + 3,10 + 0,54 + 1,87 = 6,4<\/p>\n\n\n\n<p>Ki-kare de\u011ferinin ba\u011f\u0131ms\u0131zl\u0131k derecelerine g\u00f6re en fazla ka\u00e7 olmas\u0131 gerekti\u011fi bellidir. Ba\u011f\u0131ms\u0131zl\u0131k derecesi (df = degrees of freedom) ise \u015fu \u015fekilde hesaplan\u0131r:<\/p>\n\n\n\n<p>df = (sat\u0131r say\u0131s\u0131 \u2013 1) x (s\u00fctun say\u0131s\u0131 \u2013 1)<\/p>\n\n\n\n<p>df = (2 \u2013 1) x (2 \u2013 1) = 1 x 1 = 1<\/p>\n\n\n\n<p>\u015eimdi buldu\u011fumuz ki-kare (\u03c7<sup>2 <\/sup>) de\u011ferini (6,4) a\u015fa\u011f\u0131daki ki-kare (\u03c7<sup>2 <\/sup>) da\u011f\u0131l\u0131m tablosuyla kar\u015f\u0131la\u015ft\u0131ral\u0131m.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/3.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/3.png\" alt=\"ki kare da\u011f\u0131l\u0131m\u0131\" class=\"wp-image-22284\" \/><\/a><\/figure><\/div>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/4.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/4.png\" alt=\"ki kare da\u011f\u0131l\u0131m tablosu\" class=\"wp-image-22285\" \/><\/a><\/figure><\/div>\n\n\n\n<p>Tablomuzun df de\u011feri 1 olan sat\u0131rda 6,4 de\u011feri 0,025 ile 0,01 olas\u0131l\u0131klar\u0131 aras\u0131nda bir olas\u0131l\u0131\u011fa tekab\u00fcl etmektedir.&nbsp;Bu tablodaki ili\u015fki ile ilgili olarak p de\u011ferinin 0,01 civar\u0131nda oldu\u011funu s\u00f6yleyebiliriz.<\/p>\n\n\n\n<p>A\u015fa\u011f\u0131da SPSS taraf\u0131ndan yap\u0131lan hesaplama g\u00f6sterilmi\u015ftir. SPSS taraf\u0131ndan hesaplanan \u03c7<sup>2 <\/sup>= 6,348 bizim hesaplar\u0131m\u0131zla ayn\u0131d\u0131r (Pearson chi-square de\u011feri). P de\u011feri de 0,012 olarak hesaplanm\u0131\u015ft\u0131r (asymptotic significance).<\/p>\n\n\n\n<p>Fisher\u2019in kesin testi (Fisher&#8217;s exact test) ad\u0131ndan da anla\u015f\u0131laca\u011f\u0131 gibi daha &#8220;kesin&#8221; sonu\u00e7 vermekte olup, SPSS, bu test ile p de\u011ferini 0,018 olarak&nbsp;hesaplam\u0131\u015ft\u0131r (Exact sig.). Fisher testiyle hesapanan p de\u011feri ki-kare ile hesaplanandan daha b\u00fcy\u00fck yani anlams\u0131zl\u0131k s\u0131n\u0131r\u0131na daha yak\u0131nd\u0131r.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/5.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/5.png\" alt=\"\" class=\"wp-image-22286\" \/><\/a><\/figure><\/div>\n\n\n\n<h3 class=\"wp-block-heading\">SPSS ile Ki-kare testi<\/h3>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"alignright\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.36.48.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.36.48-300x126.png\" alt=\"\" class=\"wp-image-22294\" \/><\/a><\/figure><\/div>\n\n\n\n<p><strong>Analyze &gt; Descriptive Statistics &gt; Crosstabs<\/strong> men\u00fcs\u00fcne girildi\u011finde a\u015fa\u011f\u0131daki ekranla kar\u015f\u0131la\u015ft\u0131r\u0131z. Bu ekranda \u00e7ok g\u00f6zl\u00fc tablonun sat\u0131rlar\u0131nda yer alacak de\u011fi\u015fkenleri row k\u0131sm\u0131na, s\u00fct\u00fcnlar\u0131nda yer alacak de\u011fi\u015fkenleri column k\u0131sm\u0131na yerle\u015ftirmeliyiz.<\/p>\n\n\n\n<p>Genel kural olarak ba\u011f\u0131ms\u0131z de\u011fi\u015fkenler, predikt\u00f6r fakt\u00f6rler, risk fakt\u00f6rleri, etkisini inceledi\u011fimiz giri\u015fimler ya da farkl\u0131 tedavileri uygulad\u0131\u011f\u0131m\u0131z gruplar\u0131 belirten de\u011fi\u015fkenler sat\u0131rlara (row) yaz\u0131l\u0131r.&nbsp;Ba\u011f\u0131ml\u0131 de\u011fi\u015fkenler, bekledi\u011fimiz etkiyi g\u00f6steren de\u011fi\u015fkenler, sonu\u00e7lar, sonlan\u0131mlar s\u00fctunlara yaz\u0131l\u0131r.&nbsp;Ancak sat\u0131r ya da s\u00fctuna hangisinin yaz\u0131ld\u0131\u011f\u0131 hesaplamalar\u0131 de\u011fi\u015ftirmez. E\u011fer bu tip etki(sebep) \u2013 sonu\u00e7 ili\u015fkisi bulunmayan de\u011fi\u015fkenlerle ilgili ili\u015fkiyi inceleyen analizler yap\u0131lacaksa istenildi\u011fi \u015fekilde yaz\u0131lmas\u0131nda bir sak\u0131nca yoktur.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.43.36.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.43.36-1024x704.png\" alt=\"\" class=\"wp-image-22295\" \/><\/a><\/figure><\/div>\n\n\n\n<p>Sa\u011f tarafta analize yard\u0131mc\u0131 olacak ayarlamalar\u0131n yap\u0131ld\u0131\u011f\u0131 <strong>Exact, Statistics, Cells, Format, Style<\/strong> ve <strong>Bootstrap<\/strong> se\u00e7enekleri mevcuttur.<\/p>\n\n\n\n<p><strong><em>Layer<\/em><\/strong> k\u0131sm\u0131na ise se\u00e7ti\u011fimiz sat\u0131r ve s\u00fctun de\u011fi\u015fkenlerini etkileyece\u011fini d\u00fc\u015f\u00fcnd\u00fc\u011f\u00fcm\u00fcz <strong>covariate<\/strong> ad\u0131 verilen <strong>kar\u0131\u015ft\u0131r\u0131c\u0131 de\u011fi\u015fkenler<\/strong> eklenir. Ayn\u0131 anda birden fazla ki-kare testinin birarada yap\u0131lmas\u0131 ve kar\u0131\u015ft\u0131r\u0131c\u0131 de\u011fi\u015fkenin etkisinin ar\u0131nd\u0131r\u0131lmas\u0131 i\u00e7in kullan\u0131l\u0131r. Statistics se\u00e7eneklerinde <strong><em>Cochran&#8217;s and Mantel-Haenszel statistics <\/em><\/strong>se\u00e7ilerek bu katmanlar\u0131n etkisi ayn\u0131 anda incelenebilir.<\/p>\n\n\n\n<p>En basit \u015fekilde hi\u00e7bir ayar yap\u0131lmadan al\u0131nan d\u00f6k\u00fcm a\u015fa\u011f\u0131daki gibi olacakt\u0131r:<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.43.49.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/Screen-Shot-2017-05-07-at-13.43.49-300x162.png\" alt=\"\" class=\"wp-image-22296\" \/><\/a><\/figure><\/div>\n\n\n\n<h4 class=\"wp-block-heading\"><em><strong>Hangi test se\u00e7ilmeli?<\/strong><\/em><\/h4>\n\n\n\n<ul class=\"wp-block-list\"><li>Verinin t\u00fcr\u00fcne g\u00f6re, \u00f6ncesi-sonras\u0131 gibi bir durum olmayan <strong>ba\u011f\u0131ms\u0131z de\u011fi\u015fkenler i\u00e7in \u00e7okg\u00f6zl\u00fc tablolarda <em>ki-kare (chi-square)<\/em><\/strong> se\u00e7ilir.<\/li><li><strong>Ba\u011f\u0131ml\u0131 de\u011fi\u015fkenler i\u00e7in <em>McNemar<\/em><\/strong> se\u00e7ilir. Ya statistics ayarlar\u0131ndan, ya da <strong>Analyze &gt;&gt; Non-parametrik &gt;&gt; Legacy &gt;&gt; 2 related samples<\/strong> se\u00e7ene\u011finden McNemar se\u00e7ilebilir.<\/li><li><strong>3 ve daha fazla grubu olan ordinal de\u011fi\u015fkenler i\u00e7in <em>Cochrane Q<\/em><\/strong> hesaplan\u0131r.\n<ul><li>Bu hesaplama i\u00e7in <strong>Analyze &gt;&gt; Non-parametrik &gt;&gt; Legacy &gt;&gt; K related samples<\/strong> se\u00e7ene\u011fine gidilir.<\/li><\/ul>\n<\/li><\/ul>\n\n\n\n<h4 class=\"wp-block-heading\"><em><strong>Hangi ili\u015fki katsay\u0131s\u0131 se\u00e7ilmeli?<\/strong><\/em><\/h4>\n\n\n\n<p>En temel olarak ki-kare analizinde analiz testlerinden biri ile <strong>ili\u015fki b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc g\u00f6steren<\/strong> katsay\u0131 testlerinden biri de se\u00e7ilmelidir.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>1 ya da 2 de\u011fi\u015fken <strong>nominalse<\/strong> &gt; <em><strong>Cramer\u2019s v<\/strong><\/em><\/li><li>Her ikisi de <strong>ordinalse<\/strong> &gt; <em><strong>Kendall\u2019s tau b veya c<\/strong><\/em> (sat\u0131r, s\u00fctun say\u0131s\u0131 e\u015fitli\u011fine g\u00f6re)<\/li><\/ul>\n\n\n\n<p>Dolay\u0131s\u0131yla e\u011fer ki-kare testi isteniyorsa&nbsp;<em><strong>statistics<\/strong><\/em> se\u00e7eneklerinden a\u015fa\u011f\u0131dakiler se\u00e7ilir:<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/6.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/6.png\" alt=\"\" class=\"wp-image-22297\" \/><\/a><\/figure><\/div>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/7.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/7.png\" alt=\"\" class=\"wp-image-22298\" \/><\/a><\/figure><\/div>\n\n\n\n<p><strong><em>Cells<\/em> <\/strong>se\u00e7eneklerinden de k\u0131sm\u0131nda da a\u015fa\u011f\u0131daki&nbsp;tabloda g\u00f6sterilen kutular\u0131n se\u00e7ilmesi yeterlidir.<br><\/p>\n\n\n\n<p>OK tu\u015funa bas\u0131larak analiz tamamlan\u0131r ve \u00e7\u0131kt\u0131lar de\u011ferlendirilir.<\/p>\n\n\n\n<p>Elde edilen analiz \u00e7\u0131kt\u0131s\u0131 yanda ve a\u015fa\u011f\u0131da verilmi\u015ftir. T\u00fcm \u00e7\u0131kt\u0131lar\u0131n anlamlar\u0131 ayarlarla ilgili k\u0131s\u0131mlarda a\u00e7\u0131klanm\u0131\u015ft\u0131r. <\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/8.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/8.png\" alt=\"\" class=\"wp-image-22299\" \/><\/a><\/figure><\/div>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/9.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/9.png\" alt=\"\" class=\"wp-image-22300\" \/><\/a><\/figure><\/div>\n\n\n\n<h4 class=\"wp-block-heading\"><em><strong>Statistics ayarlar\u0131<\/strong><\/em><\/h4>\n\n\n\n<h5 class=\"wp-block-heading\"><strong><em>Chi-square (ki-kare).<\/em><\/strong><\/h5>\n\n\n\n<p>Bu se\u00e7enek se\u00e7ildi\u011finde 2 sat\u0131r ve 2 s\u00fctundan olu\u015fan 2 x 2 tablolarda:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Pearson\u2019s chi-square<\/li><li>Continuity correction<\/li><li>Likelihood ratio<\/li><li>Fisher\u2019s exact test<\/li><li>Lineer-by-lineer association,<\/li><\/ul>\n\n\n\n<p>Daha fazla sat\u0131r ve s\u00fctundan olu\u015fan tablolarda ise:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Pearson\u2019s chi-square<\/li><li>Likelihood ratio<\/li><li>Lineer-by-lineer association,<\/li><\/ul>\n\n\n\n<p>sonu\u00e7lar\u0131 listelenecektir.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Pearson\u2019s chi-square (Pearson ki-kare). <\/strong>\n<ul><li>Klasik ki-kare testidir. \u00d6rneklem say\u0131s\u0131ndan son derece etkilenir. \u00c7ok geni\u015f \u00f6rneklerde en ufak sapma bile anlaml\u0131yken k\u00fc\u00e7\u00fck \u00f6rneklemlerde b\u00fcy\u00fck sapmalar bile anlams\u0131zd\u0131r.<\/li><li>\u00d6rneklem boyutu k\u00fc\u00e7\u00fcld\u00fck\u00e7e de\u011feri d\u00fc\u015fer, bu y\u00fczden de \u00f6rneklem boyutuna ba\u011fl\u0131 <strong>\u0130ki \u00f6n \u015fart\u0131<\/strong> vard\u0131r:\n<ol><li>Her kutuda en az\u0131ndan 1 vaka olmal\u0131d\u0131r.<\/li><li>Kutulardan en fazla %20\u2019sinde 5\u2019den az vaka olmal\u0131d\u0131r.<\/li><\/ol>\n<\/li><li>Yukar\u0131daki \u015fartlar\u0131n sa\u011fland\u0131\u011f\u0131 her durumda, ve 2 x 2 d\u0131\u015f\u0131ndaki her t\u00fcr tabloda <strong>Pearson chi-square <\/strong>hesaplan\u0131r<\/li><\/ul>\n<\/li><li><strong>Fisher\u2019s exact test (Fisher\u2019in kesin testi) <\/strong>\n<ul><li>Pearson ki-kare\u2019nin \u015f\u00fcpheli sonu\u00e7 verdi\u011fi k\u00fc\u00e7\u00fck \u00f6rneklem boyutlar\u0131nda kullan\u0131l\u0131r.<\/li><li>Kesin (exact) testlerden olup yakla\u015f\u0131k de\u011fil tam de\u011ferleri hesaplar<\/li><li>Eskiden sadece 2 x 2 tablolar i\u00e7in hesaplanabilirken art\u0131k her t\u00fcrl\u00fc tabloda hesaplanabilmektedir. E\u011fer herhangi ba\u015fka bir se\u00e7enek se\u00e7ilmezse SPSS sadece 2 x 2 tablolarda hesaplama yapar. E\u011fer <strong>Exact<\/strong> ayarlar\u0131ndan <strong>Monte Carlo<\/strong> se\u00e7ene\u011fi aktif hale getirilirse m x n say\u0131da kutusu olan herhangi bir \u00e7okg\u00f6zl\u00fc tablo i\u00e7in de hesaplama yap\u0131l\u0131r ve \u00e7\u0131kt\u0131da bildirilir. Bu durumda p de\u011ferinin %95 g\u00fcven aral\u0131\u011f\u0131 da bildirilir.<\/li><\/ul>\n<\/li><li><strong>Continuity correction <\/strong>\n<ul><li><strong>= <\/strong><strong>Yates&#8217; corrected chi-square (Yates\u2019 d\u00fczeltmeli ki-kare)<\/strong><\/li><li>Sadece 2 x 2 tablolarda hesaplan\u0131r<\/li><li>Beklenen ve g\u00f6zlenen de\u011fer aras\u0131ndaki farktan 0,5 \u00e7\u0131kar\u0131r. B\u00f6ylece hesaplanan ki-kare de\u011feri k\u00fc\u00e7\u00fcl\u00fcr, p de\u011feri ise b\u00fcy\u00fcr.<\/li><li>\u00d6zellikle k\u00fc\u00e7\u00fck \u00f6rneklemler fark yokken fark bulma ihtimalini azaltmak i\u00e7in \u0130ngiliz istatistik\u00e7i Yates taraf\u0131ndan \u00f6nerilmi\u015ftir.<\/li><li>En az 1 kutuda beklenen frekans 5\u2019den k\u00fc\u00e7\u00fck ise <strong>Fisher&#8217;s exact test <\/strong>yerine<strong> Yates&#8217; corrected chi-square<\/strong> Buna ra\u011fmen \u00f6zellikle k\u00fc\u00e7\u00fck \u00f6rneklemlerde gere\u011finden fazla d\u00fczeltme yapt\u0131\u011f\u0131n\u0131, s\u0131f\u0131r hipotezini reddetmesi gerekirken reddedemedi\u011fini ve b\u00f6ylece tip 2 hatay\u0131 artt\u0131rd\u0131\u011f\u0131n\u0131 savunanlar da vard\u0131r.<\/li><li>Yukar\u0131daki \u00f6rnekte SPSS taraf\u0131ndan hesaplanan c<sup>2 <\/sup>de\u011ferinin Pearson\u2019daki 6,4 de\u011ferinden Yates d\u00fczeltmesi ile 5,2\u2019ye d\u00fc\u015ft\u00fc\u011f\u00fcn\u00fc, p de\u011ferinin de 0,012 yerine 0,022\u2019ye y\u00fckseldi\u011fini g\u00f6r\u00fcyoruz. Yates d\u00fczeltmesi ile Fisher\u2019in kesin testi ile hesaplanan p de\u011feri olan 0,018\u2019den bile y\u00fcksek bir de\u011fer hesapland\u0131\u011f\u0131na dikkat ediniz.<\/li><\/ul>\n<\/li><li><strong>Linear-by-linear association<\/strong>\n<ul><li><strong>= Mantel-Haenszel test of trend <\/strong><\/li><li><strong>= Mantel-Haenszel test of Linear Association<\/strong><\/li><li>(Buradaki a\u00e7\u0131klamalar Martin Bland\u2019\u0131n York \u00dcniversitesindeki ders notlar\u0131ndan al\u0131nm\u0131\u015ft\u0131r \u2013 HA)<\/li><li>En az biri ordinal olan iki de\u011fi\u015fken aras\u0131ndaki ili\u015fkiyi verir. \u00d6rne\u011fin,\n<ul><li>bir \u00f6nceki y\u0131lki de\u011ferlendirme notu ile sonraki y\u0131l terfi etme aras\u0131ndaki ili\u015fi gibi.<\/li><li>Verilen tedavi ile taburculuktan \u00f6l\u00fcme kadar de\u011fi\u015fen bir seri sonlan\u0131m gibi<\/li><\/ul>\n<\/li><li>Bu tip durumlarda ordinal de\u011fi\u015fkenin s\u0131ras\u0131 \u00f6nemli olup giderek azalan ya da artan trendlerin de hesaba kat\u0131lmas\u0131 \u00f6nemlidir. SPSS bu ama\u00e7la 3 test yapabilir:\n<ul><li>Armitage chi-squared test for trend<\/li><li>Mantel-Haenszel test of trend, ve<\/li><li>Kendall\u2019s rank correlation tau b<\/li><\/ul>\n<\/li><li>Mantel-Haenszel test of trend, biz istesek de istemesek de SPSS taraf\u0131ndan Linear-by-linear association ad\u0131 alt\u0131nda otomatik olarak ve her t\u00fcrl\u00fc tablo i\u00e7in bildirilmektedir. Sat\u0131r ve s\u00fctun de\u011fi\u015fkenleri aras\u0131nda lineer bir ili\u015fki olup olmad\u0131\u011f\u0131n\u0131 test eder. Lineer d\u0131\u015f\u0131nda ba\u015fka t\u00fcrl\u00fc bir ili\u015fki olsa bile bunu de\u011ferlendirmedi\u011finden bu konuda bilgi vermez.<\/li><li>Lineer ili\u015fkiden kas\u0131t ise, ordinal kategorilere giderek artan say\u0131lar verip (iyile\u015fme d\u00fczeyi = 1, 2, 3, 4 vb), tedavi ya da gruplar\u0131 da numaralay\u0131p (tedavi 1 =1, tedavi 2=2), bunu bir form\u00fcl i\u00e7ine yerle\u015ftirebilmektir (iyile\u015fme d\u00fczeyi = sabit x tedavi). Buradaki sabit say\u0131dan ziyade de\u011fi\u015fen kategoriler i\u00e7in sabitteki de\u011fi\u015fim hesaplan\u0131r ve buna da Mantel-Haenszel test of trend ad\u0131 verilir.<\/li><li>Ki-kare testi yap\u0131lamaz olsa bile (\u00f6n\u015fart sa\u011flanmasa dahi) toplamda 30 vaka varsa Mantel-Haenszel test of trend yap\u0131labilir ve anlaml\u0131 sonu\u00e7lar verir.<\/li><\/ul>\n<\/li><li><strong>Likelihood ratio (olabilirlik olas\u0131l\u0131\u011f\u0131) ki-kare. <\/strong>\n<ul><li>Geni\u015f \u00f6rneklemlerde Pearson ki-kare ile ayn\u0131 sonucu verir. \u00d6zellikle az say\u0131da \u00f6rneklemin oldu\u011fu tablolarda faydal\u0131d\u0131r. LR\u2019ler bildirilir.<\/li><\/ul>\n<\/li><\/ul>\n\n\n\n<p>Bu k\u0131sma kadar olan testlerle <strong>2 de\u011fi\u015fken aras\u0131nda bir ili\u015fki ya da fark olup olmad\u0131\u011f\u0131 sorusunun cevab\u0131<\/strong> aranm\u0131\u015ft\u0131r. Bundan sonraki se\u00e7enekler ise var olan ili\u015fkinin b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc g\u00f6stermek i\u00e7in kullan\u0131lan testleri i\u00e7erir.<\/p>\n\n\n\n<p>Bir \u00e7apraz tablo de\u011ferlendirilirken bulunan de\u011ferlerin beklenen de\u011ferlerden sapmas\u0131n\u0131n b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc hesaplay\u0131p, bu b\u00fcy\u00fckl\u00fc\u011f\u00fcn g\u00f6r\u00fclme ihtimalini tablodan bulmu\u015ftuk. Bu ihtimal %5\u2019den az ise 2 de\u011fi\u015fken aras\u0131nda anlaml\u0131 ili\u015fki var demi\u015ftik. Bu ili\u015fkinin b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc tan\u0131mlamak i\u00e7in de\u011fi\u015fkenin tipine g\u00f6re a\u015fa\u011f\u0131daki se\u00e7enekler kullan\u0131l\u0131r.<\/p>\n\n\n\n<p>T\u00fcm bu ili\u015fki b\u00fcy\u00fckl\u00fc\u011f\u00fc belirten \u00f6l\u00e7\u00fctler i\u00e7in ortak bir tan\u0131mdan bahsedebiliriz. Buna <strong>Proportional Reduction in Error (PRE)<\/strong> ad\u0131 verilir. Temel olarak, ba\u011f\u0131ms\u0131z de\u011fi\u015fkenin (etken, fakt\u00f6r) bilinmesi sayesinde ba\u011f\u0131ml\u0131 de\u011fi\u015fkenin (sonlan\u0131m\u0131n) ne d\u00fczeyde tahmin edilebilece\u011finin \u00f6l\u00e7\u00fct\u00fcd\u00fcr.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\"><strong><em>Correlations (korelasyonlar).<\/em><\/strong><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>Hem sat\u0131r hem de s\u00fctunlar\u0131nda s\u0131ral\u0131 (ordinal) de\u011fi\u015fkenler olan tablolarda bu se\u00e7enek kullan\u0131l\u0131r.<\/li><li><strong>Spearman&#8217;\u0131n korelasyon katsay\u0131s\u0131 olan ro\u2019nun (Spearman\u2019s correlation coefficient, rho)<\/strong> raporlanmas\u0131n\u0131 sa\u011flar. Spearman&#8217;\u0131n ro\u2019su s\u0131ral\u0131 (ordinal) d\u00fczenlerin aras\u0131ndaki ili\u015fkinin b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc \u00f6l\u00e7er.<\/li><li>Her iki de\u011fi\u015fken de kantitatif ise <strong>Pearson korelasyon katsay\u0131s\u0131 olan r<\/strong>\u2019yi bildirir (Pearson\u2019s correlation coefficient,<em>r<\/em>). r de\u011fi\u015fkenler aras\u0131ndaki lineer ili\u015fkinin b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc bildiren bir \u00f6l\u00e7\u00fctt\u00fcr.<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><strong><em>Nominal.<\/em><\/strong><\/h5>\n\n\n\n<p>Kendi i\u00e7inde bir s\u0131ras\u0131 olmayan kategorik de\u011fi\u015fkenler i\u00e7in bu se\u00e7enekler se\u00e7ilebilir.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Lambda (<\/strong><strong>l) ve Goodman&nbsp;ve Kruskal&#8217;s tau (<\/strong><strong>t)<\/strong>\n<ul><li>Simetrik ve asimetrik Lambda ile <strong>Goodman&nbsp;ve Kruskal&#8217;s tau (<\/strong><strong>t)<\/strong> katsay\u0131lar\u0131n\u0131 verir.<\/li><li><strong>Proportional Reduction in Error (PRE) <\/strong>de\u011ferini verir.<\/li><li>1 de\u011feri ba\u011f\u0131ms\u0131z de\u011fi\u015fkenin ba\u011f\u0131ml\u0131 de\u011fi\u015fkeni m\u00fckemmel \u015fekilde \u00f6ng\u00f6rebildi\u011fini g\u00f6sterir. 0 de\u011feri ba\u011f\u0131ms\u0131z de\u011fi\u015fkenin de\u011ferinden ba\u011f\u0131ml\u0131 de\u011fi\u015fkenin \u00f6ng\u00f6r\u00fclme \u015fans\u0131 olmad\u0131\u011f\u0131n\u0131 belirtir.<\/li><li>SPSS hangi de\u011fi\u015fken ba\u011f\u0131ml\u0131 hangisi ba\u011f\u0131ms\u0131z bilemeyece\u011finden ikisini de hesaplar. Simetrik olan \u00f6l\u00e7\u00fct\u00fcn bizim i\u00e7in \u00e7ok bir anlam\u0131 yoktur.<\/li><li>Asimetriklerden do\u011fru s\u0131rada olan se\u00e7ilmelidir.<\/li><li>Lambda (l) olduk\u00e7a konservatif bir \u00f6l\u00e7\u00fctt\u00fcr.<\/li><li>Goodman ve Kruskal\u2019s tau (t) lambda\u2019dan daha iyi olsa da onlar da konservatiftir.<\/li><\/ul>\n<\/li><li>Lambda (l) ve Goodman&nbsp;ve Kruskal&#8217;s tau (t)\u2019nun bu eksiklikleri nedeniyle ki-kareden t\u00fcretilmi\u015f <strong>Phi, Cramer\u2019s V<\/strong> ve <strong>contingency coefficient<\/strong> katsay\u0131lar\u0131 da kullan\u0131lmaktad\u0131r.&nbsp;Ancak ki-kareden t\u00fcretilen a\u015fa\u011f\u0131daki katsay\u0131lar PRE de\u011ferini veremez.<\/li><li><strong>Contingency coefficient<\/strong>\n<ul><li>Ki-kare testine ba\u011fl\u0131 olarak hesaplan\u0131r. 0 ile 1 aras\u0131nda de\u011fi\u015fir. 0, sat\u0131r ve s\u00fctun de\u011fi\u015fkenleri aras\u0131nda hi\u00e7 ili\u015fki olmad\u0131\u011f\u0131n\u0131, 1 ise tam bir ili\u015fki oldu\u011funu g\u00f6sterir.<\/li><li>Maksimum alabilece\u011fi de\u011fer tablonun sat\u0131r ve s\u00fctun say\u0131s\u0131ndan etkilenir. Bu y\u00fczden farkl\u0131 sat\u0131r ve s\u00fctun say\u0131lar\u0131na sahip farkl\u0131 tablolar\u0131 birbiriyle kar\u015f\u0131la\u015ft\u0131rmada de\u011fersizdir.<\/li><\/ul>\n<\/li><li><strong>Phi&nbsp;<\/strong><strong>(<\/strong><strong>j<\/strong><strong>)<\/strong> ve<strong> Cram\u00e9r&#8217;s <\/strong><strong>n<\/strong>\n<ul><li><strong>Phi (<\/strong><strong>j<\/strong><strong>),<\/strong> ki-kare istatistik de\u011ferini \u00f6rneklem boyutuna b\u00f6l\u00fcp kare-k\u00f6k\u00fcn\u00fc alarak hesaplan\u0131r.<\/li><li><strong>Cramer&#8217;s <\/strong><strong>n<\/strong>, 2&#215;2 tablolarda hesaplanan phi (j) katsay\u0131s\u0131d\u0131r.<\/li><\/ul>\n<\/li><li><strong>Uncertainty coefficient <\/strong>\n<ul><li>Lambda\u2019ya benzer \u015fekilde ayn\u0131 \u00f6l\u00e7\u00fct\u00fc bildirir.<\/li><\/ul>\n<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><strong><em>Ordinal.<\/em><\/strong><\/h5>\n\n\n\n<p>Hem sat\u0131r hem de s\u00fctunda s\u0131ral\u0131 (ordinal) de\u011fi\u015fkenler varsa<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Gamma (<\/strong><strong>g<\/strong><strong>)<\/strong>\n<ul><li>Lambda\u2019n\u0131n ordinal de\u011fi\u015fkenler i\u00e7in olan versiyonudur<\/li><li>Her 2 de\u011fi\u015fken de ordinal olmal\u0131d\u0131r.<\/li><li>-1 ile +1 aras\u0131nda de\u011fi\u015fir. 1\u2019e (eksi ya da art\u0131) yak\u0131n olan de\u011ferler g\u00fc\u00e7l\u00fc ili\u015fki lehineyken 0\u2019a yak\u0131n de\u011ferler ili\u015fki yoklu\u011funu g\u00f6sterir.<\/li><li>Ard\u0131\u015f\u0131k giden de\u011fi\u015fkenler i\u00e7in, \u00f6rne\u011fin bir vaka de\u011fi\u015fkenlerin birinde ba\u015fka bir vakadan b\u00fcy\u00fckse, di\u011fer de\u011fi\u015fken i\u00e7in de di\u011fer vakadan b\u00fcy\u00fck olmal\u0131d\u0131r. B\u00f6yle vakalar konkordan (C), bu kurala uymayanlar diskordan (D) vakalard\u0131r. Form\u00fcl\u00fc \u015fu \u015fekildedir:<\/li><li>g = (C \u2013 D) \/ (C + D)<\/li><li>Bu form\u00fcle g\u00f6re t\u00fcm vakalar konkordan ise g tam +1 olur. Ordinal uyumun her iki de\u011fi\u015fkende m\u00fckemmel oldu\u011funu g\u00f6sterir.<\/li><li>Form\u00fcle g\u00f6re hesaplanan g de\u011feri, i\u015faretinden ba\u011f\u0131ms\u0131z olarak, PRE de\u011ferini g\u00f6sterir.<\/li><li>Lambda gibi baz\u0131 k\u0131s\u0131tl\u0131l\u0131klar\u0131 vard\u0131r. En \u00f6nemlisi de e\u015fitlik durumlar\u0131n\u0131 yok saymas\u0131d\u0131r. Bu e\u015fitlik durumunu da hesaba katan ise Kendall\u2019s tau katsay\u0131s\u0131d\u0131r.<\/li><\/ul>\n<\/li><li><strong>Kendall&#8217;s tau-b<\/strong>.\n<ul><li>Sat\u0131r ve s\u00fctun say\u0131s\u0131 e\u015fitse ge\u00e7erlidir.<\/li><li>Ordinal ya da s\u0131ral\u0131 (ranked) de\u011fi\u015fkenler i\u00e7in e\u015fit olma durumunu da hesaba katan non-parametrik bir ili\u015fki \u00f6l\u00e7\u00fct\u00fcd\u00fcr.<\/li><li>Hesaplanan katsay\u0131n\u0131n i\u015fareti (eksi ya da art\u0131) ili\u015fkinin de y\u00f6n\u00fcn\u00fc belirtir.<\/li><li>Katsay\u0131n\u0131n b\u00fcy\u00fckl\u00fc\u011f\u00fc ili\u015fkinin b\u00fcy\u00fckl\u00fc\u011f\u00fcn\u00fc g\u00f6sterir. -1 ile +1 aras\u0131nda de\u011fi\u015fir. Ancak tam -1 ya da +1 sadece tam kare \u015feklindeki tablolarda g\u00f6r\u00fcl\u00fcr.<\/li><\/ul>\n<\/li><li><strong>Kendall&#8217;s tau-c<\/strong>.\n<ul><li>Sat\u0131r ve s\u00fctun say\u0131s\u0131 e\u015fit de\u011filse kullan\u0131l\u0131r.<\/li><\/ul>\n<\/li><li><strong>Somers&#8217; d<\/strong>.\n<ul><li>Kendall\u2019s tau\u2019ya g\u00f6re daha nadir kullan\u0131l\u0131r.<\/li><li>2 ordinal de\u011fi\u015fken aras\u0131nda sat\u0131r kategorilerine g\u00f6re kolon kategorisini \u00f6ng\u00f6rme g\u00fcc\u00fcn\u00fc g\u00f6sterir.<\/li><li>-1 ile +1 aras\u0131nda de\u011fi\u015fir. 1\u2019e (eksi ya da art\u0131) yak\u0131n olan de\u011ferler g\u00fc\u00e7l\u00fc ili\u015fki lehineyken 0\u2019a yak\u0131n de\u011ferler ili\u015fki yoklu\u011funu g\u00f6sterir.<\/li><\/ul>\n<\/li><\/ul>\n\n\n\n<figure class=\"wp-block-table\"><div class=\"pcrstb-wrap\"><table class=\"\"><tbody><tr><td><strong>\u0130li\u015fkinin G\u00fcc\u00fc<\/strong><\/td><td><strong>Lambda,&nbsp;<\/strong><strong>Gamma,&nbsp;<\/strong><strong>Pearson\u2019s r<\/strong><\/td><\/tr><tr><td>Yok<\/td><td>0.00<\/td><\/tr><tr><td>Zay\u0131f<\/td><td>+&nbsp;0.01 \u2013 0.09<\/td><\/tr><tr><td>Orta<\/td><td>+&nbsp;0.10 \u2013 0.29<\/td><\/tr><tr><td>Y\u00fcksek<\/td><td>+&nbsp;0.30 \u2013 0.99<\/td><\/tr><tr><td>M\u00fckemmel<\/td><td>+&nbsp;1.00<\/td><\/tr><\/tbody><\/table><\/div><\/figure>\n\n\n\n<h5 class=\"wp-block-heading\"><strong><em>Nominal by Interval.<\/em><\/strong><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>De\u011fi\u015fkenlerden biri kategorik di\u011feri kantitatif ise <strong>Eta <\/strong>se\u00e7ilmelidir.<\/li><li>0 ile 1 aras\u0131nda de\u011fi\u015fir.<\/li><li>\u00d6zellikle ba\u011f\u0131ms\u0131z de\u011fi\u015fkenin az say\u0131da kategoriden olu\u015ftu\u011fu (cinsiyet: erkek \/ kad\u0131n), ba\u011f\u0131ml\u0131 de\u011fi\u015fkenin ise interval bir s\u00fcrekli de\u011fi\u015fken oldu\u011fu durumda (gelir miktar\u0131) Eta \u00e7ok uygun bir ili\u015fki g\u00fcc\u00fc g\u00f6stericisidir.<\/li><li>SPSS 2 farkl\u0131 Eta de\u011feri hesaplar: biri sat\u0131r de\u011fi\u015fkenini interval olarak kabul eder, di\u011feri de s\u00fctun de\u011fi\u015fkenini. Do\u011fru \u015fekilde kullan\u0131lmas\u0131 gereklidir.<\/li><\/ul>\n\n\n\n<p>Yukar\u0131da belirtilen ili\u015fki katsay\u0131lar\u0131 d\u0131\u015f\u0131nda bu ayarlar men\u00fcs\u00fcnden se\u00e7ilebilen birka\u00e7 test daha vard\u0131r:<\/p>\n\n\n\n<h5 class=\"wp-block-heading\"><strong>Kappa<\/strong><strong>.&nbsp;<\/strong><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>Ayn\u0131 objeyi numaraland\u0131ran\/skorlayan\/\u00f6l\u00e7en 2 de\u011ferlendirici aras\u0131ndaki uyumu (agreement) \u00f6l\u00e7er, <strong>Cohen&#8217;s kappa<\/strong> diye bilinir.<\/li><li>1 m\u00fckemmel uyum, 0 uyumsuzluk (\u015fans fakt\u00f6r\u00fcnden farkl\u0131 de\u011fil) anlam\u0131na gelir.<\/li><li>Hem sat\u0131r hem de s\u00fctun de\u011fi\u015fkeninin birebir ayn\u0131 isimde ve say\u0131da kategorilere sahip olmas\u0131 gereklidir.<\/li><li>Bir de\u011fi\u015fkende g\u00f6zlenen ama di\u011fer de\u011fi\u015fkende olmayan t\u00fcm de\u011ferlerin oldu\u011fu vakalar hesap d\u0131\u015f\u0131na \u00e7\u0131kar\u0131l\u0131r.<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><strong>Risk<\/strong><strong>.<\/strong><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>2 x 2 tablolarda bir fakt\u00f6r\u00fcn varl\u0131\u011f\u0131 ile (sat\u0131r) bir sonucun g\u00f6r\u00fclmesi (s\u00fctun) aras\u0131ndaki ili\u015fkiyi g\u00f6sterir.<\/li><li>Riskin g\u00fcven aral\u0131\u011f\u0131 1 de\u011ferini i\u00e7eriyorsa fakt\u00f6r ile sonu\u00e7 aras\u0131nda ili\u015fki var denilemez.<\/li><li>Fakt\u00f6r\u00fcn s\u0131kl\u0131\u011f\u0131 d\u00fc\u015f\u00fck ise risk yerine <strong>odds oran\u0131<\/strong> kullan\u0131labilir.<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><strong>McNemar<\/strong><strong>.&nbsp;<\/strong><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>2 x 2 tabloda 2 dikotom de\u011fi\u015fken aras\u0131ndaki ili\u015fkiyi non-parametrik \u015fekilde test eder.<\/li><li>\u00d6zellikle yan\u0131tlardaki de\u011fi\u015fim gibi ba\u011f\u0131ml\u0131 de\u011fi\u015fkenlerdeki fark\u0131 ve de\u011fi\u015fimi test etmek i\u00e7in kullan\u0131l\u0131r.<\/li><li>\u00d6ncesi-sonras\u0131 dizaynlarda \u00f6zellikle tercih edilir.<\/li><li>2&#215;2\u2019den b\u00fcy\u00fck tablolarda bu se\u00e7enek se\u00e7ilirse McNemar yerine <strong>McNemar-Bowker<\/strong> simetri testi <strong>(test of symmetry)<\/strong><\/li><li>Bu ayar men\u00fcs\u00fc d\u0131\u015f\u0131nda SPSS\u2019de 2 farkl\u0131 yerden daha McNemar testi hesaplanabilir<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><strong>Cochran&#8217;s and Mantel-Haenszel statistics<\/strong><strong>.&nbsp;<\/strong><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>E\u011fer iki de\u011fi\u015fken aras\u0131ndaki ili\u015fki ara\u015ft\u0131r\u0131l\u0131rken di\u011fer ba\u015fka de\u011fi\u015fkenler a\u00e7\u0131s\u0131ndan etkinin standardize edilmesi gerekiyorsa tabakalar (layers) halinde birden fazla ki-kare testinin ayn\u0131 anda yap\u0131lmas\u0131 esas\u0131na dayanan <strong>Cochran&#8217;s and Mantel-Haenszel statistics <\/strong>yap\u0131lmal\u0131d\u0131r.<\/li><li>\u00d6rne\u011fin, bir vaka-kontrol \u00e7al\u0131\u015fmas\u0131nda, sat\u0131rlar vaka ve kontrol gruplar\u0131, kolonlar da sonlan\u0131m (\u00f6ld\u00fc, ya\u015f\u0131yor) ise, yani her ikisi de kategorik de\u011fi\u015fken ise, ve sonucu etkileyebilece\u011fi d\u00fc\u015f\u00fcn\u00fclen di\u011fer fakt\u00f6rler a\u00e7\u0131s\u0131ndan standardizasyon yapmak istiyorsak (cinsiyet, meslek vb), bu istatisti\u011fi kullan\u0131r\u0131z.<\/li><\/ul>\n\n\n\n<h4 class=\"wp-block-heading\"><em><strong>Cells ayar men\u00fcs\u00fc<\/strong><\/em><\/h4>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"alignright\"><a href=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/7.png\" data-rel=\"penci-gallery-image-content\" ><img decoding=\"async\" src=\"http:\/\/www.acilci.net\/wp-content\/uploads\/2017\/05\/7.png\" alt=\"\" class=\"wp-image-22298\" \/><\/a><\/figure><\/div>\n\n\n\n<p>Bu k\u0131s\u0131mda tablonun i\u00e7indeki kutularda yer almas\u0131n\u0131 istedi\u011fimiz de\u011ferleri se\u00e7iyoruz.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\"><em><strong>Counts<\/strong><\/em><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>G\u00f6zlenen (observed) ve Beklenen (expected) de\u011ferleri i\u00e7in ilgili kutular i\u015faretlenir.<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><em><strong>Percentages<\/strong><\/em><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>Sat\u0131r (row) ve s\u00fctun (column) y\u00fczdeleri i\u00e7in ilgili kutular i\u015faretlenir.<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><em><strong>Residuals<\/strong><\/em><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li>G\u00f6zlenen ve beklenen de\u011ferler aras\u0131ndaki fark ve bunlarla ilgili hesaplar\u0131 i\u00e7erir<\/li><li><strong>Unstandardized<\/strong> = G\u00f6zlenen \u2013 Beklenen. Pozitif olmas\u0131, olmas\u0131n\u0131 bekledi\u011fimizden belirtilen say\u0131 kadar fazla vaka oldu\u011funu belirtir.<\/li><li><strong>Standardized<\/strong>. Ortalamalar\u0131 0 ve standart sapmas\u0131 1 olacak \u015fekilde yeniden hesaplan\u0131r. Kutular\u0131n sapmalar\u0131 birbirleri ile oransal olarak kar\u015f\u0131la\u015ft\u0131r\u0131labilir.<\/li><li><strong>Adjusted standardized<\/strong>. Ortalamalar\u0131 0 ve standart sapmas\u0131 1 olacak \u015fekilde yeniden hesaplanan rezid\u00fclerin ka\u00e7 standart sapma birimine denk geldi\u011fi belirtilir. Kutulardaki sapma miktar\u0131 farkl\u0131 tablolardaki verilerle kar\u015f\u0131la\u015ft\u0131r\u0131labilir. -2 ya da +2\u2019den b\u00fcy\u00fck kutular \u00f6zellikle anlaml\u0131 farkl\u0131l\u0131k i\u00e7eren kutulard\u0131r.<\/li><\/ul>\n\n\n\n<h5 class=\"wp-block-heading\"><em><strong>Z-Testi<\/strong><\/em><\/h5>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Compare columns ve Adjust p-values (bonferroni) <\/strong>se\u00e7enekleri se\u00e7ili oldu\u011fu anda her kolon kendi i\u00e7inde kar\u015f\u0131la\u015ft\u0131r\u0131l\u0131r. Birbirine benzer kutular ayn\u0131 harf ile i\u015faretlenir. B\u00f6ylece fark\u0131n kaynakland\u0131\u011f\u0131 kutular belirlenir. P de\u011feri de \u00e7oklu kar\u015f\u0131la\u015ft\u0131rma yap\u0131lmas\u0131 sebebiyle bonferroni d\u00fczeltmesi ile (anlaml\u0131l\u0131k e\u015fi\u011fi olan p de\u011ferinin kar\u015f\u0131la\u015ft\u0131rma say\u0131s\u0131na b\u00f6l\u00fcnmesi) d\u00fczeltilir.<\/li><\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u00c7ok&nbsp;g\u00f6zl\u00fc tablolar iki kalitatif&nbsp;verinin birbiriyle kar\u015f\u0131la\u015ft\u0131r\u0131lmas\u0131nda kulland\u0131\u011f\u0131m\u0131z kar\u015f\u0131la\u015ft\u0131rma metodudur. Kalitatif veriler s\u0131n\u0131fland\u0131rma&nbsp;belirtir. Bu tip veriyi tutan de\u011fi\u015fkenlere kategorik\/gruplu\/nominal de\u011fi\u015fkenler ad\u0131 verilir Bu&hellip;<\/p>\n","protected":false},"author":1561,"featured_media":493,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"inline_featured_image":false,"_lmt_disableupdate":"","_lmt_disable":"","footnotes":""},"categories":[21,10014],"tags":[],"class_list":["post-492","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-istatistik-yazilari","category-akademik-blog-yazisi"],"acf":[],"_links":{"self":[{"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/posts\/492","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/users\/1561"}],"replies":[{"embeddable":true,"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/comments?post=492"}],"version-history":[{"count":0,"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/posts\/492\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/media\/493"}],"wp:attachment":[{"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/media?parent=492"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/categories?post=492"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/tatd.org.tr\/atak\/wp-json\/wp\/v2\/tags?post=492"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}