{"id":644,"date":"2023-01-27T01:23:04","date_gmt":"2023-01-26T22:23:04","guid":{"rendered":"https:\/\/www.istatistik.gen.tr\/?p=644"},"modified":"2026-03-19T12:53:50","modified_gmt":"2026-03-19T09:53:50","slug":"acimlayici-faktor-analizinden-sonra-dogrulayici-faktor-analizi","status":"publish","type":"post","link":"https:\/\/www.istatistik.gen.tr\/?p=644","title":{"rendered":"A\u00e7\u0131mlay\u0131c\u0131\/Ke\u015ffedici fakt\u00f6r analizinden sonra do\u011frulay\u0131c\u0131 fakt\u00f6r analizi"},"content":{"rendered":"\n

A\u00e7\u0131mlay\u0131c\u0131\/Ked\u015ffedici fakt\u00f6r analizi ile do\u011frulay\u0131c\u0131 fakt\u00f6r analizi klasik kullan\u0131mlar\u0131nda ayn\u0131 sonucu vermezler. Ke\u015ffedici fakt\u00f6r analizi do\u011frulay\u0131c\u0131 analizle ayn\u0131 sonucu vermeyebilir. Bu yaz\u0131da bununla ilgili bir \u00f6rnek ve ne yap\u0131lmas\u0131 gerekti\u011fi anlat\u0131lm\u0131\u015ft\u0131r.<\/p>\n\n\n\n\n\n\n\n

Bu konudaki en iyi kaynaklardan birisi Brown\u2019\u0131n Do\u011frulay\u0131c\u0131 fakt\u00f6r analizi kitab\u0131d\u0131r. Brown, T. A. (2015). Confirmatory factor analysis for applied research (2nd ed.). The Guilford Press. <\/p>\n\n\n\n

Do\u011frulay\u0131c\u0131 fakt\u00f6r analizinden \u00f6nce ke\u015ffedici fakt\u00f6r analizi yapmak gerekli mi konusu g\u00fcn\u00fcm\u00fcz\u00fcn \u00f6nemli metodolojik problemlerinden birisi oldu. Brown\u2019a g\u00f6re ke\u015ffedici sonras\u0131 do\u011frulay\u0131c\u0131 fakt\u00f6r analizi genellikle yanl\u0131\u015f uygulan\u0131yor. \u015e\u00f6yle ki: Size ke\u015ffedici analiz yapt\u0131\u011f\u0131n\u0131zda maddeler fakt\u00f6rlere yerle\u015firken t\u00fcm fakt\u00f6rlerde y\u00fck alacak \u015fekilde da\u011f\u0131l\u0131yor. Oysa do\u011frulay\u0131c\u0131 analizde maddeler sanki sadece bir fakt\u00f6rde y\u00fck al\u0131yormu\u015f gibi tan\u0131ml\u0131yoruz. B\u00f6ylece ke\u015ffedici fakt\u00f6r analizinde elde etti\u011fimizden \u00e7ok daha k\u00f6t\u00fc bir modeli do\u011frulam\u0131\u015f oluyoruz. Bu nedenle ke\u015ffedici yapt\u0131ktan sonra do\u011frulay\u0131c\u0131 yapacaksak do\u011frulay\u0131c\u0131 analizde ke\u015ffediciden elde edilen fakt\u00f6r y\u00fcklerini kullanarak yapmal\u0131y\u0131z. Buna e\/cfa deniyor. Bu durumda da Brown diyor ki efa ile birebir ayn\u0131 sonu\u00e7lar\u0131 bulaca\u011f\u0131z. Ayn\u0131 fakt\u00f6r y\u00fckleri, ayn\u0131 chi-kare de\u011feri ayn\u0131 rmsea de\u011feri vs. Kitab\u0131n 194-202 aras\u0131 sayfalar\u0131nda \u00f6rnekle bunu g\u00f6rebilirsiniz. Peki neden do\u011frulay\u0131c\u0131 yapt\u0131k? Zaten birebir ayn\u0131s\u0131 \u00e7\u0131kmak zorunda olan bir analizi ni\u00e7in yapt\u0131k? Brown\u2019a g\u00f6re e\u011fer maddelerin birden \u00e7ok fakt\u00f6rde ald\u0131klar\u0131 y\u00fcklerin anlaml\u0131l\u0131klar\u0131n\u0131 merak etmiyorsan\u0131z efa sonras\u0131 cfa yapman\u0131n pek bir anlam\u0131 yok. Muhtemelen \u00fclkemizde klasikle\u015fmi\u015f \u201cbir madde sadece bir maddede y\u00fck almak zorundad\u0131r\u201d \u015feklinde \u00f6zetlenebilecek olan gereksiz k\u0131s\u0131tlaman\u0131n yayg\u0131nl\u0131\u011f\u0131 nedeniyle siz de bu efa sonras\u0131 cfa\u2019ya ihtiya\u00e7 duymayacaks\u0131n\u0131z. Bu nedenle Brown daha kitab\u0131n\u0131n 1. sayfas\u0131nda ve 49. sayfas\u0131nda fakt\u00f6r yap\u0131s\u0131n\u0131 \u00f6nceden biliyorsan\u0131z ya da fakt\u00f6r yap\u0131n\u0131z\u0131n ne oldu\u011funa dair g\u00fc\u00e7l\u00fc kan\u0131tlar\u0131n\u0131z varsa \u201cdolayl\u0131 olarak\u201d do\u011frulay\u0131c\u0131 yap\u0131n diye \u00f6neriyor. Siz de bu \u00f6neri do\u011frultusunda hem efa yap hem dfa yap diye \u0131srar eden hakemlere veya j\u00fcri \u00fcyelerine d\u00f6n\u00fc\u015f yapabilirsiniz.<\/p>\n\n\n\n

\u00d6rnek olmas\u0131 a\u00e7\u0131s\u0131ndan buraya t\u0131klayarak<\/a> \u00f6rnekteki datay\u0131 indirerek hem a\u00e7\u0131mlay\u0131c\u0131\/ke\u015ffedici hem de do\u011frulay\u0131c\u0131 fakt\u00f6r analizi uygulayabilirsiniz. \u0130ki sonucun birbirini tutmad\u0131\u011f\u0131n\u0131 g\u00f6receksiniz. Ke\u015ffedici fakt\u00f6r analizinde 2 fakt\u00f6r elde edeceksiniz. SPSS program\u0131nda maksimum likelihood y\u00f6ntemini kullanarak (do\u011frulay\u0131c\u0131 maksimum likelihood kullan\u0131yor o y\u00fczden) fakt\u00f6r analizi uygularsan\u0131z \u015fu sonu\u00e7lar\u0131 elde edersiniz:<\/p>\n\n\n\n

\n
\"\"<\/a><\/figure>\n<\/figure>\n\n\n\n

Ard\u0131ndan bu sonu\u00e7lar\u0131 yani do\u011frulatmak i\u00e7in diledi\u011finiz program\u0131 kullan\u0131n. AMOS, Mplus, Lisrel, SMART-PLS veya R ile elde edece\u011finiz do\u011frulay\u0131c\u0131 fakt\u00f6r analizi sonucunuz ise \u015f\u00f6yle olacakt\u0131r: <\/p>\n\n\n\n

Chi-Square de\u011feri: 65.396, sd = 13, RMSEA = 0.101, CFI = 0.942, TLI = 0.907<\/p>\n\n\n\n

G\u00f6r\u00fclece\u011fi gibi \u00f6zellikle RMSEA de\u011feri modelin do\u011frulanmas\u0131na izin vermemektedir. <\/p>\n\n\n\n

<\/p>\n\n\n\n

Yukar\u0131da ya\u015fanan durum \u015fudur: Ke\u015ffedici\/A\u00e7\u0131mlay\u0131c\u0131 fakt\u00f6r analizinden elde edilen sonu\u00e7lar do\u011frulay\u0131c\u0131 fakt\u00f6r analizinde do\u011frulanmad\u0131.<\/p>\n\n\n\n

Fakat bilmelisiniz ki do\u011frulay\u0131c\u0131 fakt\u00f6r analizinde do\u011frulanmaya \u00e7al\u0131\u015f\u0131lan model ke\u015ffedicide elde edilen sonu\u00e7 de\u011fildi. Yani yukar\u0131daki iki analiz asl\u0131nda birbirini do\u011frulam\u0131yor. \u0130ki analiz birbiriyle ayn\u0131 modeli kullanm\u0131yor.<\/p>\n\n\n\n

Bunun en \u00f6nemli sebebi, ke\u015ffedici fakt\u00f6r analizinde elde edilen fakt\u00f6r yap\u0131s\u0131n\u0131n asl\u0131nda o kadar da saf bir ayr\u0131\u015f\u0131kl\u0131\u011fa sahip olmamas\u0131d\u0131r. SPSS gibi programlar fakt\u00f6r analizi yaparken maddeleri t\u00fcm fakt\u00f6rlere az ya da \u00e7ok yerle\u015ftirir. Yani ke\u015ffedici fakt\u00f6r analizi yapt\u0131\u011f\u0131n\u0131zda her madde az ya da \u00e7ok bir fakt\u00f6rde yer al\u0131r. Yukar\u0131daki resimde verilen fakt\u00f6r analizi sonucu analiz tekrar edilip de t\u00fcm fakt\u00f6r y\u00fcklerini g\u00f6ster se\u00e7ene\u011fi aktive edilerek yap\u0131l\u0131rsa \u015f\u00f6yle g\u00f6r\u00fcn\u00fcr:<\/p>\n\n\n\n

\n
\"\"<\/a><\/figure>\n<\/figure>\n\n\n\n

Madde16 asl\u0131nda iki fakt\u00f6rde de y\u00fck almaktad\u0131r. \u0130kinci fakt\u00f6rde -0.191 gibi k\u00fc\u00e7\u00fck bir y\u00fck al\u0131yor ama yine de o fakt\u00f6rde de \u00e7al\u0131\u015fmaktad\u0131r. Siz bunlar\u0131 tamamen g\u00f6zard\u0131 ederek do\u011frulay\u0131c\u0131 fakt\u00f6r analizine ge\u00e7iyorsunuz ve o madde sanki tek fakt\u00f6rde \u00e7al\u0131\u015f\u0131yormu\u015f gibi birinci fakt\u00f6rde m2,m9,m16 vard\u0131r, ikinci fakt\u00f6rde m8,m15,m22 ve m23 vard\u0131r diye d\u00fc\u015f\u00fcnerek a\u015f\u0131r\u0131 berrakla\u015ft\u0131rma ile do\u011frulamaya \u00e7al\u0131\u015fmaktas\u0131n\u0131z. Dolay\u0131s\u0131 ile asl\u0131nda ke\u015ffediciden elde etti\u011finiz fakt\u00f6r yap\u0131s\u0131n\u0131 de\u011fil a\u015f\u0131r\u0131 berrakla\u015ft\u0131r\u0131lm\u0131\u015f ve ke\u015ffedici ile tam olarak uyu\u015fmayan bir modeli do\u011frulamaya ge\u00e7iyorsunuz.<\/p>\n\n\n\n

Bu a\u015f\u0131r\u0131 safla\u015ft\u0131rma gelene\u011fi eski bir gelenektir ve hesaplamalar\u0131n zor oldu\u011fu ge\u00e7mi\u015f y\u0131llara aittir. Bug\u00fcn hesaplama kolayl\u0131klar\u0131 sayesinde bir maddenin birden \u00e7ok fakt\u00f6rde y\u00fck almas\u0131na izin verebiliriz. Bir maddenin birden \u00e7ok fakt\u00f6r almas\u0131 demek bir maddenin birden \u00e7ok fakt\u00f6rde \u00f6l\u00e7me yapabilmesine izin vermek demektir. T\u0131pk\u0131 bir matematik sorusunun hem geometri bilgisini hem de cebir bilgisini ayn\u0131 soruda \u00f6l\u00e7mesi gibi d\u00fc\u015f\u00fcnebilirsiniz. Ger\u00e7ek hayatta maddelerimiz her fakt\u00f6rde y\u00fck al\u0131rken do\u011frulay\u0131c\u0131 analizde sanki b\u00f6yle bir durum yokmu\u015f gibi davranmak ger\u00e7ekli\u011fe ayk\u0131r\u0131 olacakt\u0131r. Hatta \u00e7ok daha k\u00f6t\u00fc\/alakas\u0131z bir modeli do\u011frulamaya \u00e7al\u0131\u015fmak olacakt\u0131r. <\/p>\n\n\n\n

Peki siz ke\u015ffediciden elde etti\u011finiz \u015fu maddelerin her fakt\u00f6rde az ya da \u00e7ok y\u00fck ald\u0131\u011f\u0131 “ger\u00e7ek” modeli do\u011frulamak istiyorsunuz diyelim. Ke\u015ffedici\/A\u00e7\u0131mlay\u0131c\u0131 fakt\u00f6r analizinden elde edilen birden \u00e7ok fakt\u00f6rde y\u00fck alma durumlar\u0131n\u0131 da dikkate alan do\u011frulama i\u015flemi e\/cfa olarak da bilinir.<\/p>\n\n\n\n

e\/cfa bir ka\u00e7 ad\u0131mda ger\u00e7ekle\u015fir:<\/p>\n\n\n\n

    \n
  1. (M\u00fcmk\u00fcnse) maximum likelihood kestirim y\u00f6ntemiyle uygulanm\u0131\u015f bir ke\u015ffedici fakt\u00f6r analizi sonucunuz olsun.<\/li>\n\n\n\n
  2. Her fakt\u00f6r i\u00e7in \u00e7apa madde belirleyin: \u00c7apa madde bir fakt\u00f6rde y\u00fcksek y\u00fck alm\u0131\u015fken di\u011fer fakt\u00f6rde s\u0131f\u0131ra en yak\u0131n y\u00fck alm\u0131\u015f madde demektir. \u00d6rne\u011fin yukar\u0131daki SPSS \u00e7\u0131kt\u0131s\u0131 incelenirse birinci fakt\u00f6rde m22, ikinci fakt\u00f6rde m23 maddeleri \u00e7apa madde olarak ele al\u0131nabilir.<\/li>\n\n\n\n
  3. Bu maddelerin d\u00fc\u015f\u00fck ilgili oldu\u011fu fakt\u00f6rdeki fakt\u00f6r y\u00fcklerini s\u0131f\u0131ra sabitleyin<\/li>\n\n\n\n
  4. Maddelerin t\u00fcm\u00fcn\u00fcn t\u00fcm fakt\u00f6rlerde yer ald\u0131\u011f\u0131ndan emin olun<\/li>\n\n\n\n
  5. fakt\u00f6rlerin fakt\u00f6r varyanslar\u0131n\u0131 1’e e\u015fitleyin<\/li>\n\n\n\n
  6. Elde edece\u011finiz sonucun t\u00fcm de\u011fi\u015fkenler i\u00e7in standardize edilmi\u015f sonu\u00e7 olmas\u0131n\u0131 sa\u011flay\u0131n. <\/li>\n<\/ol>\n\n\n\n

    R program\u0131nda lavaan i\u00e7in kullanabilece\u011finiz model kodu \u015f\u00f6yledir:<\/p>\n\n\n\n

    model <- 'factor1 =~ m8+m15+m22+m2+m9+m16+0*m23\n          factor2 =~ m8+m15+0*m22+m2+m9+m16+m23\n          factor1 ~~ 1*factor1\n          factor2 ~~ 1*factor2'\n\necfa <- cfa(model, data, std.lv=T, std.ov=T)\nsummary(ecfa, fit.measures=TRUE)\n<\/code><\/pre>\n\n\n\n
    Analizi uygulad\u0131\u011f\u0131n\u0131zda elde edece\u011finiz sonu\u00e7lar \u015f\u00f6yledir: <\/p>\n\n\n\n
    Chi Square de\u011feri : 26.997, sd = 8, RMSEA= 0.077, CFI = 979, TLI = 0.945<\/p>\n\n\n\n
    G\u00f6r\u00fclece\u011fi gibi model do\u011fruland\u0131 ve Ke\u015ffedici\/A\u00e7\u0131mlay\u0131c\u0131 fakt\u00f6r analizi ile elde edilen chi square de\u011feri (\u00e7ok yak\u0131n olacak \u015fekilde) elde edildi. <\/p>\n\n\n\n
    \u0130\u015fte ancak bu t\u00fcr bir \u00e7al\u0131\u015fma\/analiz ile ke\u015ffedici\/a\u00e7\u0131mlay\u0131c\u0131 fakt\u00f6r analizinden elde edilen fakt\u00f6r sonucunu do\u011frulam\u0131\u015f olursunuz. Yoksa ke\u015ffediciden elde edilenden daha k\u00f6t\u00fc ve gereksiz yere berrakla\u015ft\u0131r\u0131lm\u0131\u015f bir modeli do\u011frulamaya \u00e7abalam\u0131\u015f olursunuz. Do\u011frulayabilirsiniz bile, ama do\u011frulad\u0131\u011f\u0131n\u0131z \u015fey ke\u015ffediciden elde etti\u011finiz bilgi olmayacakt\u0131r. <\/p>\n\n\n\n
    e\/cfa y\u00f6ntemin \u00f6nemli avantaj\u0131 size fakt\u00f6r y\u00fcklerinizin anlaml\u0131l\u0131\u011f\u0131n\u0131 da verebiliyor olmas\u0131d\u0131r. Normalde SPSS gibi programlar ke\u015ffedici\/a\u00e7\u0131mlay\u0131c\u0131 fakt\u00f6r analizi yapt\u0131\u011f\u0131n\u0131zda sadece fakt\u00f6r y\u00fcklerini verirler. Bu fakt\u00f6r y\u00fckleri istatistiksel olarak anlaml\u0131 olmayabilir. e\/cfa sayesinde fakt\u00f6r y\u00fcklerinizin anlaml\u0131l\u0131\u011f\u0131n\u0131 da g\u00f6rebilirsiniz.  A\u015fa\u011f\u0131da R program\u0131nda lavaan \u00e7\u0131kt\u0131s\u0131n\u0131 g\u00f6rebilirsiniz. <\/p>\n\n\n\n
    Latent Variables:\n                   Estimate  Std.Err  z-value  P(>|z|)\n  factor1 =~                                          \n    m8                0.579    0.054   10.767    0.000\n    m15               0.688    0.054   12.773    0.000\n    m22               0.774    0.053   14.550    0.000\n    m2               -0.126    0.066   -1.891    0.059\n    m9               -0.137    0.066   -2.076    0.038\n    m16              -0.154    0.068   -2.273    0.023\n    m23               0.000                           \n  factor2 =~                                          \n    m8                0.269    0.050    5.423    0.000\n    m15               0.225    0.049    4.630    0.000\n    m22               0.000                           \n    m2                0.769    0.047   16.420    0.000\n    m9                0.751    0.047   15.936    0.000\n    m16               0.814    0.046   17.632    0.000\n    m23               0.628    0.049   12.917    0.000\n<\/code><\/pre>\n\n\n\n
    Ke\u015ffedici\/A\u00e7\u0131mlay\u0131c\u0131 fakt\u00f6r analizinde elde etti\u011finiz fakt\u00f6r y\u00fcklerinin birebir ayn\u0131s\u0131n\u0131 do\u011frulay\u0131c\u0131 fakt\u00f6r analizinde g\u00f6remeyebilirsiniz fakat iki analiz de ayn\u0131 kestirim y\u00f6ntemleriyle elde edilirse fakt\u00f6r y\u00fcklerinin birbirine yakla\u015fmas\u0131 beklenir. Fakt\u00f6r y\u00fcklerinin birebir ayn\u0131 \u00e7\u0131kmas\u0131 bekleniyorsa bu durumda lavaan paketi fakt\u00f6r y\u00fcklerinin efa’dan elde edilenlerle tan\u0131mlanmas\u0131na imkan tan\u0131maktad\u0131r. <\/p>\n","protected":false},"excerpt":{"rendered":"
    A\u00e7\u0131mlay\u0131c\u0131\/Ked\u015ffedici fakt\u00f6r analizi ile do\u011frulay\u0131c\u0131 fakt\u00f6r analizi klasik kullan\u0131mlar\u0131nda ayn\u0131 sonucu vermezler. Ke\u015ffedici fakt\u00f6r analizi do\u011frulay\u0131c\u0131 analizle ayn\u0131 sonucu vermeyebilir. Bu yaz\u0131da bununla ilgili bir \u00f6rnek ve ne yap\u0131lmas\u0131 gerekti\u011fi anlat\u0131lm\u0131\u015ft\u0131r.<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":true,"template":"","format":"standard","meta":{"_kad_post_transparent":"","_kad_post_title":"","_kad_post_layout":"","_kad_post_sidebar_id":"","_kad_post_content_style":"","_kad_post_vertical_padding":"","_kad_post_feature":"","_kad_post_feature_position":"","_kad_post_header":false,"_kad_post_footer":false,"_kad_post_classname":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-644","post","type-post","status-publish","format-standard","hentry","category-genel"],"_links":{"self":[{"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=\/wp\/v2\/posts\/644","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=644"}],"version-history":[{"count":21,"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=\/wp\/v2\/posts\/644\/revisions"}],"predecessor-version":[{"id":1255,"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=\/wp\/v2\/posts\/644\/revisions\/1255"}],"wp:attachment":[{"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=644"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=644"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.istatistik.gen.tr\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=644"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}