<\/p>\n
Hipotez testi<\/a> yapt\u0131\u011f\u0131m\u0131zda genellikle bo\u015f hipotezi reddedip reddetmememiz gerekti\u011fine karar vermek i\u00e7in bir alfa d\u00fczeyiyle kar\u015f\u0131la\u015ft\u0131rd\u0131\u011f\u0131m\u0131z bir p de\u011feri elde ederiz.<\/span><\/p>\n \u00d6rne\u011fin, iki pop\u00fclasyonun ortalamalar\u0131n\u0131n e\u015fit olup olmad\u0131\u011f\u0131n\u0131 belirlemek i\u00e7in 0,05 alfa d\u00fczeyini kullanarak iki \u00f6rnekli bir t testi<\/a> yapabiliriz. Testi \u00e7al\u0131\u015ft\u0131rd\u0131\u011f\u0131m\u0131z\u0131 ve 0,0023 p de\u011ferini elde etti\u011fimizi varsayal\u0131m. Bu durumda, p de\u011feri se\u00e7ilen alfa seviyesinden k\u00fc\u00e7\u00fck oldu\u011fundan, iki pop\u00fclasyonun ortalamalar\u0131n\u0131n e\u015fit oldu\u011fu y\u00f6n\u00fcndeki s\u0131f\u0131r hipotezini reddederiz.<\/span><\/p>\n P de\u011ferleri, belirli hipotezleri reddetmek veya reddetmek i\u00e7in yayg\u0131n olarak kullan\u0131lan bir \u00f6l\u00e7\u00fcd\u00fcr, ancak kullan\u0131labilecek ba\u015fka bir \u00f6l\u00e7\u00fc de vard\u0131r: Bayes fakt\u00f6r\u00fc<\/strong> .<\/span><\/p>\n Bayes fakt\u00f6r\u00fc, belirli bir hipotezin olas\u0131l\u0131\u011f\u0131n\u0131n ba\u015fka bir hipotezin olas\u0131l\u0131\u011f\u0131na oran\u0131 olarak tan\u0131mlan\u0131r. Genel olarak, alternatif bir hipotezin olas\u0131l\u0131\u011f\u0131 ile bo\u015f bir hipotezin olas\u0131l\u0131\u011f\u0131 aras\u0131ndaki oran\u0131 bulmak i\u00e7in kullan\u0131l\u0131r:<\/span><\/p>\n Bayes fakt\u00f6r\u00fc =<\/strong> veri sa\u011flanma olas\u0131l\u0131\u011f\u0131 H A<\/sub> \/ veri sa\u011flanma olas\u0131l\u0131\u011f\u0131 H 0<\/sub><\/span><\/p>\n<\/blockquote>\n \u00d6rne\u011fin Bayes fakt\u00f6r\u00fcn\u00fcn 5 olmas\u0131, veriler g\u00f6z \u00f6n\u00fcne al\u0131nd\u0131\u011f\u0131nda alternatif hipotezin s\u0131f\u0131r hipotezinden 5 kat daha muhtemel oldu\u011fu anlam\u0131na gelir.<\/span><\/p>\n Tersine, Bayes fakt\u00f6r\u00fc 1\/5 ise, s\u0131f\u0131r hipotezinin veri verilen alternatif hipotezden 5 kat daha muhtemel oldu\u011fu anlam\u0131na gelir.<\/span><\/p>\n P de\u011ferlerine benzer \u015fekilde, s\u0131f\u0131r hipotezinin ne zaman reddedilece\u011fine karar vermek i\u00e7in e\u015fikleri kullanabiliriz. \u00d6rne\u011fin, 10 veya daha fazla bir Bayes fakt\u00f6r\u00fcn\u00fcn s\u0131f\u0131r hipotezini reddetmeye yetecek kadar g\u00fc\u00e7l\u00fc kan\u0131t olu\u015fturdu\u011funa karar verebiliriz.<\/span><\/p>\n Lee ve Wagenmaker, 2015 tarihli bir makalede<\/a> Bayes fakt\u00f6r\u00fcne ili\u015fkin a\u015fa\u011f\u0131daki yorumlar\u0131 \u00f6nerdiler:<\/span><\/p>\n Bayes fakt\u00f6r\u00fc ve p de\u011ferleri farkl\u0131 yorumlara sahiptir.<\/span><\/p>\n P de\u011ferleri:<\/strong><\/span><\/p>\n Bir p de\u011feri, s\u0131f\u0131r hipotezinin do\u011fru oldu\u011fu varsay\u0131larak, bir hipotez testinin g\u00f6zlemlenen sonu\u00e7lar\u0131 kadar u\u00e7 sonu\u00e7lar elde etme olas\u0131l\u0131\u011f\u0131 olarak yorumlan\u0131r.<\/span><\/p>\n \u00d6rne\u011fin, iki pop\u00fclasyonun ortalamalar\u0131n\u0131n e\u015fit olup olmad\u0131\u011f\u0131n\u0131 belirlemek i\u00e7in iki \u00f6rnekli bir t testi yapt\u0131\u011f\u0131n\u0131z\u0131 varsayal\u0131m. Test 0,0023’l\u00fck bir p de\u011feriyle sonu\u00e7lan\u0131rsa, bu, iki pop\u00fclasyonun ortalamalar\u0131n\u0131n ger\u00e7ekten e\u015fit olmas\u0131 durumunda bu sonucu elde etme olas\u0131l\u0131\u011f\u0131n\u0131n yaln\u0131zca 0,0023<\/strong> oldu\u011fu anlam\u0131na gelir. Bu de\u011fer \u00e7ok k\u00fc\u00e7\u00fck oldu\u011fu i\u00e7in s\u0131f\u0131r hipotezini reddediyoruz ve iki pop\u00fclasyonun ortalamalar\u0131n\u0131n e\u015fit olmad\u0131\u011f\u0131n\u0131 s\u00f6yleyecek yeterli kan\u0131t\u0131m\u0131z oldu\u011fu sonucuna var\u0131yoruz.<\/span><\/p>\n Bayes fakt\u00f6r\u00fc:<\/strong><\/span><\/p>\n Bayes fakt\u00f6r\u00fc, alternatif hipotez alt\u0131nda g\u00f6zlemlenen verilerin olu\u015fma olas\u0131l\u0131\u011f\u0131n\u0131n, s\u0131f\u0131r hipotezi alt\u0131nda g\u00f6zlemlenen verilerin olu\u015fma olas\u0131l\u0131\u011f\u0131na oran\u0131 olarak yorumlan\u0131r.<\/span><\/p>\n \u00d6rne\u011fin, bir hipotez testi yapt\u0131\u011f\u0131n\u0131z\u0131 ve Bayes fakt\u00f6r\u00fcn\u00fcn 4 oldu\u011funu varsayal\u0131m. Bu, ger\u00e7ekte g\u00f6zlemledi\u011finiz veriler g\u00f6z \u00f6n\u00fcne al\u0131nd\u0131\u011f\u0131nda, alternatif hipotezin s\u0131f\u0131r hipotezinden 4 kat daha muhtemel oldu\u011fu anlam\u0131na gelir.<\/span><\/p>\n Baz\u0131 istatistik\u00e7iler Bayes fakt\u00f6r\u00fcn\u00fcn p de\u011ferlerine g\u00f6re bir avantaj sundu\u011funa inan\u0131yor \u00e7\u00fcnk\u00fc iki rakip hipotezin lehine ve aleyhine kan\u0131tlar\u0131n \u00f6l\u00e7\u00fclmesine yard\u0131mc\u0131 oluyor. \u00d6rne\u011fin, bir s\u0131f\u0131r hipotezinin lehine veya aleyhine kan\u0131t \u00f6l\u00e7\u00fclebilir ve bu, bir p de\u011feri kullan\u0131larak yap\u0131lamaz.<\/span><\/p>\n Hangi yakla\u015f\u0131m\u0131 kullan\u0131rsan\u0131z kullan\u0131n (Bayes fakt\u00f6r\u00fc veya p de\u011ferleri), s\u0131f\u0131r hipotezini reddetmek isteyip istemedi\u011finize dair bir e\u015fik de\u011ferine karar vermeniz gerekir.<\/span><\/p>\n \u00d6rne\u011fin, yukar\u0131daki tabloda 9’luk bir Bayes Fakt\u00f6r\u00fcn\u00fcn “alternatif hipotez i\u00e7in orta derecede kan\u0131t” olarak s\u0131n\u0131fland\u0131r\u0131laca\u011f\u0131n\u0131, 10’luk bir Bayes Fakt\u00f6r\u00fcn\u00fcn ise “alternatif hipotez i\u00e7in g\u00fc\u00e7l\u00fc kan\u0131t” olarak s\u0131n\u0131fland\u0131r\u0131laca\u011f\u0131n\u0131 g\u00f6rd\u00fck.<\/span><\/p>\n Bu anlamda Bayes fakt\u00f6r\u00fc de ayn\u0131 sorundan muzdariptir: 0,06’l\u0131k bir p de\u011feri “anlaml\u0131 de\u011fil” olarak kabul edilirken, 0,05’lik bir p de\u011feri anlaml\u0131 kabul edilebilir.<\/span><\/p>\n Daha fazla okuma:<\/strong><\/span><\/p>\n P de\u011ferlerinin ve istatistiksel anlaml\u0131l\u0131\u011f\u0131n a\u00e7\u0131klamas\u0131 Hipotez testi yapt\u0131\u011f\u0131m\u0131zda genellikle bo\u015f hipotezi reddedip reddetmememiz gerekti\u011fine karar vermek i\u00e7in bir alfa d\u00fczeyiyle kar\u015f\u0131la\u015ft\u0131rd\u0131\u011f\u0131m\u0131z bir p de\u011feri elde ederiz. \u00d6rne\u011fin, iki pop\u00fclasyonun ortalamalar\u0131n\u0131n e\u015fit olup olmad\u0131\u011f\u0131n\u0131 belirlemek i\u00e7in 0,05 alfa d\u00fczeyini kullanarak iki \u00f6rnekli bir t testi yapabiliriz. Testi \u00e7al\u0131\u015ft\u0131rd\u0131\u011f\u0131m\u0131z\u0131 ve 0,0023 p de\u011ferini elde etti\u011fimizi varsayal\u0131m. Bu durumda, p de\u011feri se\u00e7ilen alfa […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[],"class_list":["post-594","post","type-post","status-publish","format-standard","hentry","category-rehber"],"yoast_head":"\n\n
\n\n
\n Bayes fakt\u00f6r\u00fc<\/span><\/strong><\/th>\n Terc\u00fcme<\/strong><\/span><\/th>\n<\/tr>\n \n > 100<\/span><\/td>\n Alternatif bir hipotez i\u00e7in a\u015f\u0131r\u0131 kan\u0131t<\/span><\/td>\n<\/tr>\n \n 30 \u2013 100<\/span><\/td>\n Alternatif bir hipotez i\u00e7in \u00e7ok g\u00fc\u00e7l\u00fc kan\u0131tlar<\/span><\/td>\n<\/tr>\n \n 10 \u2013 30<\/span><\/td>\n Alternatif bir hipotez i\u00e7in g\u00fc\u00e7l\u00fc kan\u0131tlar<\/span><\/td>\n<\/tr>\n \n 3 \u2013 10<\/span><\/td>\n Alternatif bir hipotez i\u00e7in orta d\u00fczeyde kan\u0131t<\/span><\/td>\n<\/tr>\n \n 1 \u2013 3<\/span><\/td>\n Alternatif bir hipotez i\u00e7in anekdot niteli\u011finde kan\u0131tlar<\/span><\/td>\n<\/tr>\n \n 1<\/span><\/td>\n Kan\u0131t yok<\/span><\/td>\n<\/tr>\n \n 1\/3 \u2013 1<\/span><\/td>\n Bo\u015f hipotez i\u00e7in anekdot niteli\u011finde kan\u0131tlar<\/span><\/td>\n<\/tr>\n \n 1\/3 \u2013 1\/10<\/span><\/td>\n S\u0131f\u0131r hipotezi i\u00e7in orta d\u00fczeyde kan\u0131t<\/span><\/td>\n<\/tr>\n \n 1\/10 \u2013 1\/30<\/span><\/td>\n Bo\u015f hipotez i\u00e7in g\u00fc\u00e7l\u00fc kan\u0131t<\/span><\/td>\n<\/tr>\n \n 1\/30 \u2013 1\/100<\/span><\/td>\n Bo\u015f hipotez i\u00e7in \u00e7ok g\u00fc\u00e7l\u00fc kan\u0131t<\/span><\/td>\n<\/tr>\n \n <1\/100<\/span><\/td>\n Bo\u015f hipotez i\u00e7in a\u015f\u0131r\u0131 kan\u0131t<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Bayes fakt\u00f6rleri ve P de\u011ferleri<\/strong><\/span><\/h2>\n
\u00c7\u00f6z\u00fcm<\/span><\/strong><\/h2>\n
\u0130statistiksel ve Pratik \u00d6neme \u0130li\u015fkin Basit Bir A\u00e7\u0131klama<\/p>\n","protected":false},"excerpt":{"rendered":"