Bu varsay\u0131m kar\u015f\u0131lanmad\u0131\u011f\u0131nda regresyon modelinde heteroskedastisitenin<\/a> mevcut oldu\u011fu s\u00f6ylenir.<\/span><\/p>\n Bu ger\u00e7ekle\u015fti\u011finde modelin regresyon katsay\u0131lar\u0131n\u0131n standart hatalar\u0131 g\u00fcvenilmez hale gelir.<\/span><\/p>\n Bunu hesaba katmak i\u00e7in, heteroskedastisiteye kar\u015f\u0131 “sa\u011flam” olan ve bize regresyon katsay\u0131lar\u0131 i\u00e7in ger\u00e7ek standart hata de\u011ferleri hakk\u0131nda daha iyi bir fikir verebilecek olan sa\u011flam standart hatalar\u0131<\/strong> hesaplayabiliriz.<\/span><\/p>\n A\u015fa\u011f\u0131daki \u00f6rnek, R’deki bir regresyon modeli i\u00e7in sa\u011flam standart hatalar\u0131n nas\u0131l hesaplanaca\u011f\u0131n\u0131 g\u00f6sterir.<\/span><\/p>\n R’de, bir s\u0131n\u0131ftaki 20 \u00f6\u011frencinin kazand\u0131\u011f\u0131 s\u0131nav puanlar\u0131 ve \u00e7al\u0131\u015f\u0131lan saatler hakk\u0131nda bilgi i\u00e7eren a\u015fa\u011f\u0131daki veri \u00e7er\u00e7evesine sahip oldu\u011fumuzu varsayal\u0131m:<\/span><\/p>\n R’de tahmin de\u011fi\u015fkeni olarak saatleri<\/strong> ve yan\u0131t de\u011fi\u015fkeni olarak puan\u0131<\/strong> kullanan bir regresyon modeline uyum sa\u011flamak i\u00e7in lm()<\/a> i\u015flevini kullanabiliriz:<\/span><\/p>\n Regresyon modelinde heteroskedastisitenin bir sorun olup olmad\u0131\u011f\u0131n\u0131 g\u00f6rsel olarak kontrol etmenin en kolay yolu bir art\u0131k grafi\u011fi olu\u015fturmakt\u0131r:<\/span> <\/p>\n X ekseni yan\u0131t de\u011fi\u015fkeninin uygun de\u011ferlerini, y ekseni ise kar\u015f\u0131l\u0131k gelen art\u0131klar\u0131 g\u00f6sterir.<\/span><\/p>\n Grafikten, uyum de\u011ferleri artt\u0131k\u00e7a art\u0131klar\u0131n varyans\u0131n\u0131n da artt\u0131\u011f\u0131n\u0131 g\u00f6rebiliriz.<\/span><\/p>\n Bu, de\u011fi\u015fen varyans\u0131n regresyon modelinde muhtemelen bir sorun oldu\u011funu ve model \u00f6zetinin standart hatalar\u0131n\u0131n g\u00fcvenilmez oldu\u011funu g\u00f6sterir.<\/span><\/p>\n Sa\u011flam standart hatalar\u0131 hesaplamak i\u00e7in lmtest<\/strong> paketindeki coeftest()<\/strong> fonksiyonunu ve sandvi\u00e7<\/strong> paketindeki vcovHC()<\/strong> fonksiyonunu a\u015fa\u011f\u0131daki gibi kullanabiliriz:<\/span><\/p>\n Saat<\/strong> tahmin de\u011fi\u015fkenine ili\u015fkin standart hatan\u0131n \u00f6nceki model \u00f6zetinde 1,075’ten bu model \u00f6zetinde 1,2072’ye y\u00fckseldi\u011fini unutmay\u0131n.<\/span><\/p>\n Orijinal regresyon modelinde de\u011fi\u015fen varyans mevcut oldu\u011fundan, bu standart hata tahmini daha g\u00fcvenilirdir ve saat<\/strong> tahmin de\u011fi\u015fkeni i\u00e7in bir g\u00fcven aral\u0131\u011f\u0131 hesaplan\u0131rken kullan\u0131lmal\u0131d\u0131r.<\/span><\/p>\n Not<\/strong> : vcovHC()<\/strong> i\u015flevinde hesaplanacak en yayg\u0131n tahmin t\u00fcr\u00fc ‘HC0’d\u0131r, ancak di\u011fer tahmin t\u00fcrlerini bulmak i\u00e7in belgelere<\/a> ba\u015fvurabilirsiniz.<\/span><\/p>\n A\u015fa\u011f\u0131daki e\u011fitimlerde R’de di\u011fer ortak g\u00f6revlerin nas\u0131l ger\u00e7ekle\u015ftirilece\u011fi a\u00e7\u0131klanmaktad\u0131r:<\/span><\/p>\n R’de de\u011fi\u015fen varyans i\u00e7in White testi nas\u0131l yap\u0131l\u0131r?<\/a> Do\u011frusal regresyonun varsay\u0131mlar\u0131ndan biri, model art\u0131klar\u0131n\u0131n yorday\u0131c\u0131 de\u011fi\u015fkenin her seviyesinde e\u015fit \u015fekilde da\u011f\u0131lm\u0131\u015f olmas\u0131d\u0131r. Bu varsay\u0131m kar\u015f\u0131lanmad\u0131\u011f\u0131nda regresyon modelinde heteroskedastisitenin mevcut oldu\u011fu s\u00f6ylenir. Bu ger\u00e7ekle\u015fti\u011finde modelin regresyon katsay\u0131lar\u0131n\u0131n standart hatalar\u0131 g\u00fcvenilmez hale gelir. Bunu hesaba katmak i\u00e7in, heteroskedastisiteye kar\u015f\u0131 “sa\u011flam” olan ve bize regresyon katsay\u0131lar\u0131 i\u00e7in ger\u00e7ek standart hata de\u011ferleri hakk\u0131nda daha iyi bir fikir […]<\/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-3909","post","type-post","status-publish","format-standard","hentry","category-rehber"],"yoast_head":"\n \u00d6rnek: R’de sa\u011flam standart hatalar\u0131n hesaplanmas\u0131<\/strong><\/h2>\n
#create data frame\n<\/span>df <- data. frame<\/span> (hours=c(1, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4,\n 4, 5, 5, 5, 6, 6, 7, 7, 8),\n score=c(67, 68, 74, 70, 71, 75, 80, 70, 84, 72,\n 88, 75, 95, 75, 99, 78, 99, 65, 96, 70))\n\n#view head of data frame\n<\/span>head(df)\n\n hours score\n1 1 67\n2 1 68\n3 1 74\n4 1 70\n5 2 71\n6 2 75\n<\/strong><\/pre>\n #fit regression model\n<\/span>fit <- lm(score ~ hours, data=df)\n\n#view summary of model\n<\/span>summary(fit)\n\nCall:\nlm(formula = score ~ hours, data = df)\n\nResiduals:\n Min 1Q Median 3Q Max \n-19,775 -5,298 -3,521 7,520 18,116 \n\nCoefficients:\n Estimate Std. Error t value Pr(>|t|) \n(Intercept) 71.158 4.708 15.11 1.14e-11 ***\nhours 1.945 1.075 1.81 0.087 . \n---\nSignificant. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1\n\nResidual standard error: 10.48 on 18 degrees of freedom\nMultiple R-squared: 0.154, Adjusted R-squared: 0.107 \nF-statistic: 3.278 on 1 and 18 DF, p-value: 0.08696<\/strong><\/pre>\n #create residual vs. fitted plot\n<\/span>plot(fitted(fit), reside(fit))\n\n#add a horizontal line at y=0 \n<\/span>abline(0,0)<\/strong> <\/pre>\n![\"\"](\"https:\/\/statorials.org\/wp-content\/uploads\/2023\/08\/voler1.jpg\")
<\/p>\n library<\/span> (lmtest)\nlibrary<\/span> (sandwich)\n\n#calculate robust standard errors for model coefficients\n<\/span>coeftest(fit, vcov = vcovHC(fit, type = ' HC0<\/span> '))\n\nt test of coefficients:\n\n Estimate Std. Error t value Pr(>|t|) \n(Intercept) 71.1576 3.3072 21.5160 2.719e-14 ***\nhours 1.9454 1.2072 1.6115 0.1245 \n---\nSignificant. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1\n<\/strong><\/pre>\n Ek kaynaklar<\/strong><\/span><\/h2>\n
R’de do\u011frusal regresyon \u00e7\u0131kt\u0131s\u0131 nas\u0131l yorumlan\u0131r<\/a>
R’de art\u0131k arsa nas\u0131l olu\u015fturulur<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"