Comparison Of Some Variance Component Estimation Methods With Respect to Type I Eror Rate

Genc S., MENDEŞ M., KOCABAŞ Z., Soysal M.

JOURNAL OF TEKIRDAG AGRICULTURE FACULTY-TEKIRDAG ZIRAAT FAKULTESI DERGISI, cilt.9, sa.2, ss.67-74, 2012 (ESCI) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 9 Sayı: 2
  • Basım Tarihi: 2012
  • Dergi Adı: JOURNAL OF TEKIRDAG AGRICULTURE FACULTY-TEKIRDAG ZIRAAT FAKULTESI DERGISI
  • Derginin Tarandığı İndeksler: Emerging Sources Citation Index (ESCI), TR DİZİN (ULAKBİM)
  • Sayfa Sayıları: ss.67-74
  • Anahtar Kelimeler: Analysis of Variance, Variance Components, Probability of Type I Error, Simulation
  • Ankara Üniversitesi Adresli: Evet

Özet

In this study; Variance components and probability of Type I Error were estimated by Analysis of Variance (ANOVA), Maximum Likelihood (ML) and Restricted Maximum Likelihood (REML) when the assumptions of analysis of variance were violated. For this purpose, random numbers from equal and not equal variances Z (0,1), t (10), (3) chi(2), beta(5,2) distributions with the various sample sizes (n=5, 10, 20, 30, 40, 50) and group numbers (k=3, 4, 10) were generated by simulation technique. Depending on the findings of this study with 100000 simulation, it is concluded that probability of type I error (alpha) estimated by ANOVA and ML was not protected in small sample sizes (n=5, 10) even if assumption of analysis of variance were met. Whenever variance components were estimated by REML, probability of type I error protected as % 5. All methods (ANOVA, ML, REML) were affected when homogeneity was violated at this affect shows clearly by measured heterogeneity. When the group number (k=4,10) increased that was affect to change probability of type I error.