Jb. Duarte et al., Estimators of variance components in the augmented block design with new treatments from one or more populations, PESQ AGROP, 36(9), 2001, pp. 1155-1167
This work compares by simulation estimates of variance components produced
by the ANOVA (analysis of variance), ML (maximum likelihood), REML (restric
ted maximum likelihood), and MIVQUE(0) (minimum variance quadratic unbiased
estimator) methods for augmented block design with additional treatments (
progenies) stemming from one or more origins (crosses). Results showed the
superiority of the MIVQUE(0) estimation. The ANOVA method, although unbiase
d, showed estimates with lower recision. The ML and REML methods produced d
ownwards biased estimates for error variance (sigma (2)(e)), and upwards bi
ased estimates for genotypic variances (sigma (2)(g)), particularly the ML
method. Biases for the REML estimation became negligible when progenies wer
e derived from a single cross, and experiments were of larger size with rat
ios sigma (2)(g)/sigma (2)(e) >0.5. This method, however, provided the wors
t estimates for genotypic variances when progenies were derived from severa
l crosses and the experiments were of small size (n< 120 observations).