Comparison of Variance Components Methods for One Way Random Effects Model in Cotton

In this study, emphasis is placed on comparing the variance components for one way model in cotton by ten methods: analysis of variance (ANOVA), Quasi-maximum-likelihood method (QML), Maximum-likelihood method (ML), Full maximum likelihood Procedure (FML), Restricted maximum-likelihood method (REML), Modified maximum-likelihood method (MML), Federer’s estimator (FE), Moment (MOM), Klotz-Milton-Zacks (KMZ) and Stein estimator (SE) methods. The results showed that the estimation of variance components in some methods were found equal to each other and some other methods gives different values. The environmental variance for ANOVA, QML, ML, FML, REML, MML, FE and MOM methods by equating mean square of error to its expected value in analysis of variance. The MOM method was registered the highest values of genetic variance, followed by ANOVA, REML and FE methods and followed by QML, ML and FML methods for all studied traits in cotton. While, the lowest values of genetic variance were found with MML, KMZ and SE methods. The cluster analysis for the methods of genetic variance estimates based on studied traits contained into four clusters i.e., the cluster I (MOM), the cluster II (ANOVA, REML and FE), the cluster III (QML, ML and FML) and the cluster IV (MML, KMZ and SE). These results indicate a similarity of the methods in the same each cluster and differences between the four clusters. The differences of these methods due to differed in calculated the genetic variance. The ten studied methods for BSH and ratio estimations were showed the same results for genetic variance for all studied traits, and quite the opposite for ratio.