线性混合模型最佳线性无偏预测(BLUP)不仅适用于数据不平衡和误差方差异质试验的分析,而且对随机效应的排序会更准确.在实际试验分析中由于真实方差参数值未知而采用估计值时,BLUP转变为所谓经验性BLUP(eBLUP).为了探讨eBLUP在作物区域试验品种评价的效果,本文以我国2012—2014年长江流域油菜区域试验12套产量资料为例,对eBLUP在品种主效应和特定环境中效应的估计、排序及差异比较t测验等方面与方差分析综合比较.结果表明,对品种主效应,eBLUP与方差分析算术平均值仅有较小差异,品种排序在eBLUP与算术平均值法相同;对特定环境中品种效应,eBLUP与算术平均值法有较大差异,品种排序在eBLUP较算术平均值法更准确;用Kenward-Roger法估算基于eBLUP的效应差异t测验的自由度,无论对品种主效应还是对特定环境中品种效应,eBLUP和方差分析有着相近的显著性(α=0.05)测验效果.%The mixed model and its best linear unbiased prediction (BLUP) are more suitable for analysis of the trails with unbal-anced data and heterogeneous errors, and BLUP can provide more accurate ranking for random effects. In analysis of practical trials, the variance parameters are unknown and their estimates have to be used. In this case, the BLUP become empirical BLUP (eBLUP). To investigate the performance of eBLUP in variety valuation of regional crop trials in China, this paper compared es-timates, ranking andt-test for both variety main effects and location-specific variety effects between using ANOVA and eBLUP based on 12 yield data sets from the rape variety evaluation trials of China from 2012–2014. The method of Kenward and Roger was used for approximating denominator degrees of freedom int-test for effect difference comparison based on eBLUP. The re-sults showed that in view of variety main effects, there was only small discrepancy between eBLUP and arithmetic mean of ANOVA, variety ranking was the same between them; in view of location-specific variety effects, there was large discrepancy between eBLUP and arithmetic mean, variety ranking was more accurate in eBLUP than arithmetic mean; int-test for both variety main effects and location-specific variety effects, eBLUP and ANOVA provided similar variety pairs with significant (α = 0.05) difference.
展开▼
