作物品种区域试验中品种均值估计的模型和方法——算术平均值、加权最小二乘估值和BLUP的比较

Models and Methods for Estimating Variety Means in Regional Crop Trials——Comparisons of Arithmetic Mean,Weighted Least Squares Estimates and BLUPs

针对作物区域试验中的品种均值估计问题 ,根据混合线性模型的一般原理 ,总结和提出多种加权最小二乘估计(WLSE)和最佳线性无偏预测 (BLUP)的方法 ,推导了这些方法的平衡数据计算简式 ;同时 ,利用 14套 2年多点的棉花区试资料和一套 4年多点的棉花品种试验对这些方法的预测效果进行验证比较。结果表明 ,与算术平均值相比 ,以环境内误差方差倒数加权的WLSE估值的预测精度 (包括预测差的大小和品种排名的一致性 )明显不同 ,但其高低因数据而异 ;其他WLSE估值以及BLUP的预测结果差别不大 ,和算术平均值以及相互间的相关系数和秩相关系数均在 0 93以上。

Based on the mixed linear model, several weighted least squares estimates(WLSEs)and best linear unbiased predictors(BLUPs) were summarized and proposed for estimating variety means in regional crop trials, and the corresponding calculating formulae were derived and presented for balanced data The data of 14 rounds of 2-year-multi-location regional cotton trials and a 4-year-multi-location cotton trial were used to compare the predictive efficiencies of arithmetic means, WLSEs and BLUPs The results showed that the predictive differences and variety ranks of the WLSE weighted by the reciprocals of error variances within environments (WLSEe) differed significantly from that of the arithmetic means, but the predictive accuracy of WLSEe increased or decreased irregularly in different trials; the predictive results of other WLSEs and BLUPs were similar to that of the arithmetic means, the correlation coefficients and rank correlation coefficients between them were all above 0 93