小麦骨干亲本及其衍生后代产量分析模型及比较方法

Models and Comparison Methods for Yield Analysis of Wheat Backbone Parents and Its Derivative Progeny

为探讨小麦骨干亲本及其衍生后代的性状比较模型及分析方法,基于R语言程序,选取线性固定模型、重复内区组分群混合模型2种分析模型及Tukey多重比较、Scott-Knott分组2种分析方法,对2008年杨凌矮秆小麦品种区域试验的小区产量进行分析。结果表明,线性固定模型和混合模型的处理效应估计标准误的平均值分别为140.35及96.75,重复内区组分群混合模型处理效应估计标准误的平均值比线性固定模型降低31.06%。对混合模型分析结果进行Tukey多重比较(显著性水平0.05),可将252个处理分成有字母重叠的35个组,但是采用Scott-Knott分组可分成无字母重叠的4个组,可以看出不完全区组试验在大数量处理且数据缺失不平衡时使用混合模型进行处理的效应估计误差较小。通过Scott-Knott检验可得到字母不重叠的分组结果,它在大样本时结果表示简洁且易于解释。建议今后在此类试验中采用混合模型及Scott-Knott分组方法分析小区产量。

In order to explore the models and analysis methods of wheat backbone parents and its derivative progeny on character comparision,Base on R language program,The 2 models of linear fixed model and linear mixed model on grouping block within repeat,and the 2 analysis methods of Tukey＇s multiple comparison test and the Scott-Knott grouping method were chosen for studying on the data of plot yield from yangling dwarf wheat variety regional trials in 2008.The results showed that,the Standard error mean on the estimated treatment effects of linear fixed model and linear mixed model value were 140.35 and 96.75,respectively,the Standard error mean on the estimated treatment effects of linear mixed model on grouping block within repeat value was lowered 31.06%than linear fixed model mean value.The Results using linear mixed model,for Tukey＇s multiple comparison tests （Significance level 0.05）,would allow 252 treatment were divided into 35 groups with overlapping letters.However,the Scott-Knott grouping method would allow all treatment were divided into four groups without overlapping letters.As can be seen,the error would be smaller if we used mixed models to estimate treatment effect when a large number of treatments and unbalanced missing-data in in- complete block experiments.Through the Scott-Knott test,It can be got the results without overlapping letters.The results also have simple expression and easy to explain in large sample size,the linear mixed model and the Scott-Knott grouping method was suggested to apply in the same trials in the future,and would obtain the better results.