几种模型在玉米区域试验的适用性与品种评价中的差异研究

为了综合评价玉米区域试验的结果，采用Shukla模型、Finlay-Wilkinson模型、Eberhart-Russell模型和加性主效乘式互作模型(AMMI)等4种常用模型对中国5个类型的玉米品种区域试验资料进行了拟合与分析。结果表明，没有一个模型对所有试验资料的拟合效果都最佳，Finlay-Wilkinson模型对5个试验拟合效果均最差，AMMI模型对3个试验的数据拟合效果最佳，Shukla模型和Eberhart-Russell模型则分别对一个试验拟合效果最佳。各个模型在玉米品种产量差异显著性检验和稳定性排序等方面存在较大的差异。因此，在玉米区域试验的实际分析中应采用AMMI模型或者利用模型拟合信息量准则选用最佳的模型进行分析，以提高玉米区域试验对品种高产与稳产评价的准确性。

Suitability and Evaluation Difference of Models for Corn Regional Trials

Stability can comprehensively reflect quality and yields in corn. Our objective was to evaluate results of corn regional trials. Five data sets from various maize regional trials were fitted and analyzed by use of 4 typical models, including Shukla model, Finlay-Wilkinson regression model, Eberhart-Russell regression model and AMMI model. The results showed that, no one of the used models was the best in fitting the data for all trials, the Finlay-Wilkinson model was bad for all of 5 trials and the AMMI model was best for 3 trials; Shukla model and Eberhart-Russell model were best separately only for one trial. There was also obvious difference in stability ranking and statistical test about maize yield between the used models. Therefore, the AMMI model or the model selected based on information criterion in data fit was suggested for analysis of maize regional trials, in order to enhance the accurate in test of yield difference and assess of stability of maize variety.