复杂数据下基因型与环境互作的统计分析

发布时间：2018-05-25 01:31

本文选题：基因型与环境互作 + 缺失数据　；参考：《北京工业大学》2016年硕士论文

【摘要】：多环境试验是最基础最常用的农业试验,基因型与环境互作(GE)在作物中是比较普遍的现象。但是,在多环境试验中,由于各种各样的原因,试验数据往往存在缺失。本文主要研究在数据存在缺失情况下,基因型与环境互作效应分析方法。由于主效应可加互作效应可乘(AMMI)模型、基因型主效应加基因型与环境互作效应(GGE)模型和联合回归模型(又称为F-W模型)是常用的GE效应分析模型,本文主要考虑这三种模型。本文对基因型效应和环境效应是否看作随机变量分别考虑。首先,将基因型效应和环境效应均看作固定效应,即在固定效应情形下,给出了有缺失数据时AMMI模型和GGE模型的基于EM算法的参数估计。然后,将基因型效应看作固定,将环境效应看作随机,考虑了F-W模型在异方差混合模型下的参数估计,给出了其极大似然估计,由于该极大似然估计通常没有显式解,需通过迭代法求得,而EM算法由于其优良性常用于缺失数据情形下,本文又给出了对应的基于EM算法的参数估计方法。最后,本文通过一个实际数据来说明本文给出的固定效应下GGE模型的基于EM算法的参数估计的效果。结果表明,在数据缺失比例不高时,缺失位置数据用给出的参数估计方法填补后,能够很好的恢复原始数据的模式。同时又以一个缺失数据集为例,对本文所给方法对品种一般适应性、特殊适应性、环境鉴别力和代表性等的影响进行了研究。
[Abstract]:Multi-environment test is the most basic and most commonly used agricultural experiment, genotype and environment interaction is a common phenomenon in crops. However, due to various reasons, the test data are often missing in the multi-environment test. In this paper, the analysis method of genotypic and environmental interaction effect is studied in the absence of data. As the main effect additive interaction effect can be multiplied by AMMIM model, the GGE model and the combined regression model (also called F-W model) are commonly used to analyze GE effects. These three models are mainly considered in this paper. Whether genotype effect and environmental effect are considered as random variables are considered in this paper. First, the genotype effect and the environmental effect are considered as fixed effects, that is, in the case of fixed effects, the parameter estimation of AMMI model and GGE model with missing data is given based on EM algorithm. Then, the genotype effect is regarded as fixed, the environmental effect is considered as random, the parameter estimation of F-W model under heteroscedasticity mixed model is considered, and its maximum likelihood estimate is given, because the maximum likelihood estimate usually has no explicit solution. The EM algorithm is usually used in the case of missing data because of its superiority. The corresponding parameter estimation method based on EM algorithm is given in this paper. Finally, an actual data is given to illustrate the effect of parameter estimation of GGE model based on EM algorithm under the fixed effect given in this paper. The results show that when the missing ratio is not high, the missing position data can be restored to the original data model well after the missing position data is filled by the given parameter estimation method. At the same time, taking a missing data set as an example, the effects of the methods given in this paper on the general adaptability, special adaptability, environmental discrimination and representativeness of varieties were studied.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：S33

【相似文献】