Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/36732
標題: 以不同統計模型分析基因型與環境交感效應之研究
The Study on the Analysis of Genotype Environment Interaction Using Different Statistical Models
作者: 陳璿宇
Chen, Hsang-Yu
關鍵字: genotype environment interaction;基因型與環境交感效應;A.N.O.V.A.;regression model;principal components analysis model;AMMI model;partial least square model;biplot;變異數分析;回歸模型;主成分分析模型;AMMI模型;淨最小平方法模型;雙軸圖
出版社: 農藝學系
摘要: 
品種種植在不同環境下,隨著環境的改變,品種間有著不同的反應。此種隨著環境改變造成品種産生不同反應的現象稱爲基因型與環境交感效應。區域試驗資料中,包括累加性的主效應與非累加性的基因型與環境交感效應。選擇有效的基因型與環境交感效應的統計方法除了能夠了解基因型如何受到環境的影響外,更能找出穩定性高的基因型。
本研究以農業試驗所1998年秋作玉米五個品系(1-5)在六個地區(E1-E6)所得之產量的資料,使用變異數分析、回歸分析模型、PCA模型與AMMI模型探討其對交感效應的解釋能力;以及PCA模型、AMMI模型與PLS模型分析併變因子與基因型及環境之間的交互作用。
結果發現,就變異數分析的結果來說,品系效應、環境效應與交感效應均為顯著,品系效應與環境效應各佔了處理變異的41.19與45.31﹪,表示品系間與環境間都存在著相當大的變異。其中交感效應平方和佔了處理變異的13.5﹪卻使用了69﹪的自由度,比品系效應與環境效應所使用的自由度多。變異數分析只能初步的劃分品系、環境的主效應與交感效應,並不能解釋品系與環境對交感效應的影響。
回歸分析模型中,使用Perkins 和 Jinks 回歸模型與Eberhart 和 Russell回歸模型分析交感效應,結果兩種模型得到的結果相同。回歸模型所計算的變異佔處理變異的88.61 ﹪,但是就交感效應方面來說,只佔了交感效應15.66 ﹪的變異。在包括六個區域的試驗中,發現品系5為最穩定的品系。
PCA模型方面,PCA模型解釋了97.65﹪的處理變異,品系3為最穩定的品系,品系5為次穩定的品系。在PCA模型中,因為選取的前兩個成分納入了主效應的影響,所以在品系與環境間、併變因子與品系和環境之間的相互關係方面,並不是很明確。
在AMMI模型方面,AMMI模型共解釋了99.51﹪的處理變異,其中共解釋了95.4﹪的交感效應變異。AMMI模型的結果中品系3為最穩定的品系,品系5為次穩定的品系。在品系或環境併變因子方面,AMMI模型能夠清楚地描述併變因子與品系和環境之間的關係。
而PLS模型在解讀品系或是環境併變因子與交感效應之間的相互關係上,得到了不錯的結果。在PLS模型中,雖然環境併變因子與交感效應的相互關係似乎略顯薄弱,但是在品系併變因子方面,PLS模型能夠在雙軸圖上清楚的找出影響產量的品系併變因子。
綜合言之,AMMI模型在解釋變異能力方面表現最好;在敘述併變因子與交感效應相互關係方面,PLS模型則是一個最好的模型。若是不考慮併變因子,AMMI能提供一個快速且有效率的分析方法。

Cultivars grown in multi-environments react differently to environmental changes. This differential response of cultivars from one environment to another is called genotype ×environment interaction. There are additive main effect and nonadditive genotype ×environment interaction in the regional trial data. Applying an efficient statistical method to analyze genotype ×environment interaction can explore how the cultivars were affected by the environments. This study used the yield data provided by T.A.R.I. which includes five lines grown in six environments in 1998. It was shown that not only the ability of ANOVA model, regression model, PCA model, and AMMI model for interpreting if the interaction effect, but also shown how covariables affect cultivars and environments by using PCA model, AMMI model, and PLS model.
The results showed that line effect, environment effect, and interaction effect were significant in the ANOVA model. The line effect and environment effect explained 41.19 and 45.31% of the treatment variation, respectively. It was also shown that there was highly variation within lines and within environments. The interaction sum of squares explains 13.5% of treatment variation by using 69% degrees of freedom. The ANOVA model only partitions line effect, environment effect, and interaction effect. It could not intepret how lins and environments were affected by interaction.
The Perkins and Jinks model and Eberhart and Russell model were used to analyze the genotype ×environment interaction in regression models. It was appeared that two models had the same result, and both of them explained about 88.61% of the treatment variation. However, only 15.66% of the interaction variation were explained. It was found that line 5 was the most stable among all tested lines in the regional trial.
The PCA model interpreted 97.65% of treatment variation. It was found that line 3 was the most stable and line 5 was the sub-stable. The explicit relationship between lines and environments and the relationship of covariables between lines and between environ- ments were not found because the first two components only contain the main effect in the PCA model.
The AMMI model intepreted 99.51% of treatment variation, and accounted for 95.4% of interaction variation. It was also found that line 3 was the most stable and line 5 was the sub-stable. The relationship between lines and environments and identified major en- vironmental (or line's) covariables that caused interactions between lines and between environments were explained clearly in the AMMI model.
There were excellent results in interpreting relationship between lines and environ- ments and the relationship of covariables between lines and between environments in PLS model, especially in line covariables. Although the relationship between environ- mental covariables and interaction was not clearly, however, it was easily found that the relationship between line covariables and yield using the biplot.
The ability in interpreting variation of AMMI model is the best. The PLS model clearly describes the relationship between lines and environments and the relationship of covariables between lines and between environments. However, the AMMI model is a quick and efficient statistical model without considering covariables.
URI: http://hdl.handle.net/11455/36732
Appears in Collections:農藝學系

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.