毛豆區域試驗多變量穩定性之研究

Abstract

作物產量之穩定性分析 (stability analysis) 針對區域試驗 (regional trail) 中的產量進行分析，經由產量要素在不同環境下之表現或交感作用，探討作物產量在不同環境中是否具有穩定表現。在台灣，多使用Finlay與Wilkinson模式的迴歸係數，藉由建立品系自身產量與環境指標 (environment index) 的迴歸模式，計算其迴歸係數作為穩定性介量。 本研究使用高雄場毛豆區域試驗 (regional experiment) 之資料，產量計算往往是經由數個不同的性狀線性組合而成，而對於穩定性分析方法的研究應可擴展至多變量方法的領域，以給予育種家更為客觀以及全面的資訊，並減少以往使用產量構成要素運算而得之原始產量可能所造成的偏差。本研究運用集群分析 (cluster analysis) 以及因素分析 (factor analysis) 等多變量方法 (multivariate method) 分析。先以集群分析對毛豆產量性狀進行分群，並進一步使用因素分析擷取性狀背後之因素，經由產生因素得點作為替代毛豆產量之指標。在取得指標後結合Finlay與Wilkinson模式之迴歸係數進行穩定性分析，探討使用多變量方法運用在穩定性分析中的可行性。研究結果顯示，在經由適當的性狀選取之後，可由因素分析等多變量方法擷取出穩健並與經濟產量高度相關的因素，藉由此因素所得到之標準化因素得點可作為產量指標，用以分析同一年組中各品系的產量表現。雖然該指標之應用仍具有部分限制，本產量指標仍具相當的發展潛力。往後若能將其運用在更多真實資料並確立一標準分析流程，可望實際運用於實務之中。

The stability analysis analyzes crop yields in different environments to confirm if varieties have stable phenotypes. Stability analysis discusses the problem that the interaction between genotype and environment by many parameter methods. In Taiwan, Finlay and Wilkinson coefficient, which builds linear regression models with yield (dependent variable) and environment index (Independent variable), is regularly used. However, when yields involve multiple traits, it may be plausible to analysis the data using multivariate methods. We analyzed regional experiment data of soybean in Kaohsiung District Agricultural Research and Extension Station by multivariate methods. In this study, we used cluster analysis to group agronomic traits related to yield and then make new yield index by factor analysis. After getting index, we chose Finlay and Wilkinson coefficient as stability analysis to find stable varieties. In conclusion, after choosing proper traits, we can get the factor scores, which are highly correlated with yields, by factor analysis and defined the scores as the index of yields. Though the index has some restrictions, it is still a potential index to find stable varieties in regional experiment.