獨立性環境指標在落花生產量穩定性分析之建立與應用

Independent Environmental Indices for Stability Analysis in Peanut Yield Trials

呂秀英(Hsiu-Ying Lu)

《中華農業研究》 40卷4期 (1991/12) Pp. 385-395

https://doi.org/10.29951/JARC.199112.0003

摘要

直線迴歸模式能有效地估計數量性狀對環境的適應能力。然而，傳統的直線迴歸模式中係以各環境均值與總平均值的離差作爲該一試驗的環境效應估值，再以各品種收量隨此環境指標作迴歸所得的迴歸係數來評價品種的穩定性。由於該環境指標與分析資料間並非獨立，實不符合迴歸分析中自變數為固定值的一般假設，因此，本研究以落花生品系試驗資料為實例說明，重新設定直線迴歸之各種獨立性環境指標，並與傳統環境指標評估品系穩定性的結果作一比較討論，結果發現：利用逢機、分層及極端不同取樣方法將所有品系分爲兩組，彼此利用他組之環境效應估值作爲該組的環境指標，與傳統環境指標之評估結果幾無差異，應是最理想的獨立性環境指標。若不計迴歸離差均方上的影響，則亦可將各迴歸品系的資料自整個資料中去除後，以剩餘品系資料的環境效應估值作為獨立性環境指標，但必須每一品系各有不同的環境指標，計算上比較麻煩。選定一對照品系，以其環境離差作爲各參試品系的共同環境指標估值，必須選擇對環境反應與參試品系相同型式的標準品系。至於將資料所有重複逢機劃分爲指標組及受試組的方法所建立的獨立性環境指標，只有在重複次數夠多時，始能獲致較準確的穩定性評估結果。

關鍵字

穩定性分析 ； 獨立性環境指標 ； 落花生

並列摘要

Linear regression model is useful to evaluate the adaptation of a quantitative character to the change of environments. However, in the conventional regression analysis for stability, mean performance of individual genotypes is regressed on the environmental index measured by the mean performance of all genotypes grown in that environment. The use as independent variable of values which are not independent of the phenotypic variable regressed on to them, therefore the assessment of stability should correspond most closely to the average response of the trial individuals. In the present study, regression analysis on peanut yield trials were performed to construct different measures of independent environmental values and to compare their relative stability. The results showed that division of the genotypes by random, stratified and extreme method into two subsets using one subset to measure the genotype-environment interactions and the second subset to assess the environments made little difference on the general conclusion of regression analysis, thus can be regarded as ideal independent environmental indices. Inclusion of assessment genotype to assess its environmental values is useful when the effect on deviations from regression mean squares (S^2d) is ignored, but its use becomes difficult since there requires the calculation of a separate environmental index for each regression. Use of check genotypes which respond in the same way as the trial genotypes to assess the environments may also be satisfactory. Division of replicates at random into two subsets is preferred only when there are many replicates in the trial.

並列關鍵字

Stability analysis ； Independent environmental index ； Peanut