查詢結果分析
來源資料
相關文獻
- The Robustness of Different Statistical Models for Prediction of Amylose Content of Rice Based on Near-Infrared Spectroscopy
- Estimating the Starch-iodine Blue Value of Residual Liquid of Rice by Near-infrared Reflectance Spectroscopy
- 近紅外線光譜分析上以淨最小平方法決定白米粉末直鏈澱粉含量的模型
- 評估以不同統計方法建立白米粉末直鏈澱粉含量的近紅外線光譜模型
- 利用均方誤矩陣於迴歸分析中主成份之選擇
- 迴歸分析自變數間相關性之探討
- 臺灣稻米支出彈性值的檢驗
- 開工前產糖率預測
- 資本結構的影響因素探討與實證研究
- 改良式脊迴歸分析法
頁籤選單縮合
題 名 | The Robustness of Different Statistical Models for Prediction of Amylose Content of Rice Based on Near-Infrared Spectroscopy=預測近紅外線光譜所分析稻米直鏈澱粉含量之不同統計模式的穩健性 |
---|---|
作 者 | 郭寶錚; | 書刊名 | 中華農學會報 |
卷 期 | 2:5 2001.10[民90.10] |
頁 次 | 頁464-474 |
分類號 | 434.111 |
關鍵詞 | 直鏈澱粉含量; 近紅外線反射光譜; 共線性; 複線性迴歸; 主成份迴歸; 淨最小平方迴歸; Amylose content; NIRS; Multicollinearity; MLR; PCR; PLSR; |
語 文 | 英文(English) |
中文摘要 | 近紅外線反射光譜研究可提供一個迅速且便捷的方法去預測一些品質性狀的含量。複線性迴歸(MLR)常會因所選波長間的共線性而造成嚴重的問題。主成分迴歸(PCR)及淨最小平方迴歸(PLSR)能被用於解這決這些嚴重的共線性問題。本研究的目的在於評估以MLR、PCR及PLSR建立稻米直鏈澱粉含量檢量方程式的穩健性,並比較三種統計方法在建模及預測上的表現。663個水稻樣本用來分析直鏈澱粉含量,複線性迴歸模式中由於所選出波長間的共線性問題而顯出較大的隨機干擾,因此造成複線性迴歸模式中估計而得的迴歸係數變異性膨脹,且經刪除極端值後模式表現顯著改變。主成分迴歸模式則較複線性迴歸模式穩健,淨最小平方迴歸模式最具穩健性。NIR可應用於迅速篩選不同直鏈澱粉含量的水稻品系,並且淨最小平方法的建模及預測表現優於主成份迴歸及複線性回歸。 |
英文摘要 | Near-infrared reflectance spectroscopy (NIRS) can provide a fast and convenient method to predict the content of some quality traits. Multiple linear regression (MLR) often causes sever problems because of the multicollinearity among the selected wavelengths. Principal component regression (PCR) and partial least squares regression (PLSR) can be implemented to overcome the serious multicolinearity problems. The objective of this study was to evaluate the robustness of MLR, PCR, and PLSR in developing the calibration equation for amylase content of rice. The performances of model-building and prediction using three different statistical methods were also compared. A total of 663 rice samples were included in analyzing the amylase content. MLR showed large random noise due to the serious multicollinearity problems among the selected wavelengths. Hence the inflated variability of the estimated regression coefficients was found in MLR model and the changes of model performance were significant after deleting the outlying observation. PCR performed more stable than MLR, while PLSR was the most robust calibration method for amylase content. The performances of model-building and prediction demonstrate that NIRS can be used for rapid screening of rice lines for amylase content in breeding program. |
本系統中英文摘要資訊取自各篇刊載內容。