頁籤選單縮合
題 名 | 田間試驗之空間混合模式和傳統變方分析法比較=Comparisons of Spatial Mixed Model and Conventional Analysis of Variance for Field Experiments |
---|---|
作 者 | 蔣國司; 劉弘一; 何玫儀; | 書刊名 | 作物、環境與生物資訊 |
卷 期 | 1:2 2004.06[民93.06] |
頁 次 | 頁79-88 |
分類號 | 430.34 |
關鍵詞 | 空間混合模式; 空間自身相關; 平均經驗變異數; 平均預測變異數; 地理統計學; 半序列變異圖; Spatial autocorrelation; Spatial mixed model; Average empirical variance; Average predicted variance; Geostatistics; Semivariogram; |
語 文 | 中文(Chinese) |
中文摘要 | 有學者提出以空間混合模式(spatial mixed model) 來解決此空間自身相關的問題(Littell et al. 1996, Stroup 2002) ,它是使用地理統計學(geostatistics) 的半序列變異圖(semivariogram) 參數來描述共變異數(covariance)的結構。Stroup (2002) 提出以檢定力分析(power analysis) 為工具,來比較空間混合模式與逢機完全區集設計之變方分析(RCBD ANOVA),但在多處理數時,比較上有所不便。故本文採用逢機完全區集設計下之試驗數據,以Besag and Kempton (1986) 提出的平均經驗變異數(average empirical variance, Emp)和平均預測變異數(average predicted variance, Pre)為比較標準,來比較結合地理統計學(geostatistics) 之空間混合模式和傳統變方分析之有效性及精確性。本研究藉由分析Mercer and Hall (1911) 在英國的Rothamsted 試驗站所收集的數據,來比較此兩種分析方法,結果發現當數據具空間效應時,空間混合模式比RCBD ANOVA 更能有效地解釋空間的變異性。又經進一步模擬不同地理統計學之參數下的數據對這兩種分析方法的影響,發現在有效性上,兩者有相似的結果,但在精確性上,空間混合模式遠勝過RCBD ANOVA。綜合研究結果,顯示田間試驗數據具空間變異性時,空間混合模式優於傳統變方分析。 |
英文摘要 | Recently, spatial mixed models have been proposed to solve the problem of spatial autocorrelation. In these models, the semivariogram of geostatistics were used to depict the covariance structure (Littell et al. 1996, Stroup 2002). Stroup (2002) compared the performance of spatial mixed models and ANOVA by the power analysis. Nevertheless, the approach is sometimes inconvenient when more treatments exist. Thus, the objective of this study utilized another criterion to compare the spatial mixed model and the conventional ANOVA among RCBD. The criterion mentioned is to use the average empirical variance (Emp) and the average predicted variance (Pre) (Besag and Kempton 1986) as the tool in order to evaluate their validity and precision. First, we compared the two methods which had been used to analyze the data collected at Rothamsted experimental station in U.K. (Mercer and Hall 1911). It was found that spatial mixed models were more efficient than RCBD ANOVA. Moreover, we further simulated the two methods among the different geostatistical parameters. In validity, the results were similar for the two methods. However, for the precision, spatial mixed models were superior to RCBD ANOVA. Therefore, the method of spatial mixed models gave a good insight into spatial variation compared with the conventional analysis of variance for field experiment data. |
本系統中英文摘要資訊取自各篇刊載內容。