查詢結果分析
來源資料
相關文獻
- The Comparison between Anderson's and Multiple Regression's Algorithm in Calculating the Partial Correlation
- 微控制器嵌入式系統的開發環境
- 語音壓縮標準G.723.1在C62X DSP上的設計
- 淺談常態與零壹分布之動差遞迴公式
- Multiple Recusive Random Number Generators of Orders Two and Three
- 系統風險變動下新上市公司股票的長期報酬行為:遞迴迴歸之應用
- 磁場導向控制感應電動機控制器參數之自我調適
- An Overview of RNN-Based Mandarin Speech Recognition Approaches
- 應用於複雜運動環境之聯合物體運動參數計算演算法
- Analysis and Performance Evaluation of a Recursive Integration Scheme
頁籤選單縮合
題 名 | The Comparison between Anderson's and Multiple Regression's Algorithm in Calculating the Partial Correlation=安德生和多回歸演算法在計算淨相關係數的比較 |
---|---|
作 者 | 鄧志堅; | 書刊名 | 科技學刊 |
卷 期 | 10:3 2001.05[民90.05] |
頁 次 | 頁195-207 |
分類號 | 440.11 |
關鍵詞 | 淨相關係數; 遞迴; 複迴歸分析; Partial correlation; Recursion; MATLAB; Multiple regression; |
語 文 | 英文(English) |
中文摘要 | 基本上有兩種方法可計算淨相關係數。一種是多?歸另一種是常相關係數的遞迴式。多迴歸方法由以西結於1950年使用,而遞迴的公式由安德生於1984使用來計算變數xi和x間剔除變數xη+1, xq-2,...,xp效果的淨相關係數。雖然有一統計軟體,如MINITAB,偏好多回歸方法,並沒有研究做過兩者優越性的比較。本篇文章試圖指出這兩種方法,安德生演算法和多迴歸方法,在大部份的情況下都產生相同的結果;但當多變量資料內變數間是高度相關時,安德生的方法較優。在此情況下,安德生演算法能得到正確答案但多迴歸方法卻不能。我們提出一個有三個M形態的檔案的MATLAB程式來實化該演算法。我們修正前一組MATLAB程式來得另外一組MATLAB程式來解決不當的一般相關係數的輸入而造成有些淨相關係數的絕對值大於1所造成的問題。我們用兩個例子說明、比較用這兩種方法所產生的淨相關係數。我們指出SAS FACTOR步驟計算出的淨相關係數並不是使用安德生的演算法。我們推薦安德生的演算法優於多迴歸演算法。 |
英文摘要 | There are basically two methods available of calculating the partial correlation. One is multiple regression and the other recursion of the ordinary correlation. The multiple regression method was used by Ezekiel early in 1950 while the recursion formula was used by T.W. Anderson in 1984 to calculate the partial correlation between variables xi and xj excluding the effects of variables xη+1, xq-2,...,xp. Although a statistical package, such as MINITAB, favors the multiple regression method, no research has been carried to compare the superiority of these two methods. This paper attempts to show that both Anderson's algorithm and multiple regression method are identical in calculating the partial correlation in most circumstances; however, when some variables of the multivariable data are highly correlated, then Anderson's algorithm does appear to be favorable. In this situation, Anderson's algorithm is able to determine the correct solution while the multiple regression is not. A MATLAB program with three M-files was provided to implement the Anderson's algorithm. Another MATLAB program was modified from the former to deal with improper ordinary correlation input that gave partial correlation values greater than one in absolute value. We used two numerical examples to compare the results of the partial correlation generated by the two above methods. We point out that the SAS FACTOR procedure does not use Anderson's algorithm to calculate the partial correlation. Finally we recommend the Anderson's algorithm over the multiple regression algorithm. |
本系統中英文摘要資訊取自各篇刊載內容。