頁籤選單縮合
題名 | 無複製點多因子實驗中離群值的偵測=Finding Outliers in Unreplicated Factorial Designs |
---|---|
作者姓名(中文) | 廖本煌; | 書刊名 | 中國統計學報 |
卷期 | 37:4 1999.12[民88.12] |
頁次 | 頁371-383 |
分類號 | 319.9 |
關鍵詞 | 多因子設計; 常態機率圖; 離群值; Factorial designs; Normal probability plot; Outliers; |
語文 | 中文(Chinese) |
中文摘要 | 多因子設計中,若出現離群值 (outlier) 時,對變異數分析會產生較重大的影響。對於偵測離群值的方法及基準,在過去為一些人所探討。許多人用模擬法或近似法來確認離群值,但這些方法並非均是很容易的,當模式中因子的個數增加時,常使得分析更為複雜。在實際上,許多工程師需要一個簡單的方法來偵測離群值,以幫助他們作實驗分析。本文的目的除了指出在文獻上的一些方法之間是對等,並且提出一個運用圖形來偵測離群值的簡易方法。 |
英文摘要 | The presence of one or more outliers in factorial or fractional factorial designs can seriously affect the analysis of variance. Numerous criteria and techniques for outliers detection have been proposed and discussed for decades. Many of the procedures rely upon critical levels obtaining from either simulation studies or approximate theory, and become very complex when the model under consideration inc1ude several factors. In practice, engineers need a simple method to help the analysis of experiments. The purpose of the present paper is to describe and exemplify the normal probability plotting which provides to check against the spurious points. We also discussed some relevent issues in handling outliers inc1uding showing the equivalence among some well-known test statistics. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。