迴歸模型的影響力點分析

在統計學中，離群值是一個，遠離其他觀察值的觀察點，離群值有分成幾種不同的種類，我們指出在不同方向上的離群值會對迴歸模型造成不同方面的影響，且儘管該筆資料是離群值也不一定是對迴歸模型具有影響力之資料。我們首先介紹不同方向上的離群值判斷方法，在知道有哪些種類的離群值後，我們同樣也想知道它會對我們的迴歸模型產生多少的影響力，在本篇研究中，針對影響力分析的診斷方法之組成元素進行探討，用權數來擾動我們想要分析的該筆資料，在影響力診斷方法中之組成元素是否具有共通點，與各自所具有之特色為何，再探討在資料模擬中不同的影響力診斷方法之差異。

關鍵字

離群值；影響力

並列摘要

In statistics, the outlier is the one that keeps away from other observations, and the outliers are classified as several different categories. We point out that outliers in different aspects have different effects on the regression model. Although the data is outliers, it is not necessarily has influence on the regression model. We first introduce the outlier diagnostic method in different directions. After knowing what kinds of outliers are they, we also want to know how much influence it will have on our regression model. In this paper, we will discuss the components of the diagnostic method. The weights are used to perturb the data we want to analysis. Do the components of the diagnostic method have common relationship, and what are the characteristics of each. And then discuss the differences in diagnostic methods in the data simulation