利用拔靴法於数据分析使用 MATLAB 方法.pdf.pdfVIP

管理科學研究 Vol.1, No.1, 2003 第49-56 頁 利用拔靴法於數據分析：使用MATLAB 方法 Bootstrapping in dara analysis ：A matlab approach 1 2 吳信宏 郭廣洋 摘要 拔靴法為無母數且為電腦密集的方法，採用機率的推論預估母體參數，不需要對母 體有任何理論假設。由於拔靴法相當的複雜且不易了解，因此並未廣泛的運用於社會科 學領域。雖然拔靴法可以透過使用Lotus 1-2-3 計算，但是在效能上面仍然不盡理想。本 文的目的在於介紹與解釋如何使用MATLAB 電腦語言簡化拔靴法的使用。 關鍵詞：MATLAB 、拔靴、資料分析、無母數 Abstract The bootstrap is a non-parametric but computer intensive method for making probability-based inferences about population parameters without theoretical assumptions. However, due to its complexity and a lack of understanding, the bootstrap has not been widely used in social science fields. Using Lotus 1-2-3 in bootstrapping is helpful but still somewhat ineffective. The purpose of this paper is to introduce and explain how the bootstrap method can be applied much easier using MATLAB computer language. Keywords : MATLAB, Bootstrap, Data analysis, Non-parametric 1. Introduction The bootstrap method first introduced in 1979 by Bradly Efron has become one of the important techniques in data analysis (Efron, 1979; Franklin and Wasserman, 1991; Wu and Wang, 1996-97; Wu, 2001). It is a non-parametric but computer-intensive method for making probability-based inferences about population parameters without theoretical assumptions. In a statistical context, the bootstrap method describes a way to generate an entire distribution of a population starting from only a sample. Because of its complexity and a lack of understanding, the bootstrap method has not been widely used in social science fields (Woodroof, 2000). Two major reasons are summarized as follows (Woodroof, 2000): (1) There is