Local Influence on Functional Data for Outlier Detection

• Main Authors: TSAI, YI-JIE, 蔡依潔
• Other Authors: HUANG, YUFEN
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/15592262869310980230

碩士 === 國立中正大學 === 數學系統計科學研究所 === 105 === Functional data consist of a collection of functions which are smooth curves or surfaces. This type of data has been vastly increas- ing common and attracted a lot of interest from a wide range of fields in the past decade. Consequently, several new statistical tools for func- tional data are continuously developed. Statistical analysis may be affected with the presence of outliers in the data which might lead to inaccurate conclusion. Therefore, detection of such influential curves is a necessary step to assure data quality for right decision making. Two major tools of sensitivity analysis are the influence function by Hample (1974) and the local influence by Cook (1986). In this thesis, we propose a method based on local influence to identify the outliers for functional data and make comparisons with existing methods. Two real data examples and two simulation studies show the excellent per- formance of our proposed approach.