ncku-talk2017-02-22

Colloquium

DATE 2017-02-22　16:10-17:00

PLACE 數學館3174教室

SPEAKER 黃郁芬教授（中正大學數學系）

TITLE Influence Analysis on Linear Regression for Symbolic Interval Data

ABSTRACT Nowadays, with the advent of computers, data sets become inevitably large than before. This brings the diﬃculty in performing standard statistical analysis. Hence, such huge data sets might be aggregated in some fashion and the resulting summary data may be represented by lists, intervals, histograms and the like, which are called symbolic data. Linear regression is one of the most popular and useful tools to analyze the data for studying the relationship between a response variable and its explanatory variable(s). During the ﬁtting process, observations that are suspicious can greatly inﬂuence the results of the analysis. Therefore, detection of such inﬂuential points becomes an essential task. Many literatures have studied for the inﬂuence analysis in linear regression for classical data. However, to our knowledge, a study in the inﬂuence analysis on regression for symbolic data has not been explored in the literature. In this paper, we develop three sample versions of the inﬂuence function motivated by Hampel (1974) to identify suspicious concepts which cause seriously adverse eﬀects on the linear regression analysis results for symbolic interval data. Also we illustrate these proposed methods with simulation studies and real data examples.

Colloquium

DATE	2017-02-22　16:10-17:00

PLACE	數學館3174教室

SPEAKER	黃郁芬教授（中正大學數學系）

TITLE	Influence Analysis on Linear Regression for Symbolic Interval Data

ABSTRACT	Nowadays, with the advent of computers, data sets become inevitably large than before. This brings the diﬃculty in performing standard statistical analysis. Hence, such huge data sets might be aggregated in some fashion and the resulting summary data may be represented by lists, intervals, histograms and the like, which are called symbolic data. Linear regression is one of the most popular and useful tools to analyze the data for studying the relationship between a response variable and its explanatory variable(s). During the ﬁtting process, observations that are suspicious can greatly inﬂuence the results of the analysis. Therefore, detection of such inﬂuential points becomes an essential task. Many literatures have studied for the inﬂuence analysis in linear regression for classical data. However, to our knowledge, a study in the inﬂuence analysis on regression for symbolic data has not been explored in the literature. In this paper, we develop three sample versions of the inﬂuence function motivated by Hampel (1974) to identify suspicious concepts which cause seriously adverse eﬀects on the linear regression analysis results for symbolic interval data. Also we illustrate these proposed methods with simulation studies and real data examples.