A Comparision of Diagnostic Measures in Linear Regression

회귀진단을 위한 새로온 척도의 제안 및 상호비교

This paper is to study the various diagnostic measures for detecting outliers and influential cases in linear regression. In this paper we review the most common diagnostic measures and show the inter-relationships the exist among them. Based on the PRESS(Predicted REsidual Sum of Squares ) offered by Allen(1974) as a criterion for model selection, we propose three measures for detecting outliers and influential cases. Examples are given illustrating various diagnostic measures including Proposed measures.