Sankhya: The Indian Journal of Statistics

2000, Volume 62, Series A, Pt. 1, pp. 135--143

PROCEDURES FOR THE IDENTIFICATION OF MULTIPLE INFLUENTIAL OBSERVATIONS USING LOCAL INFLUENCE

By

M. MERCEDES SUÁREZ RANCEL and MIGUEL A. GONZÁLEZ SIERRA, University of La Laguna, Spain

SUMMARY. Some authors point out that one of the attractions of the local influence concept is that it assesses the effect of joint perturbations on the data cases. Thus, in a local sense the results are free from masking and swamping effects that present difficulties to individual case-deletion methods. However, the available outliers identification methods using local influence often do not succeed in detecting them because, they are based on measures which are not resistant to the masking and swamping effects, they are affected by the observations they are supposed to indentify, and they don't find suitable cut off points. We introduce two test procedure for the detection of multiple locally influential observations that mitigate these problems.

AMS (1991) subject classification. 62J20.

Key words and phrases. Local influence; masking; multiple influence observation; swamping.

Full paper (PDF)