THE DETECTION OF INFLUENTIAL SUBSETS IN LINEAR-REGRESSION BY USING ANINFLUENCE MATRIX

This paper presents a new method to identify influential subsets in li near regression problems. The procedure uses the eigenstructure of an influence matrix which is defined as the matrix of uncentred covarianc es of the effect on the whole data set of deleting each observation, n ormalized to include the univariate Cook statistics on the diagonal. I t is shown that the eigenstructure of the influence matrix is useful t o identify influential subsets and a procedure for detecting influenti al sets is proposed. The method is illustrated with two examples.