A CLUSTERING ALGORITHM FOR DETECTING INFLUENTIAL SUBSETS IN MULTIVARIATE METHODS

A clustering algorithm is proposed for detecting influential subsets of observations in multivariate methods such as principal component analysis, exploratory factor analysis and covariance structure analysis where the influence functions have been derived. It makes clusters hierarchically by optimizing a criterion defined by a specified aspect of influence among some different aspects such as the influence on the estimate, on its precision and on the goodness of fit. A numerical example is given to show the usefulness of the proposed method in maximum likelihood factor analysis (MLFA).