Simplicial and Minimal-Variance Distances in Multivariate Data Analysis

Abstract In this paper, we study the behaviour of so-called k -simplicial distances and -minimal-variance between a point sample. The family includes Euclidean distance, Mahalanobis Oja’s simplex distance many others. We give recommendations about choice parameters used to calculate distances, including size sub-sample simplices improve computation time, if needed. introduce new which call distances. Each these is constructed using polynomials in sample covariance matrix, with aim providing an alternative inverse that applicable when data degenerate. explore some applications considered outlier detection clustering, compare how affected for different parameter choices.