J. Fortiana, E. Boj, A. Esteve
Linear and quadratic geometry of vector spaces defined as column spaces of observational data matrices underlies a good portion of methods in Multivariate Data Analysis, broadening this appellation to encompass multivariate linear and generalized linear models. By means of a Euclidean metric, either the Pythagorean l^2 or a Mahalanobis-type metric, statistical statements and properties translate into equivalent statements and properties in distance
geometry. Several Distance-Based (DB) methods exploit this dictionary to adapt classic statistical procedures to situations where observables are non-numerical. In this perspective, for instance DB linear prediction appears as an operation on an old metric returning a new one, a transformation which may be viewed as a mapping from the convex cone of Euclidean metrics to itself.
Palabras clave / Keywords: distance-based statistics, joint metrics, distance-correlation
Programado
Sesión GT02-1: Análisis Multivariante y Clasificación (AMyC-1). Organizadora: Eva Boj del Val
29 de mayo de 2018 10:30
Sala 5
