J. J. Salamanca Jurado
Our main goal in this work is to find an hyperplane which best models a random variable in the Lorentz-Minkowski spacetime, L^n. This space is an affine space endowed with a pseudo-metric, specifically, a non-degenerate
bilinear form with signature (- + ... +). We focus on the class of hyperplanes whose induced scalar product is Euclidean: the spacelike hyperplanes. The regression problem is considered by endowing the statistical parameters with natural geometric meaning. To get a suitable error function, we consider the total distance function to a spacelike hyperplane. As a result, we obtain different behaviors from the Euclidean case. We prove existence and uniqueness of the regression spacelike hyperplane in a wide class of random variables.
Palabras clave / Keywords: regression hyperplane, pseudo-metric space, random variable
Programado
Sesión M05 Análisis Multivariante
30 de mayo de 2018 15:50
Sala 3
