SOME GEOMETRICAL ASPECTS OF DATA ANALYSIS AND STATISTICS

摘要: In this paper we discuss some desirable properties that a distance between probability spaces must satisfy, and from these considerations introduce the information metric in several different approaches. Once is introduced, its applications to data analysis statistics are explored. For instance, classic of variance methods may be obtained through geometrical approach, interpretation likelihood estimation given terms metric, by defining natural way individuals statistical population. Finally consider geometric general class elliptic distributions, exhibiting differential sectional curvatures, Ricci tensor, geodesic equations and, cases, evaluation Riemannian distance, called Rao for family distributions. Some hypothesis tests also discussed.