Statistical manifolds are statistical models (Q2490679): Difference between revisions

A statistical model is defined as an immersed submanifold \(M\) in Cap\((\Omega) \) -- the space of all probability measures on the sample space \(\Omega\); the Fisher metric \(g^{F}\) on a statistical model has been introduced by [\textit{C. R. Rao} et al., Differential geometry in statistical inference. IMS Lecture Notes-Monograph Series, 10. Hayward, CA: Institute of Math. Statistics (1987; Zbl 0694.62001)]. A statistical manifold is a Riemannian manifold \(\langle M,g\rangle \) equipped with a \(3\)-symmetric tensor \(S\) on \(M\) [in: \textit{S. L. Lauritzen}, Differential geometry in statistical inference. IMS Lecture Notes-Monograph Series, 10. Hayward, CA: Institute of Mathematical Statistics, p. 163--216 (1987; Zbl 0694.62001)]. The author proves that any statistical manifold is a statistical model. This gives a positive answer to questions asked by \textit{S. L. Lauritzen} (loc. cit.) and [\textit{S. Amari} and \textit{H. Nagaoka}, Methods of information geometry. Translations of Mathematical Monographs. 191 (Providence), RI: American Mathematical Society (2000; Zbl 0960.62005)].