Invariant Geometric Structures on Statistical Models
We review the notion of parametrized measure models and tensor fields on them, which encompasses all statistical models considered by Chentsov , Amari  and Pistone-Sempi . We give a complete description of n-tensor fields that are invariant under sufficient statistics. In the cases \(n= 2\) and \(n = 3\), the only such tensors are the Fisher metric and the Amari-Chentsov tensor. While this has been shown by Chentsov  and Campbell  in the case of finite measure spaces, our approach allows to generalize these results to the cases of infinite sample spaces and arbitrary n. Furthermore, we give a generalisation of the monotonicity theorem and discuss its consequences.
KeywordsBanach Space Measure Space Sample Space Regularity Assumption Logarithmic Derivative
- 2.Ay, N., Jost, J., Lê, H.V., Schwachhöfer, L.: Information geometry, book in preparationGoogle Scholar
- 3.Amari, S.: Differential Geometrical Theory of Statistics. In: Amari, S.-I., Barndorff-Nielsen, O.E., Kass, R.E., Lauritzen, S.L., Rao, C.R. (eds.) Differential Geometry in Statistical Inference. Lecture Note-Monograph Series, vol. 10. Institute of Mathematical Statistics, California (1987)Google Scholar
- 6.Chentsov, N.: Category of mathematical statistics. Dokl. Acad. Nauk USSR 164, 511–514 (1965)Google Scholar
- 8.Bauer, M., Bruveris, M., Michor, P.: Uniqueness of the Fisher-Rao metric on the space of smooth densities (2014). arXiv:1411.5577