English

Parametrized measure models

Differential Geometry 2017-07-17 v3

Abstract

We develope a new and general notion of parametric measure models and statistical models on an arbitrary sample space Ω\Omega which does not assume that all measures of the model have the same null sets. This is given by a diffferentiable map from the parameter manifold MM into the set of finite measures or probability measures on Ω\Omega, respectively, which is differentiable when regarded as a map into the Banach space of all signed measures on Ω\Omega. Furthermore, we also give a rigorous definition of roots of measures and give a natural definition of the Fisher metric and the Amari-Chentsov tensor as the pullback of tensors defined on the space of roots of measures. We show that many features such as the preservation of this tensor under sufficient statistics and the monotonicity formula hold even in this very general set-up.

Keywords

Cite

@article{arxiv.1510.07305,
  title  = {Parametrized measure models},
  author = {Nihat Ay and Jürgen Jost and Hông Vân Lê and Lorenz Schwachhöfer},
  journal= {arXiv preprint arXiv:1510.07305},
  year   = {2017}
}

Comments

29 pages, final version to appear in Bernoulli Journal

R2 v1 2026-06-22T11:28:29.243Z