English

Duality for real and multivariate exponential families

Probability 2021-08-10 v2

Abstract

Consider a measure μ\mu on Rn\R^n generating a natural exponential family F(μ)F(\mu) with variance function VF(μ)(m)V_{F(\mu)}(m) and Laplace transform exp(μ(s))=Rnexp(\<s,x)μ(dx). \exp(\ell_{\mu}(s))=\int_{\R^n} \exp(-\<s,x\>)\mu(dx). A dual measure μ\mu^* satisfies μ(μ(s))=s.-\ell'_{\mu^*}(-\ell'_{\mu}(s))=s. Such a dual measure does not always exist. One important property is "μ(m)=(VF(μ)(m))1,\ell"_{\mu^*}(m)=(V_{F(\mu)}(m))^{-1}, leading to the notion of duality among exponential families (or rather among the extended notion of T exponential families TFT\hskip-2pt F obtained by considering all translations of a given exponential family FF).

Keywords

Cite

@article{arxiv.2104.05510,
  title  = {Duality for real and multivariate exponential families},
  author = {Gérard Letac},
  journal= {arXiv preprint arXiv:2104.05510},
  year   = {2021}
}

Comments

25 pages

R2 v1 2026-06-24T01:04:58.122Z