Stability of doubly-intractable distributions
Statistics Theory
2020-08-13 v3 Numerical Analysis
Numerical Analysis
Probability
Statistics Theory
Abstract
Doubly-intractable distributions appear naturally as posterior distributions in Bayesian inference frameworks whenever the likelihood contains a normalizing function . Having two such functions and we provide estimates of the total variation and Wasserstein distance of the resulting posterior probability measures. As a consequence this leads to local Lipschitz continuity w.r.t. . In the more general framework of a random function we derive bounds on the expected total variation and expected Wasserstein distance. The applicability of the estimates is illustrated within the setting of two representative Monte Carlo recovery scenarios.
Cite
@article{arxiv.2004.07310,
title = {Stability of doubly-intractable distributions},
author = {Michael Habeck and Daniel Rudolf and Björn Sprungk},
journal= {arXiv preprint arXiv:2004.07310},
year = {2020}
}
Comments
16 pages, to appear in Electronic Communications in Probability