References & Citations
Statistics > Methodology
Title: Inference for Empirical Wasserstein Distances on Finite Spaces
(Submitted on 11 Oct 2016 (v1), last revised 26 Apr 2017 (this version, v2))
Abstract: The Wasserstein distance is an attractive tool for data analysis but statistical inference is hindered by the lack of distributional limits. To overcome this obstacle, for probability measures supported on finitely many points, we derive the asymptotic distribution of empirical Wasserstein distances as the optimal value of a linear program with random objective function. This facilitates statistical inference (e.g. confidence intervals for sample based Wasserstein distances) in large generality. Our proof is based on directional Hadamard differentiability. Failure of the classical bootstrap and alternatives are discussed. The utility of the distributional results is illustrated on two data sets.
Submission history
From: Max Sommerfeld [view email][v1] Tue, 11 Oct 2016 11:54:18 GMT (367kb,D)
[v2] Wed, 26 Apr 2017 11:29:55 GMT (357kb,D)
Link back to: arXiv, form interface, contact.