Id: arxiv/2603.08939 Url: https://arxiv.org/abs/2603.08939 Tags: math.ST Score: 0 Title: Shape-constrained density estimation with Wasserstein projection Domain: math.ST Source: arxiv Authors: Takeru Matsuda, Ting-Kam Leonard Wong Reviews: 0 Subject: math.ST Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-03-09 Page: https://delphi.science/r/arxiv/2603.08939 Statistical inference based on optimal transport offers a different perspective from that of maximum likelihood, and has increasingly gained attention in recent years. In this paper, we study univariate nonparametric shape-constrained density estimation via projection with respect to the $p$-Wasserstein distance, with a focus on the quadratic case $p = 2$. By considering shape constraints given by displacement convex subsets of the Wasserstein space, Wasserstein projection estimation is a convex optimization problem. We focus on two fundamental examples, namely non-increasing densities on…