Id: arxiv/2611.25007 Url: https://arxiv.org/abs/2611.25007 Tags: math.ST, stat.AP Score: 0 Title: Geometry, Local Minimax Risk, and the Limits of Sample Average Approximation in Linear Stochastic Optimization Domain: math.ST Source: arxiv Authors: Hanbo Yang, Ming Yuan Reviews: 0 Subject: math.ST Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-11-18 Page: https://delphi.science/r/arxiv/2611.25007 We study how the local geometry of a linear stochastic optimization problem determines its statistical difficulty, measured by expected decision regret. The optimal local rate is governed by how fast the support function of the feasible set grows away from its linearization at the true mean; depending on the instance, it can be $n^{-1/2}$, $n^{-1}$, $n^{-2}$, or a non-power rate, and sample average approximation (SAA) attains these rates in a Gaussian mean model. At smooth instances, SAA also attains the classical sharp local minimax constant. We then allow the data distribution to vary in a…