A Distributionally Robust Estimator that Dominates the Empirical Average
Statistics Theory
2024-06-21 v4 Statistics Theory
Abstract
We leverage the duality between risk-averse and distributionally robust optimization (DRO) to devise a distributionally robust estimator that strictly outperforms the empirical average for all probability distributions with negative excess kurtosis. The aforesaid estimator solves the robust mean squared error problem in closed form.
Keywords
Cite
@article{arxiv.2402.10418,
title = {A Distributionally Robust Estimator that Dominates the Empirical Average},
author = {Nikolas Koumpis and Dionysis Kalogerias},
journal= {arXiv preprint arXiv:2402.10418},
year = {2024}
}