English

From $p$-Wasserstein Bounds to Moderate Deviations

Probability 2022-05-27 v1

Abstract

We use a new method via pp-Wasserstein bounds to prove Cram\'er-type moderate deviations in (multivariate) normal approximations. In the classical setting that WW is a standardized sum of nn independent and identically distributed (i.i.d.) random variables with sub-exponential tails, our method recovers the optimal range of 0x=o(n1/6)0\leq x=o(n^{1/6}) and the near optimal error rate O(1)(1+x)(logn+x2)/nO(1)(1+x)(\log n+x^2)/\sqrt{n} for P(W>x)/(1Φ(x))1P(W>x)/(1-\Phi(x))\to 1, where Φ\Phi is the standard normal distribution function. Our method also works for dependent random variables (vectors) and we give applications to the combinatorial central limit theorem, Wiener chaos, homogeneous sums and local dependence. The key step of our method is to show that the pp-Wasserstein distance between the distribution of the random variable (vector) of interest and a normal distribution grows like O(pαΔ)O(p^\alpha \Delta), 1pp01\leq p\leq p_0, for some constants α,Δ\alpha, \Delta and p0p_0. In the above i.i.d. setting, α=1,Δ=1/n,p0=n1/3\alpha=1, \Delta=1/\sqrt{n}, p_0=n^{1/3}. For this purpose, we obtain general pp-Wasserstein bounds in (multivariate) normal approximations using Stein's method.

Keywords

Cite

@article{arxiv.2205.13307,
  title  = {From $p$-Wasserstein Bounds to Moderate Deviations},
  author = {Xiao Fang and Yuta Koike},
  journal= {arXiv preprint arXiv:2205.13307},
  year   = {2022}
}

Comments

58 pages

R2 v1 2026-06-24T11:29:31.206Z