Wasserstein-p Bounds via Cumulant-Based Edgeworth Expansion for $α$-Mixing Random Fields
Tianle Liu, Morgane Austern
math.PR
Sep 13, 2023 · v6
TL;DR
A Lean 4/Mathlib companion machine-checks the main theorem interfaces and the full genogram argument for the Edgeworth expansions, relative to seven stated literature inputs.
Abstract
We establish normal-approximation bounds in the Wasserstein-$p$ distance for $α$-mixing random fields, extending results previously available mainly for independent or locally dependent variables. For standardized sums indexed by $T\subset\mathbb{Z}^d$, we obtain rates $O(|T|^{-β})$, where $β\in(0,1/2]$ depends on $p$, the moment order, the dimension, and the mixing decay; the optimal square-root rate holds under sufficiently fast polynomial mixing. For integer $p\geq2$, a refined bound retains the stronger mixing power produced by the expansion, covers the endpoint of finite $(p+2)$-th moments, and identifies exact logarithmic corrections at critical decay. Under finite-range dependence, the endpoint moment condition gives the optimal rate for every real $p\geq1$. The proof combines a cumulant-based Edgeworth expansion with Stein's method. Its main technical tool is a constructive graph calculus that organizes high-order dependence and preserves cancellations before mixing bounds are applied. The resulting Wasserstein estimates also imply two-regime non-uniform normal-approximation bounds with sharper sample-size decay in the far tail. A Lean 4/Mathlib companion machine-checks the principal theorem interfaces and the complete paper-internal genogram argument for the expansions, relative to seven explicitly identified literature and standard inputs.
Problem
Normal-approximation bounds in Wasserstein-p distance were mainly available for independent or locally dependent variables. The goal is to extend them to standardized sums of α-mixing random fields indexed by subsets of Z^d.
Approach
The proof combines a cumulant-based Edgeworth expansion with Stein's method. Its main tool is a constructive graph calculus (genograms) that organizes high-order dependence and preserves cancellations before mixing bounds are applied. A Lean 4/Mathlib companion machine-checks the principal theorem interfaces and the paper-internal genogram proof chain, assuming seven explicitly identified literature and standard inputs.
Results
Rates O(|T|^{-β}) are obtained, with β depending on p, the moment order, the dimension and the mixing decay; the optimal square-root rate holds under fast polynomial mixing. For integer p≥2 a refined bound covers the finite (p+2)-th moment endpoint with exact logarithmic corrections at critical decay. The estimates also give two-regime non-uniform normal-approximation bounds.