Step Size in Stein's Method of Exchangeable Pairs

TL;DR

This paper investigates how the step size in exchangeable pairs, especially from reversible Markov chains, influences the accuracy of normal and Poisson distribution approximations using Stein's method, providing rigorous bounds and insights.

Contribution

It provides the first rigorous analysis of how step size affects error bounds in Stein's method for exchangeable pairs, particularly for normal approximation.

Findings

Smaller step sizes improve normal approximation bounds.

Error bounds relate to the spectrum of the Markov chain.

Principles for Poisson approximation are similar to normal case.

Abstract

Stein's method of exchangeable pairs is examined through five examples in relation to Poisson and normal distribution approximation. In particular, in the case where the exchangeable pair is constructed from a reversible Markov chain, we analyze how modifying the step size of the chain in a natural way affects the error term in the approximation acquired through Stein's method. It has been noted for the normal approximation that smaller step sizes may yield better bounds, and we obtain the first rigorous results that verify this intuition. For the examples associated to the normal distribution, the bound on the error is expressed in terms of the spectrum of the underlying chain, a characteristic of the chain related to convergence rates. The Poisson approximation using exchangeable pairs is less studied than the normal, but in the examples presented here the same principles hold.