Rigorous confidence bounds for MCMC under a geometric drift condition

Abstract

We assume a drift condition towards a small set and bound the mean square error of estimators obtained by taking averages along a single trajectory of a Markov chain Monte Carlo algorithm. We use these bounds to construct fixed-width nonasymptotic confidence intervals. For a possibly unbounded function $f:\stany \to R,$ let $I=\int_{\stany} f(x) π(x) dx$ be the value of interest and ${\hat{I}}_{t,n}=(1/n){\sum }_{i=t}^{t+n-1}f(X_i)$ its MCMC estimate. Precisely, we derive lower bounds for the length of the trajectory $n$ and burn-in time $t$ which ensure that $$P(|{\hat{I}}_{t,n}-I|\le \epsilon )\ge 1-\alpha .$$ The bounds depend only and explicitly on drift parameters, on the $V-$norm of $f,$ where $V$ is the drift function and on precision and confidence parameters $\epsilon ,\alpha .$ Next we analyse an MCMC estimator based on the median of multiple shorter runs that allows for sharper bounds for the required total simulation cost. In particular the methodology can be applied for computing Bayesian estimators in practically relevant models. We illustrate our bounds numerically in a simple example.

References in corpus (8)

• Batch means and spectral variance estimators in Markov chain Monte Carlo
• Renewal theory and computable convergence rates for geometrically ergodic Markov chains
• Quantitative bounds on convergence of time-inhomogeneous Markov chains
• Optimal Hoeffding bounds for discrete reversible Markov chains
• Explicit error bounds for lazy reversible Markov Chain Monte Carlo
• Nonasymptotic bounds on the estimation error for regenerative MCMC algorithms
• Gibbs Sampling, Exponential Families and Orthogonal Polynomials
• Gibbs Sampling for a Bayesian Hierarchical General Linear Model

Cited by in corpus (11)

• Spectral gaps for a Metropolis-Hastings algorithm in infinite dimensions
• Explicit error bounds for Markov chain Monte Carlo
• Variable transformation to obtain geometric ergodicity in the random-walk Metropolis algorithm
• Markov Chain Monte Carlo Estimation of Quantiles
• Nonasymptotic bounds on the estimation error of MCMC algorithms
• Information Geometry Approach to Parameter Estimation in Markov Chains
• Complexity Results for MCMC derived from Quantitative Bounds
• Hit-and-run for numerical integration
• Error bounds of MCMC for functions with unbounded stationary variance
• On a Metropolis-Hastings importance sampling estimator
• Dimension-independent Markov chain Monte Carlo on the sphere