A0230
Title: Lugsail lag windows and their application to MCMC
Authors: Dootika Vats - Indian Institute of Technology, Kanpur (India) [presenting]
James Flegal - University of California - Riverside (United States)
Abstract: Lag windows are commonly used in the time series, steady-state simulation, and Markov chain Monte Carlo literature to estimate the long-range variances of estimators arising from correlated data. We propose a new lugsail family of lag windows specifically designed for improved finite sample performance. We use this lag window for weighted batch means and spectral variance estimators in Markov chain Monte Carlo simulations to obtain strongly consistent estimators that exhibit positive first-order bias and are asymptotically unbiased. This quality is beneficial when calculating effective sample size and using sequential stopping rules to help avoid premature termination. Further, we calculate the bias and variance of lugsail estimators and demonstrate little loss compared to other estimators. The finite sample properties of lugsail estimators are studied in various examples.
