"Everything should be made as simple as possible, but not simpler." - Albert Einstein

Seminar Announcement

Quadratic forms of Markov Chains with applications

Yves Atchade, University of Michigan

Monday, March 21, 2011

4:00 p.m., room 223, Weber Bldg

ABSTRACT

This talk will present a martingale approximation approach to study the limiting behavior of quadratic forms of Markov chains. The method yields a decomposition of the quadratic form which is analogous to the well known Hoeffding decomposition for U-statistics of i.i.d. data. As application, the talk will focus on the class of lag-window estimators for asymptotic variance estimation in Markov Chain Monte Carlo. These
estimators can be naturally handled as quadratic forms of the underlying Markov Chain. In dealing with lag-window estimators, we will discuss both regular asymptotics results and the so-called "fixed-b" asymptotics results.

This is a joint work with Matias Cattaneo, Dept of Economics, University of Michigan.