Student Seminar Series - July 26, 2006
University of Minnesota
School of Statistics
College of Liberal Arts

Application of the Monte Carlo Likelihood Approximation


Yiqun Mou


Wednesday, July 26, 2006
10:00 AM, 127 Ford Hall
Minneapolis, East Bank Campus

Refreshments at 9:30 AM
300 Ford Hall


Abstract


Monte Carlo Markov Chain (MCMC) is a very powerful numerical method of doing probability calculations. There are many probabilities or expectations in statistics that can not be calculated analytically, and MCMC is the main tool that people refer to. Eventually, MCMC will be widely used in all areas of statistics. Likelihood inference is an area where MCMC is very useful. MCMC can be used to calculate the maximum likelihood estimator, likelihood ratio statistics, and Fisher information. Only in simple cases, these statistics can be calculated analytically. In this paper, I will discuss a functional approach of MCMC to the approximation of likelihood function, which is called Monte Carlo likelihood approximation (MCLA). MCLA tries to approximate the whole likelihood function with a single sample and is very convenient to do likelihood inference, which will be illustrated in the examples later.