STA 4276F: Markov chain Monte Carlo algorithms (Fall, 2003)

This course is designed for Ph.D. level students in statistics, who are interested in learning the application and theory of Markov chain Monte Carlo (MCMC) algorithms as they are used in statistics.

Instructor: Professor Jeffrey S. Rosenthal, Department of Statistics, University of Toronto
Sidney Smith Hall, room 6024; phone (416) 978-4594;; 'jeff' at ''

Time: Thursdays 12-2. First class Sept 11. Last class Dec 4.

Place: Ramsey Wright Zoology Building, room 141. [New!]

Textbook: None, but various references will be suggested. (The course notes from a recent Lancaster course, or the slides from a recent introductory lecture, may also be somewhat useful.)

NEW! See the STA4276 notes, and some R commands (to obtain R, visit CRAN).

Content: We will describe such standard MCMC algorithms as the Metropolis-Hastings algorithm and the Gibbs sampler. We will then explore such issues as asymptotic convergence in distribution, geometric ergodicity, quantitative convergence rate bounds, optimal scaling, and perfect simulation algorithms. The issues will be discussed both in terms of theoretical analysis, and in terms of simulation output.

Prerequisites: STA 2111F or permission of the instructor. Students should have a strong background in real analysis and probability theory (including basic measure theory) and stochastic processes (including Markov chains).

Evaluation: See separate sheet.

See also my Markov chain Java applets.

This document is available at