Simple Yet Powerful: The Gibbs Sampler by Catherine Torres ’23, Statistics Colloquium, Wednesday, April 5, 1:10 – 1:50 pm, North Science Building 114, Wachenheim.
Abstract: The Gibbs sampler is a powerful Markov Chain Monte Carlo (MCMC) sampling technique that allows one to sample from a known multivariate probability distribution using the conditional distributions instead of the joint distribution. By sampling from the conditional distributions, Gibbs sampler allows one to sample from distributions that are very difficult to directly sample from. After briefly introducing Markov Chains and their properties, as well as the intuition behind Monte Carlo simulations, we will go through Casella and George’s paper “Explaining the Gibbs Sampler”. This paper takes us through the intuition behind Gibbs sampling as well as its practical application for both Bayesian statistics and classical statistical methods.