sdunbarne / gibbssampler

Section 12 of a tentative book on Markov chains with examples of application of Gibbs sampling

GibbsSampler

Section 12 of a tentative books on Markov chains with examples of application of Gibbs sampling.

The Gibbs sampler is an algorithm for generating random variables from a marginal distribution indirectly, without having to calculate the density. Gibbs sampling is based only on elementary properties of Markov chains.

The simple case of a ( 2 \times 2 ) table with multinomial sampling clearly illustrates the Markov chain nature of the process.

The simple case of a bivariate normal table with Gibbs sampling clearly illustrates Gibbs sampling for continuous distributions.

A simple Bayesian model for a spam filter illustrates typical Gibbs sampling.