New Bayesian methods for testing independence in large sparse two-way contingency tables
by Professor Murray Aitkin
Abstract: This talk describes a new Bayesian method for testing for independence of row and column classifications in large (or small) sparse two-way contingency tables. It uses the posterior distribution of the likelihood ratio between the independence and the saturated models as the test criterion. This is easily simulated with diffuse priors for any table, and can be interpreted without any asymptotic repeated-sampling or posterior distribution assumption.
Four examples are given of sparse tables of various sizes, including two from simple social networks.
For More Information: for further information please contact Owen Jones: email email@example.com