exactLoglinTest: A Program for Monte Carlo Conditional Analysis of Log-linear Models

Nuisance parameters are parameters that are not of direct interest to the inferential question in hand. In a frequentist or likelihood paradigm, a common tool for eliminating nuisance parameters is to condition on their sufficient statistics. The same technique is useful (though rarely used) in a Bayesian settings, as it eliminates the need to put priors on nuisance parameters. For log-linear models, conditional analysis suffers from two main drawbacks.