Introduction
Welcome.
This is a BEAST 2 package for comparing two models using the Model Switch Integration method outlined in
Computing Bayes Factors Using Thermodynamic Integration by Nicolas Lartillot and Hervé Phillipe.
It is used to estimate the Bayes factor between two competing models.
About the method
This method progressively moves the analysis from using one model to using the other, ie. it "switches"" between them slowly.
In this package, we take the idea of progressively changing between the models further, by moving incrementally at every step of the MCMC chain.
The advantage of this method is that it works in situations where sampling from the prior is infeasible - meaning that other approaches such as path sampling/stepping stone cannot be used.
One example of situations such as these are structured coalescent models with many demes.
This also gives a much more accurate estimate for the Bayes factor than the harmonic mean estimator or AICM.
News and Updates