Parsimony-based deterministic approximation of likelihoods in population genetics

In this work we investigated how corner cutting works for a popular model of sequence evolution, namely the infinite sites model. We started our investigation from an ideal (and as such impractical) parsimony criterion based on exact posterior probabilities, then we constructed a practically applicable, simple criterion that empirically performs closely as well as the ideal one for small datasets. The results indicate that although the basic method of corner cutting does not provide a close approximation to the likelihood, it may be useful for estimating quantities of secondary interest, e.g. posterior sampling distributions and derivatives of the likelihood with respect to model parameters. We also identified possible improvements and directions for further investigation. References: