Bayesian inference of total biogeographic history under an event-based model
(English)Manuscript (preprint) (Other academic)
We propose a Bayesian approach to infer the historical biogeography under an event-based model, where the total biogeographic histories are sampled from its posterior probability distribution using Metropolis-coupled Markov chain Monte Carlo. Total histories are stochastically mapped on a phylogeny followed by invoking a biogeographical model, which defines the biogeographical events dispersal, vicariance, persistence and extinction. A hypothetic order of probabilities for these events to happen is applied in the priors of the analysis, where a stick-breaking process is used to pick variables from a flat Dirichlet distribution. In comparison to the two most commonly used methods, the proposed method delivers relatively similar reconstructions, albeit with some differences such as favouring extinctions more. These differences are linked to either the treatment of total histories, or to the fundamental statistical differences of the three approaches. In conculsion, this method favours extinctions more than the compared methods, but the main difference is more complex and is instead linked to the treatment of total histories.
Biogeography, Bayesian inference, stochastic mapping, dispersal, vicariance, persistence, extinction
IdentifiersURN: urn:nbn:se:su:diva-71308OAI: oai:DiVA.org:su-71308DiVA: diva2:484567