Extended Data Fig. 6: Pairwise comparisons of 8 birth-death (BD) branching process models.
From: Prolonged morphological expansion of spiny-rayed fishes following the end-Cretaceous
TESS calculated Bayes Factors (2ln(BF)) from the pairwise comparisons of marginal likelihoods for 8 birth-death (BD) branching-process models with the following priors: 1.) constant diversification rate with uniform (random) sampling, 2.) constant diversification rate with diversified sampling (sampling results in even coverage of all clades), 3.) decreasing speciation rate with uniform sampling, 4.) decreasing speciation rate with diversified sampling, 5.) constant diversification rate with a rate shift 50 Mya (‘Episodic BD’) with uniform sampling, 6.) constant diversification rate with a rate shift 50 Mya (‘Episodic BD’) with diversified sampling, 7.) constant speciation rate with a single mass extinction event occurring at any point in time (‘Mass Extinction BD’) with uniform sampling, and 8.) constant speciation rate with a single mass extinction event occurring at any point in time (‘Mass Extinction BD’) with diversified sampling. Observations of the pairwise comparisons note strong preference for a model that assumes uniform (random) sampling and strong support for either a constant rate BD model, or an episodic BD model that assumes a shift 50 mya. There is moderate Bayesian support (2ln(BF) = 6.4) that among the models assuming uniform sampling, the constant rate model is preferred over the model with a shift 50 Mya.
