Data from: Estimating diversification rates on incompletely-sampled phylogenies: theoretical concerns and practical solutions

Molecular phylogenies are a key source of information about the tempo and
mode of species diversification. However, most empirical phylogenies do
not contain representatives of all species, such that diversification
rates are typically estimated from incompletely sampled data. Most
researchers recognize that incomplete sampling can lead to biased rate
estimates, but the statistical properties of methods for accommodating
incomplete sampling remain poorly known. In this point of view, we
demonstrate theoretical concerns with the widespread use of analytical
sampling corrections for sparsely sampled phylogenies of higher taxonomic
groups. In particular, corrections based on “sampling fractions” can lead
to low statistical power to infer rate variation when it is present,
depending on the likelihood function used for inference. In the extreme,
the sampling fraction correction can lead to spurious patterns of
diversification that are driven solely by unbalanced sampling across the
tree in concert with low overall power to infer shifts. Stochastic
polytomy resolution provides an alternative to sampling fraction
approaches that avoids some of these biases. We show that stochastic
polytomy resolvers can greatly improve the power of common analyses to
estimate shifts in diversification rates. We introduce a new stochastic
polytomy resolution method (TACT: Taxonomic Addition for Complete Trees)
that uses birth-death-sampling estimators across an ultrametric phylogeny
to estimate branching times for unsampled taxa, with taxonomic information
to compatibly place new taxa onto a backbone phylogeny. We close with
practical recommendations for diversification inference under several
common scenarios of incomplete sampling.