Fast and Accurate Estimation of Species-Specific Diversification Rates Using Data Augmentation

Diversification rates vary across species as a response to various factors, including environmental conditions and species-specific features. Phylogenetic models that allow accounting for and quantifying this heterogeneity in diversification rates have proven particularly useful for understanding clades diversification. Recently, we introduced the cladogenetic diversification rate shift model, which allows inferring multiple rate changes of small magnitude across lineages. Here, we present a new inference technique for this model that considerably reduces computation time through the use of data augmentation and provide an implementation of this method in Julia. In addition to drastically reducing computation time, this new inference approach provides a posterior distribution of the augmented data, that is the tree with extinct and unsampled lineages as well as associated diversification rates. In particular, this allows extracting the distribution through time of both the mean rate and the number of lineages. We assess the statistical performances of our approach using simulations and illustrate its application on the entire bird radiation.[Birth-death model; data augmentation; diversification; macroevolution.]

Automated summary

Diversification rates vary across species due to environmental conditions and species-specific features. A new inference technique is presented here that reduces computation time by using data augmentation, allowing for the estimation of posterior distribution of tree with extinct and unsampled lineages as well as associated diversification rates. Simulation results demonstrate the statistical performance of this approach, which is applied to the study of bird radiation.