COMPARING METHODS FOR TIME-SCALING TREES OF EXTINCT TAXA

As phylogenetic approaches to paleobiology become increasingly common, it becomes necessary to assess the effect of methodology on our results. The temporal scaling of phylogenies is important in tree-based analyses that measure rates of trait evolution, reconstruct patterns of lineage diversity or estimate phylogenetic signal of a trait (a quantitative measure of fit to a Brownian model of trait evolution). In the recent paleobiological literature, various routines have been applied for scaling the phylogenetic branches to time. The most basic method for time-scaling a tree of extinct taxa produces many zero-length branches as an artifact and several alternatives have been suggested to deal with this. One alternative involves the use of a morphological clock, but this applied to phylogenies for which there is no information about character changes along branches, such as supertrees. Comparing time-scaling methods for such trees is thus especially important.

I present a new method for time-scaling phylogenies based on a stochastic weighted sampling procedure. This method uses estimates of sampling rates in the fossil record to produce large samples of time-scaled trees which bracket the uncertainty in branch lengths. Analyses are done across samples of trees to account for this uncertainty, rather than analyzing a single time-scaled tree. This new method also resolves soft polytomies and allows for the possibility of ancestor-descendant relationships among taxa.

In this study, I investigate the effect of this and other time-scaling methods on analyses of rates, phylogenetic signal and diversity. Using simulated fossil records and corresponding phylogenetic datasets, I produce time-scaled trees under various time-scaling methods, apply the analysis of interest and measure the fidelity of the result with the ‘true’ answer from the original simulation. In addition, I will also apply the new time-scaling method to real phylogenetic data in order to demonstrate the utility of this method in general.