Skyline fossilized birth-death priors resolve deep root attraction in total-evidence dating
Zhang, Chi; Ronquist, Fredrik; Stadler, Tanja (2022), Skyline fossilized birth-death priors resolve deep root attraction in total-evidence dating, Dryad, Dataset, https://doi.org/10.5061/dryad.573n5tb41
Several total-evidence dating studies under the fossilized birth-death (FBD) prior produced very old age estimates which are not supported by the fossil record, and this phenomenon has been termed "deep root attraction'' (DRA). For the Hymenoptera and Eutheria data, utilizing diversified extant sampling in the FBD process has been shown to greatly reduce the DRA effect. However, it is unclear how to proceed in data analyses where typically both the random and diversified samplings are only rough model approximations of the underlying procedure, while having drastic effects on the results. We investigate a more flexible skyline FBD (SFBD) process where the diversification and fossil-sampling rates can vary over time, and the extant sampling can either be diversified or random, on resolving DRA. The rates among different time intervals can either be independent and identically distributed, or autocorrelated according to an Ornstein-Uhlenbeck (OU) process. By re-analyzing the Hymenoptera and Eutheria data, we show that these variants of the SFBD priors resolve DRA, and also unify age estimates under random and diversified sampling assumptions. The SFBD priors also provide more insights on the diversification dynamics that the constant-rate FBD process unable to recover. This study reveals the importance of relaxing the constant-rate assumption in the FBD prior in total-evidence dating.
1. OUSimTraj.R: code to simulate trajectories of 1000 replicates under the OU process (Fig. 1)
2. *.xml: XML inputs for BEAST2 analyses.
div: diversified extant sampling
rnd: random extant sampling
cst: constant-rate FBD
iid10: skyline FBD with independent and identically distributed rates among 10 time intervals
ou10: skyline FBD with Ornstein-Uhlenbeck smoothing hyperprior rates among 10 time intervals
ou20: skyline FBD with Ornstein-Uhlenbeck smoothing hyperprior rates among 20 time intervals
drs: d-r-s parameterization for the birth, death and sampling rates, instead of direct parameterization
3. *.pdf: Maximum clade credibility trees
same as above