A critique of Thompson and Ramírez-Barahona (2023) or: how I learned to stop worrying and love the fossil record
Data files
Mar 19, 2024 version files 3.28 MB
-
raw_smithandbrown.tre.txt
3.28 MB
-
README.md
2.63 KB
Apr 04, 2024 version files 20.08 MB
Abstract
A recent study published in Biology Letters by Thompson and Ramírez-Barahona (2023) argued that, according to analyses of diversification on two massive molecular phylogenies comprising thousands of species, there is no evidence that angiosperms (i.e., flowering plants) were affected by the Cretaceous-Paleogene mass extinction. Here I critique these conclusions from both methodological and philosophical perspectives. I demonstrate that the methods used in their study possess statistical limitations that strongly reduce the power to detect a true mass extinction event using data similar to those analyzed by Thompson and Ramírez-Barahona (2023). Additionally, I use their study as a springboard to examine the relationship between phylogenetic and fossil evidence in diversification studies.
README: Data and code for: A critique of Thompson and Ramírez-Barahona (2023) or: how I learned to stop worrying and love the fossil record
This dataset contains two software files, both written in R, used to conduct the analyses for the publication. The first, critique_kpg_paper.R, takes one data file as an input: raw_smithandbrown.tre.txt, the seed plant phylogeny published by Smith and Brown (2018). With help from functions provided by several free, open-source R packages, this file particularly contains analyses of which models of evolution best fit the Smith and Brown tree. This was done following the methodology of Thompson and Ramírez-Barahona (2023). The second software file, kpg_simulations.R, simulates trees of similar size and age to the Smith and Brown phylogeny to examine whether the results described in Thompson and Ramírez-Barahona (2023) might be due to Type II error. It conducts similar model-fitting tests as written in critique_kpg_paper.R on trees for which the true generating parameters are known.
Description of the data and file structure
raw_smithandbrown.tre.txt is the only input data file in this repository, and it is only input for critique_kpg_paper.R. It is a file containing a phylogeny of seed plants.
All R code files can be run line-by-line in an integrated development environment like RStudio. One warning is that some of the analyses contained in kpg_simulations.R are computationally expensive, particularly when generating multiple very large simulated phylogenies, and they can take several hours to complete.
The repository contains several supplemental figures for analyses described in the text, as well as the five simulated phylogenies described in the text.
Sharing/Access information
The data contained in this repository can also be accessed on GitHub (link in the Related Works section).
The tree published by Smith and Brown (2018) was derived from GitHub (specifically "GBMB.tre") linked in the Related Works section.
Code/Software
To run the two code files, you will need the following packages (the versions with which analyses were conducted are also listed):
- ape (version 5.6.2)
- TESS (version 2.1.2)
- diversitree (version 0.9.16)
- geiger (version 2.0.10)
- TreeSim (version 2.4)
- paleotree (version 3.4.5)
- dplyr (version 1.1.3)
- TreeTools (version 1.9.1)
- phytools (version 1.2.0)
Methods
Here I examine the conclusions of Thompson and Ramírez-Barahona (2023) by performing a limited number of tests with TESS and CoMET similar to those found in their paper. All tests involve the phylogeny published by Smith and Brown (2018) and simulated trees of similar sizes.
To test the degree to which TESS and CoMET can accurately identify the generating models of simulated trees, I extracted branching times from the tree and created three simulated trees in TESS to which I could fit the data using the following branching process models: constant birth-death, episodic birth-death, and a birth-death model including a mass extinction. The marginal likelihoods of each model were estimated using stepping-stone sampling in TESS, and the relative support of the models was compared using Bayes factors.
To test the degree to which TESS and CoMET can accurately identify mass extinction events in an analysis like that of Thompson and Ramírez-Barahona (2023), I usedTreeSim (Stadler 2011) to simulate phylogenies of similar size and age to the Smith and Brown phylogeny. I estimated diversification rates using the “tess.analysis” function to test whether a mass extinction would be detected by the CoMET model.