Supplementary code and data for "The paradox of Predictability Provides a Bridge Between 1 Micro- and Macroevolution".

Masahito Tsuboi
Lund University
Web page: www.masahitotsuboi.com
ORCID: 0000-0002-0144-2893

Niklas Hohmann
Utrecht University
email: n.h.hohmann [at] uu.nl
Web page: www.uu.nl/staff/NHohmann
ORCID: 0000-0003-1559-1838

Kjetil L. Voje
ORCID: 0000-0003-2556-3080

Melanie Hopkins
ORCID: 0000-0002-3580-2172

Meghan Balk
ORCID: 0000-0003-2699-3066

Sophie Nilén
ORCID: 0009-0002-0996-0182

Erik I. Svensson
ORCID: 0000-0001-9006-016X

Lee Hsiang Liow
ORCID: 0000-0002-3732-6069

Gene Hunt
ORCID: 0000-0001-6430-5020

R (version >= 4) and the RStudio IDE

Download the code from GitHub. In RStudio, open the file paradox_of_predictability.Rproj. This open the RProject, installs the renv package (if it is not already installed), and set the working directory correctly. Next, run

renv::restore()

to install all required packages in the correct version. Now you are set up to interact with the code and data.

Then you can inspect and run the code in the directory code/. The main analysis is contained in the script code/analysis.R.

• code : folder with R code
• data : folder with raw and modified data
  • contains various files with raw time series data, named after publication and year
  • output : folder with data extracted from time series
  • var_div_dat : folder with data extracted from time series, with added data on timescale (paleo or neo). This data was added manually
• figs : folder for figures
• renv : renv package folder
• .gitignore : untracked files
• .Rprofile : R session settings
• LICENSE : Apache 2.0 license file
• paradox_of_predicability.Rproj : R Project file
• README.md : readme file
• renv.lock : lock file for renv package

Bibliographical references of the data are shown in Table S1 of the publication.

Raw data is stored in .RData files as list (of lists). The first layer of the list contains 10 variables of the following name and meaning.

$nsamp : number of time steps $nvar : number of variables $M : mean values of variables at each time step $S : list of phenotypic variance-covariance matrices (P-matrices) at each time step $nn : sample size at each time step $tt : time lapse between steps (the first entry is 0) $time.units : unit of time laper. Either "Myr" for million years or "yr" for years. $taxon : taxon name (Genus and species) $sex : sex. Either "male" or "female". $reference : bibliographical reference

Note that the unit of raw data is measurement of size and shape in the ratio scale (length and area) except for the whorl count (ordinal) in Bellamya (see Table S1). Whorl count however essentially reflects length in the ratio scale, and thus treated here as a ratio scale trait. In all cases, traits in the original scales were natural-log transformed before estimating the P-matrix. Values presented in "$S" are therefore are mean-scaled variances and covariances, which are unit-less elasticities in percentage of the mean (see Hansen and Houle 2008, Voje et al 2023 for measurement theory behind this operation). For the trait units and further details about each measurements, please consult original sources.

Apache 2.0, see LICENSE file for details.

This work was funded by

1. The Center for Advanced Studies in Norwegian Academy of Science and Letters to Thomas F. Hansen and Christophe Pélabon and
2. The Japanese Society for the Promotion of Science, Overseas Research Fellowship to Masahito Tsuboi (Grant identification number: 202270014)