Protea-traits-physiology-fitness

Data and code for analysis of associations among traits, physiology, and fitness in five species of Protea.

The data and code provided here should allow anyone who is interested to reproduce results reported in the Annals of Botany paper that links to this repository. Note: Some of the scripts were used during development and debugging. The final structure of the running model (represented by the configuration variables in analysis.R) may not allow one or more of the development/debugging scripts to run as is.

Data

• Individual_Data_for_Analysis.csv - The complete set of data used in the analyses consisting of

  • 450 individual lines of data, each corresponding to a different individual plant
  • 4 sites (Observations from McGregor were not included in analyses)
  • 6 species (Observsation on Leucadendron salignum and Protea neriifolia were not included in analyses)
  • 51 variables

• Protea_Trait_Data_Full.csv - Data from Mitchell et al. (The American Naturalist 185:525-537; 2015)

R scripts

• analysis.R - Driver script for path analyses. Variables that affect the analysis that is run are in ALL CAPS near the head of the script. The values in the script as posted are those used to produce results presented in the paper. Other options invoke various configurations used during development and debugging.

• Bivariate_models_R2.R - Script to estimate R2 for bivariate associations.

• loo.R - Script for leave-one-out cross validation of various models explored during development and debugging.

• mr.R - Multiple regression analysis of structural and performance traits. Used to validate path model during development.

• path.R - Path analysis.

• plot.R - Plotting functions to visualize certain aspects of posterior distributions.

• traits.R - A utility module to define the trait sets used and ensure consistency across models.

• Variance_Partitioning.R - Script to estimate variance partitioning.

Stan scripts

• mr.stan - Multiple regression model coded in Stan to match the path model.

• path-hard-center.stan - Path model with hard centering of random effects. See Stan documentation for details on the difference between hard- and soft-centering.

• path.stan - Path model as used in analysis presented in the paper with soft-centering.