MS submitted as a Research Paper to Ecography
Ben Weinstein corresponding author, email: benjamin.weinstein@stonybrook.edu Dept. of Ecology and Evolution, Stony Brook University, Stony Brook, New York 11794; USA
Juan Luis Parra, email: juanl.parra@udea.edu.co
Grupo de Ecología y Evolucíon de Vertebrados, Instituto de Biología, Facultad de Ciencias Exactas y Naturales, Universidad de Antioquia, Medellín, Colombia
Catherine H. Graham, email: Catherine.Graham@stonybrook.edu Dept. of Ecology and Evolution, Stony Brook University, Stony Brook, New York 11794; USA
For the greatest transparency and reproducibility, we have included all R files used to create this analysis. The entire repository can be found here: https://github.com/bw4sz/Pred_Obs. For git users, the repo can be cloned
git clone https://github.com/bw4sz/Pred_Obs.git
The supplamental documents are labeled by their filename and their order.
Using the scape function from the R package PEZ, we simulate different occurrence patterns based on models of trait evolution that included phylogenetic signal and/or repulsion in an occurrence trait. We test our result for sensitivity to tree size and shape.
This is the main section of analysis which takes in the species assemblages, localities and runs the niche models (SDMUpdated.R). The predicted habitat suitability maps are used to create predicted environment assemblages at each geographic location for which we have observed assemblage lists. The phylogenetic distance to the closest related species is calculated for each species in each assemblage and parameters are fit using hierarchical bayesian approaches. The posterior estimates are compared to simulated posteriors created in Appendix 2
Evaluation metrics of our ensemble niche models for all species, as well as correlation values among test statistics and modeling methods. We also compared the sensitivity and specificity of species prediction based on thresholds of habitat suitability and found a 0.05 quantile best minimized overprediction and underprediction.
Analysis of the severity of spatial patterns of geographic barriers between sister taxa and time since divergence. We also fit mixed effects models to explore the statistical relevance of this relationship at multiple phylogenetic depths.
Fitting alternative bayesian, glm, and glmmm models to the data and comparing results to randomization assemblages based on tip swapping. These approaches evaluate whether we could arrive at our result based solely on the topology and relatedness of the taxa.
ReadME file you are currently reading orienting the reader to the goals and relavance of each of the appendices.