Data from: Improving the robustness of phylogenetic independent contrasts: Addressing abrupt evolutionary shifts with outlier- and distribution-guided correlation

Description

Notes

Methods

Phylogenetic tree simulation: Two types of phylogenetic trees were simulated: balanced trees with fixed topologies and randomly generated trees using a coalescent model. The random trees introduced variability in branching rates to reflect diverse phylogenetic scenarios. Tree sizes included 16,128 and 256 species.

Trait data simulation: Trait data were generated under both Brownian motion (BM) and abrupt evolutionary shift scenarios. For abrupt shifts, two traits (X1 and X2) were simulated with independent evolution except for a significant shift at the root branch. Gradual evolution data were simulated under BM with varying levels of noise.

Statistical analysis: Multiple phylogenetic comparative methods were applied to the datasets, including:

• PIC-OGC: A hybrid framework integrating Pearson and Spearman correlations to handle outliers and non-normal data.
• Robust regression methods (PIC-MM, PIC-L1, etc.).
• PGLS models optimized across evolutionary scenarios (BM, λ, OU fixed/random, EB).
• PGLMM : Phylogenetic generalized linear mixed model.
• MR-PMM: Multi-response phylogenetic mixed model.
• Benchmarks for evaluating true and spurious correlations were constructed using simulation parameters.

Data processing: All simulations and analyses were conducted using R (version 4.1.3). Packages including phytools, ape, phylolm, ROBRTand MCMCglmm were employed for tree generation, PIC calculation, and statistical modeling. The dataset was pre-processed to include raw and processed outputs for reproducibility and ease of use.