The Comparative Analysis of Lineage-Pair Traits.

Area of Science:

• Evolutionary biology
• Ecology
• Phylogenetics
• Comparative methods

Background:

• Comparative analyses in ecology and evolution often rely on pairwise lineage data, such as diet niche overlap and reproductive isolation (RI).
• Existing statistical methods struggle to account for the inherent non-independence of these lineage-pair traits, leading to untested assumptions and potential biases.
• The covariance structure of lineage-pair traits, influenced by evolutionary relatedness, has not been explicitly formulated, hindering robust statistical modeling.

Purpose of the Study:

• To develop a statistical framework that accurately models the non-independence of lineage-pair traits arising from phylogenetic signal.
• To create statistically robust methods for analyzing relationships among pairwise-defined variables in evolutionary and ecological studies.
• To provide a user-friendly tool for implementing these advanced statistical approaches.

Main Methods:

• Development of models to describe how phylogenetic signal in characters generates covariance among lineage pairs.
• Incorporation of lineage-pair covariance matrices into modified phylogenetic generalized least squares (PGLS) and a new phylogenetic beta regression.
• Simulation testing to compare the performance of new methods against existing approaches, including node averaging.

Main Results:

• The developed methods, incorporating lineage-pair covariance, significantly outperform previous approaches in simulation tests.
• The heuristic method of node averaging was found to be detrimental to model performance, more so than the non-independence it aimed to correct.
• Re-analysis of empirical datasets, including avian hybridization data, showed improved model fit and revealed stronger relationships between pair age and RI.

Conclusions:

• The study provides a statistically robust framework for analyzing non-independent lineage-pair traits, addressing a critical gap in comparative methods.
• The new methods and the `phylopairs` R package offer a more straightforward and reliable way for empiricists to test evolutionary hypotheses.
• Accurate modeling of lineage-pair covariance is crucial for uncovering reliable patterns and relationships in ecological and evolutionary research.