Likelihood-Based Parameter Estimation for High-Dimensional Phylogenetic Comparative Models: Overcoming the...

Area of Science:

• Evolutionary biology
• Phylogenetics
• Quantitative genetics

Background:

• Distance-based multivariate phylogenetic comparative methods are used for high-dimensional traits.
• These methods have limitations including low statistical power and high Type I error.
• They are restricted to simple evolutionary models like Brownian motion.

Purpose of the Study:

• To propose an alternative method for high-dimensional trait evolution analysis.
• To overcome statistical limitations of distance-based methods.
• To provide a robust framework for estimating evolutionary models and handling complex data.

Main Methods:

• Parametric bootstrapping and maximum pseudolikelihood estimation.
• Development of pairwise composite likelihood methods.
• Implementation in the R package phylocurve.

Main Results:

• Pairwise composite likelihood methods demonstrate appropriate Type I error.
• These methods exhibit high statistical power across various evolutionary scenarios.
• The new framework allows for estimation of alternative evolutionary models and handling of missing data.

Conclusions:

• Pairwise composite likelihood methods provide a robust framework for high-dimensional trait evolution.
• This approach overcomes limitations of distance-based methods.
• The phylocurve R package facilitates these advanced phylogenetic comparative analyses.