How Well Does Your Phylogenetic Model Fit Your Data?

Area of Science:

• Statistical Sciences
• Phylogenetic Inference
• Molecular Phylogenetics

Background:

• Goodness of fit testing is fundamental in statistics for evaluating model-data alignment.
• Phylogenetic model goodness of fit has received less attention than model selection, despite its importance.
• Existing phylogenetic goodness of fit tests are often flawed due to model errors and data complexities.

Purpose of the Study:

• To emphasize the necessity of goodness of fit measures in phylogenetic analysis.
• To advocate for the development of "useful" tests that diagnose inadequate model-to-data fits.
• To guide the adaptation or rejection of phylogenetic models based on fit assessment.

Main Methods:

• Discussion of proposed and existing goodness of fit methods within maximum likelihood and Bayesian frameworks.
• Analysis of the strengths and limitations of current approaches.
• Identification of critical open statistical problems in phylogenetic goodness of fit assessment.

Main Results:

• Current goodness of fit tests in phylogenetics are often inadequate, leading to their neglect.
• There is a critical need for diagnostic tests that explain *why* a model-to-data fit is poor.
• Effective goodness of fit assessment is vital for refining phylogenetic models or rejecting unsuitable classes.

Conclusions:

• Goodness of fit assessment is an underdeveloped but critical area in phylogenetic inference.
• Developing "useful" diagnostic tests will enhance the reliability of phylogenetic models.
• Further research is encouraged to address the statistical challenges in this field.