Related Experiment Videos
The importance of proper model assumption in bayesian phylogenetics
Alan R Lemmon1, Emily C Moriarty
1Section of Integrative Biology, University of Texas, 1 University Station C0930, Austin, Texas 78712, USA. alemmon@evotutor.org
Systematic Biology
|June 19, 2004
Summary
Proper model selection in Bayesian phylogenetics is crucial. Incorrectly specified models, especially those ignoring rate heterogeneity, can significantly bias phylogenetic tree estimates and parameter values.
Area of Science:
- Evolutionary Biology
- Computational Biology
- Phylogenetics
Background:
- Model misspecification is a common issue in Bayesian phylogenetic analyses.
- Accurate phylogenetic inference relies on appropriate statistical models for nucleotide substitution.
Purpose of the Study:
- To investigate the impact of model misspecification on Bayesian phylogenetic analyses.
- To quantify biases in bipartition posterior probabilities and parameter estimates under different model assumptions.
Main Methods:
- Conducted >5,000 Bayesian phylogenetic analyses using six nested nucleotide substitution models.
- Examined biases in bipartition posterior probabilities and parameter estimates across varying branch lengths and model complexity.
Main Results:
- Model misspecification, particularly ignoring rate heterogeneity, strongly biases bipartition posterior probabilities.
- Biases are influenced by branch lengths, with under- and overparameterization causing issues in different phylogenetic zones.
- Parameter estimates like branch lengths and gamma shape are also biased by underparameterization.
Conclusions:
- Researchers must carefully select appropriate models using a priori methods and a posteriori adequacy tests.
- Ensuring model adequacy is vital for reliable phylogenetic inference and parameter estimation in Bayesian analyses.