- Maximum likelihood assignment based on arbitrary number of partitions now completed. Working on stochastic assignment.
- Now have functions for:
- partitioning data into k groups,
- inferring an ancestral state painting based on this partition
- inferring a continuous model with k OU regimes based on this painting
- need to add my function for the proper model-comparison bootstrap of likelihood ratio values rather than the likelihoods themselves. See my original entry on this approach.
Currently running the bootstraps on the inferred models for two peaks and three peaks compared to a single partition under either BM or OU.
As three peaks includes 2 peaks as a subset, it necessarily performs better. However, the relative increase is much smaller. Comparing likelihood ratios should help tease this apart further.
Discussed target papers for this research.
- Target a short paper focusing on the biological aspects of identifying two peaks in the Labrid data. Working on with Peter. Will probably add the three peak comparison (running now).
- Develop a commentary on phylogenetic signal and bootstrapping methods
- Full manuscript with the complete inference of multiple niches that integrates over uncertainty