User:Carl Boettiger/Notebook/Comparative Phylogenetics/2010/04/08

{| width="800"
 * style="background-color: #EEE"|[[Image:owwnotebook_icon.png|128px]] Comparative Phylogenetics
 * style="background-color: #F2F2F2" align="center"|  |Main project page
 * style="background-color: #F2F2F2" align="center"|  |Main project page


 * colspan="2"|
 * colspan="2"|

Research

 * Wrote summary function using bootstrap confidence intervals for the bootstraps of the likelihoods directly.
 * On interpreting the bootstraps of likelihoods directly:
 * 1) The extent to which the distributions themselves are distinct provides an indication of the ability to distinguish between models on the given phylogenetic tree and the actual data.  For instance, if the tree was a star tree and the data produced by either BM or OU1, then all distributions would fall on top of one another.
 * 2) The extent to which the observed likelihood matches those produced by the bootstraps under that model gives an indication of how well that model is fitting the actual data.
 * 3) May also be interesting to look at the bootstrap confidence intervals around theta parameters, to see if they are distinct.

Labrids



 * observed likelihoods of BM & OU1 are indistinguishable from one another but outside the 95% confidence interval of the other distributions. The observed likelihoods lie easily within the distribution produced by those models for both BM & OU1.


 * Observed likelihood for OU2 falls within the bootstrap likelihoods produced under OU2.


 * Observed likelihood for OU3 and OU4 fall significantly below the likelihoods expected by data simulated under these models, suggesting that OU2 is indeed the best fit.

Anoles

 * The observed likelihood scores of the first three models (BM, OU1, OU3) lie within the 95% confidence intervals for those models.
 * The observed likelihood scores of OU4 and OU(LP) lie within the 95% confidence intervals of all models.

Possible topics
1. Comparative Methods Workflow in R -- hackathon 2. Reliability: evaluating the robustness of current methods 3. Future of methods
 * Interfaces between programs
 * Extensibility of existing software
 * Redundancy of approaches
 * User / developer dialog and interface
 * Model choice criteria
 * Robustness to assumptions in tree, etc
 * Bayesian implementations
 * High-dimensional data
 * databases, standards

People

 * Marc Lajeunesse
 * Liam Revell
 * Hilmar Lapp
 * Todd Vision
 * Sam Price
 * Brian O'Meara

Notes and Reading

 * Federal Research Public Access Act, mentioned in Nature article on rise of open access.
 * another Nature article. on open access.
 * I'm mentioned in Science Careers on Open Notebook Science.
 * Article discussed by Jean-Claude Bradley, friendfeed, Nature blogs, qbit, Duncan Hull, Anthony, library, Deepak


 * }