README.md

In cboettig/wrightscape: wrightscape

wrightscape

A fast C code implementation of maximum likelihood calculations for a multi-peak OU process with independent alpha, sigma, and theta parameters for each peak on phylogenetic trees (an extension of the ouch model). Includes R package wrapper as interface to C code. Likelihood calculation presented in http://www.carlboettiger.info/files/phylo-covariance.pdf, later the basis for: Jeremy M. Beaulieu, Dwueng-Chwuan Jhwueng, Carl Boettiger and Brian O’Meara, (2012). Modeling Stabilizing Selection: Expanding the Ornstein-Uhlenbeck Model of Adaptive Evolution, Evolution 66 (8) 2369-2383. doi:10.1111/j.1558-5646.2012.01619.x

File overview

regimes.c: Test the regimes data structure Modify to handle a single regime Implement a transition density calculation where transitions occur at nodes alone <<< Pagel & Meade, Huelsenbeck papers -- what about transitions at nodes?

linear.c: BM not really subset of OU, as alpha = 0 not acceptible OU calculation. Modify the OU model to handle appropriately small alpha with Taylor Expansion or as BM

max likelihood: Implement and test maximum likelihood searches over parameters against existing methods

mcmc.c: Start implementing a basic MCMC solver!

matrix_method.c: nonlinear transition densities? Test the stationary distribution?

Error handling

This should probably be done mostly at the R level, enforcing the strict formats and conventions of the input (i.e. if not doing regimes, all states should be listed at 0. States must be numbered incrementally from 0 to n-1, where n is the number of regimes present.)

STEPPING STONES

• Recursive Joint Prob method using BM, compare to analyticals
• Bayesian OU vs BM in RJMCMC
• Use Pagel's BayesTraits run to "paint tree" with uncertainty, then repeat OUCH analysis with uncertainty in branch color

cboettig/wrightscape documentation built on May 13, 2019, 2:12 p.m.