General

Implement SIMMAP in R
SIMMAP creates a stochastic simulation of discrete character trait evolution on a phylogenetic tree. This is a realization from the Bayesian posterior distribution of possible transitions given the tip data and the tree.
Desired functions:
A) Stochastic simulation under a given model
B) Infer the stochastic modelNotes:
a) Should implement an arbitrary size finite state Markov model
b) Should provide fast visualization or realizations3 votes 
Ancestral state reconstruction under OU models
Currently ancestral state reconstruction for continuous characters is done only in Brownian motion. OUCH will estimate the root node but not the ancestral states of all internal nodes under OU and multiOU models.
1 vote 
Tree painting should be inferred by the model
Brownie and OUCH both take a "painted tree" to indicate which branches of the phylogeny operate under the same regime (assigned the same color). Different colors indicate different selective regimes. simmap can be used to infer possible paintings, (as can parsimony reconstruction), but it would be preferable to have the painting inferred as part of the model fitting of the OU models, directly from the data.
0 votesThis is one of the primary goals of my thesis proposal. The full approach is outlined in the notebook, involving calculating the joint probability of the hybrid discretecontinuous model directly. Meanwhile, I’m attempting an inverse approach that generates a distribution of paintings first, then does the continuous inference.

Partition continuous data into discrete groups intelligently
It is often desirable to divide continuous data into multiple discrete classes to explore certain hypotheses and leverage certain methods that can't be applied directly to continuous data. This is often done relatively arbitrarily.
A simple version of the method would break the data into a userdetermined number of groups. A more sophisticated approach would attempt to estimate the number of clusters. By using modelbased approaches, this could be done in a maximumlikelihood or Bayesian manner.
0 votesDividing data into K partitions by maximum likelihood is now implemented. Also implemented a bootstrap function for the partitioning.
 Don't see your idea?