Difference between revisions of "R Hackathon 1/TransitionProbability"

From Phyloinformatics
Jump to: navigation, search
Line 1: Line 1:
'''Transition Probability Models in R'''
+
'''Modeling Discrete Character Evolution in R'''
  
 
The evolution of a discrete character along a branch of a phylogeny can be model as a Markovian process, where the probability of moving the current state to a different state is governed by a rate matrix.  We can learn about discrete character evolution by calculating and comparing the likelihoods under models which make different assumptions about this matrix.  For example, we might compare a Jukes-Cantor model (where the probability of going from state 0 to state 1 is equal to the probability in the reverse direction) to an asymetrical model where the two rates are allowed to differ.  If we found the latter model was a significantly better fit, we might conclude that the trait in question evolves in a directional fashion.
 
The evolution of a discrete character along a branch of a phylogeny can be model as a Markovian process, where the probability of moving the current state to a different state is governed by a rate matrix.  We can learn about discrete character evolution by calculating and comparing the likelihoods under models which make different assumptions about this matrix.  For example, we might compare a Jukes-Cantor model (where the probability of going from state 0 to state 1 is equal to the probability in the reverse direction) to an asymetrical model where the two rates are allowed to differ.  If we found the latter model was a significantly better fit, we might conclude that the trait in question evolves in a directional fashion.

Revision as of 16:05, 12 December 2007

Modeling Discrete Character Evolution in R

The evolution of a discrete character along a branch of a phylogeny can be model as a Markovian process, where the probability of moving the current state to a different state is governed by a rate matrix. We can learn about discrete character evolution by calculating and comparing the likelihoods under models which make different assumptions about this matrix. For example, we might compare a Jukes-Cantor model (where the probability of going from state 0 to state 1 is equal to the probability in the reverse direction) to an asymetrical model where the two rates are allowed to differ. If we found the latter model was a significantly better fit, we might conclude that the trait in question evolves in a directional fashion.

Estimating transition rates and calculating the likelihood of different models for discrete character evolution can be accomplished with both the geiger and the ape packages in R.

An example using the geiger package