Notes 20100727 114043 CS 798 Split Network Parsimony Talk Continued

From SnOwy - Ed's Wiki Notebook

Split networks -- so why?

• experiment
  • generative model
• for a quartet where nearest neighbours have differing length instead of identical lengths
  • create a MSA {i, j, k, Jukes-Cantor}; where i, j, k are the endge lengths
  • run
    1. ML
    2. BMCMC
    3. MP
• can each of the above methods perform when nearest neighbours have segments that are evolving at different rate?
  • parsimony: long branch attraction problem
  • Bayesian, likelihood too have problems: faulty assumption.
• as sequence length gets longer, the Hill curve against proportion of correct models gets steeper

Split decomposition

• works using a distance matrix
• ...