Monday

Phylogenetics paper

  • Grayscale the figures

Moore & Coop PBG 270:

Phylo methods for exploring correlations between character traits and lineage rates.

  • Models ((Are the generalized birth-death models (e.g. Kendall 1948) good limit to interesting process? somtimes.  ))

  1. detecting variation in rates

  2. locating shifts along branches

  3. locating shifts in time

  4. correlates of differential rates (_e.g. _traits; our subject for today/course).

  5. Parameter estimation

BiSSE methods

Discussing Maddison 2006, (Maddison, 2006)

An “over-represented traits” may be due to a more rapidly diversifying character or a character with faster transitions into it then out of it.  Foreshadowing comes early: Maddison uses the median to avoid convergence problems influencing the mean… We’ll see a lot of rather flat likelihood surfaces.

Statistical concerns that entangle this:

  • biased sampling of states

  • entangled rates? (to some extent represented by models)

  • Short branches, identifiable issues.

Older methods:

Sister-clade comparisons: how unbalanced does your tree need to be, to have 95% confidence?  40:1 under the birth-death models.  Hmm, not such a good way…  Seems robust, if low power, but Maddison notes that it can also be biased.  Note that under any Markov process, any partition pattern is equally probable (only distribution of times matters).

Writing

References