search

arborworkflows / Visomics

ARBOR Revolutionary Workflows for the Tree of Life.
Apache License 2.0
3 stars 1 forks source link

learn more about geiger #22

Closed zackgalbreath closed 11 years ago

zackgalbreath commented 11 years ago
  1. The "OU" model is inconsistent. That is, if you run it multiple times on the same data, you get different results. Maybe there are some other parameters we could set to make it more reproducible?
  2. As far as I can tell, you cannot use the "BM" model to generate an output tree. Search for "transform(" within this document for more info: http://cran.r-project.org/web/packages/geiger/geiger.pdf.
zackgalbreath commented 11 years ago

I had a good chat with Luke & his lab.

He confirmed that the OU model does have some randomness within it. As part of its output, it generates a likelihood score. He said that this shouldn't vary much from one run to another, so that's a good indication of whether or not the code is working as intended. He also indicated that a Brownian Motion output tree would simply be scaled up or down, but they're considering adding "BM" support to transform() anyway.

Anyway, they were very helpful, and willing to field any future questions should they arise.