Marginalization of missing taxa in partitions

[Nathanael Walker-Hale]

Hi guys,

I would just like to clarify RAxML's treatment of partitions in which some taxa have completely missing data.

Assuming a partition including some taxa with completely missing data, does this missing data affect the likelihood of the partition relative to the likelihood calculated independently on the induced subtree (i.e. not marginalizing)?

Best wishes,

Nathanael

[Alexandros Stamatakis]

Hi Nathanael,

No, at least as long as you estimate individual per-partition branch
lengths, i.e., the likelihood of the induced subtree will be identical
to that of the full subtree including the missing data.

This observation actually gave rise to discovering and describing the
phenomenon of terraces in phylogenetic tree space, see, for instance:

https://academic.oup.com/bioinformatics/article/26/12/i132/284771
https://www.ncbi.nlm.nih.gov/pubmed/21680810

Alexis


>
> Best wishes,
>
> Nathanael
>
> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology

www.exelixis-lab.org

[Nathanael Walker-Hale]

Hi Alexis,

Awesome, thanks very much for clarifying.

Best,

Nathanael