The use of outgroups

[Dylan O'Hearn]

Hello,

From reading the BEAST book and seeing various comments, I understand that the use of outgroups is not recommended, being kind of inconsistent with Bayesian phylogenetics.  On the other hand, I see that outgroups are used very often, even in published papers about BEAST by the creators, where example datasets with outgroups are used for validation. 

So, is there an updated perspective on outgroups?  Or maybe is it just recommended that if you use one, you should also run the same analysis without one to see whether it has an effect?  Just looking for some philosophical perspective on this. 

Thanks! 

[Alexei Drummond]

The main problem with outgroups: the way they are selected and sampled is usually inconsistent with the assumptions of the tree priors. For example the birth-death tree priors assume that you have either complete or random sampling within the clade containing *all* of your taxa. Generally speaking when outgroups are used they are not sampled as densely as the ingroup and therefore the tree containing both the outgroup and the ingroup will not be sampled with the same density in all subclades. Whereas a molecular clock-based inference can provide a root estimate without an outgroup and associated problems. Keeping this in mind, a sensitivity analysis that involves estimating the root both with and without the outgroup is a reasonable option.

Alexei

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/b0184024-5efb-4570-a4b9-fc13735b060dn%40googlegroups.com.