ExaML - Bootstrap replicate start trees

[Karen]

Dear all,

recently in some workgroups we discussed whether it is reasonable
a) to use the best tree found in ML treesearches
b) parsimony starttrees (default) or
c) completely random trees

as start trees when bootstrapping.
In additional question to b) in case of parsimony trees (this is the default, correct?) should the parsimony trees drawn from each BS replicate
or from the original dataset / and or is there any recommendation? What is the common setting you (the developers) would use?

Thanks, & best, Karen

[Alexandros Stamatakis]

Hi Karen,

> recently in some workgroups we discussed whether it is reasonable
> a) to use the best tree found in ML treesearches

That.s probably too biased, so I would never just use one tree as
starting tree.

> b) parsimony starttrees (default) or
> c) completely random trees

I think for BS it doesn't matter that much as you already have a
randomization component via the bootstrap re-sampling itself.

If a single BS replicate however always yields the same parsimony tree
(which can happen for datasets with strong signal), I'd then rather go
for random starting trees.

> as start trees when bootstrapping.
> In additional question to b) in case of parsimony trees (this is the
> default, correct?)

Not really the default as starting trees are generated in a
pre-processing step, thus there is no real default.

> should the parsimony trees drawn from each BS replicate
> or from the original dataset / and or is there any recommendation? What
> is the common setting you (the developers) would use?

the parsimony tree should be computed on the BS replicate.

Alexis

>
> Thanks, & best, Karen
>
> --
> 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

[Karen]

Hi Alexis,

thanks a lot for the fast reply and the helpful comments !

Best Karen