fixing the ucld.mean parameter to 1

[Omar]

I ran an analysis where I fixed the ucld.mean parameter to 1 because I
had no calibration sources. Even though I get no dates this way, I
still can see the relative timing of speciation events, right?
However, I would like someone to explain why running an analysis this
way results in time being measured in units of substitutions per site.
Is it because 1 unspecified time unit = 1 substitution per site?
should we better say 1 unspecified time unit = 1 expected substitution
per site?
Thank you for your help. I know I posted the same question a few days
ago, but I got no replies and I need this for a paper I have to
resubmit in just a few days, so I apologize for insisting.

[Aloysius_H]

It doesn't look like you got an answer to this question more than 4 years ago, but I have the same question. The tree BEAST produces when ucld.mean is fixed to 1 (estimate box unchecked) produces an ultrametric tree, not a tree with unequal branch lengths (as in MrBayes or RaXML) in which branch lengths are in substitutions/site. 

[Andrew Rambaut]

In BEAST, normally all trees are time trees (and thus ultrametric). You can get the branch lengths in units of substitutions for any given branch by multiplying its time by its rate (there is an option in BEAUti to create trees with branch lengths of units of substitutions). If you fix the mean rate to 1, then the node heights (or ages) will be the equivalent of the number of substitutions with this mean rate.

Which is more useful to you depends on what you are trying to do.

Andrew

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To post to this group, send email to beast...@googlegroups.com.
To unsubscribe from this group, send email to beast-users...@googlegroups.com.
Visit this group at http://groups.google.com/group/beast-users?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Aloysius_H]

Thanks, Andrew. What I am still struggling with is if I have no absolute time calibrations on any of the nodes  and I wish to produce a tree that reflects best estimates of the relative splitting times, does it really make sense to set ucld.mean to 1 as some have suggested in comments to this group, (particularly if there is obvious rate heterogeneity across the tree)?  I want branch lengths to be proportional to time, not to number of substitutions.

[Andrew Rambaut]

It doesn't matter what value you use, if you fix the rate you will have the same relative node ages. You can rescale the trees and root height parameter post hoc with some rate. It is really just convenience to use the value 1 if you don't have an a priori rate to use. The relaxed clock still allows relative rates to vary across the tree but constrains the mean to be 1. 

Andrew 

To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.

To post to this group, send email to beast...@googlegroups.com.

[Aloysius_H]

Gotcha...that's helpful..thanks!