Extending a BEAST run arbitrarily

[sammysheep]

Hi,

I am very pleased that one can view log files in tracer while BEAST is
running to see how things are converging. Sometimes it becomes
obvious the chain length is too short for the standard 10% burn-in.
One can alter the burn-in within tracer to make adjustments but at the
loss of samples.

Is there a way to an extend a beast run instead after the initial
run? For example, if I did 100M instances at first, could I, starting
where I left off, append another 50M for a total of 150M? Obviously,
turn off -overwrite, but wasn't sure if this was possible in the
current version.

Thanks for the help!

- Sammy

[Andrew Rambaut]

Hi Sammy,

Turning off -overwrite will not extend the logs - it is just a safety feature to avoid overwriting existing logs accidentally. There is no way to continue a BEAST run once it has finished. However, if you think you only need another 50M to add to your 100M why not simply do another independent run? That way you can compare the two runs to look that they have converged on the same sampling distribution and then combine them.

Generally you always want to run BEAST for many times (>10?) the burnin period or you will not have sufficient independent samples (the burnin will be inversely correlated with the efficiency of mixing).

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.
> For more options, visit this group at http://groups.google.com/group/beast-users?hl=en.
>

[Schoenhofer, Axel]

As you can run BEAST and view log files in the process just specify longer runs and throw away the trees you do not need. I sample 20.000 trees and throw away 15.000 when Tracer results are ok. (you can alter burnin in Tracer to see how that effects convergence and ESS). Sometimes running multiple runs is better as BEAST can start of with a bad tree and then has a hard time converging anyway.
Greetings
Axel

Dr. Axel L. Schönhofer
Department of Biology,
Life Sciences North, Room 230
San Diego State University
5500 Campanille Drive
San Diego, CA 92182-4614
USA
Phone: (00)1-619-594-0717
E-mail: Axel.Sch...@uni-mainz.de

-----Ursprüngliche Nachricht-----
Von: beast...@googlegroups.com [mailto:beast...@googlegroups.com] Im Auftrag von Andrew Rambaut
Gesendet: Dienstag, 18. Oktober 2011 08:57
An: beast...@googlegroups.com
Betreff: Re: Extending a BEAST run arbitrarily