LogCombiner

[phyl...@gmail.com]

Hi Beast-ies,

I am trying to combine two log files with logcombiner, They both have
been run with a relaxed clock using a skyline model on the same
dataset. The only difference between these two is the length of the
MCMC and the frequency of sampling. So for instance Run1 was ran with
a chain length of 50M with sampling every 5000 generations thus
creating 10,000 samples. Run 2 was ran with a chain length of 250M
sampling every 25,000 generations thereby creating 10,000 samples. Now
when I try to combine both of these logfiles in logcombiner it appears
that the program always removes the burning of the first file but
never removes the burnin for the second file. This happens
irrespective of what order the files are added to logcombiner. I have
set the burnin to be the the number of states as in Tracer. So the
respective burnins are 5000000 and 25000000.


Can anyone shed some light on this please,

Many thanks.

[Andrew Rambaut]

Hi,

It may be a bug caused by having different sampling frequencies in the
two files. However, I wouldn't advise combining files at different
frequencies because the ESS calculated on the combined file will be
wrong. I suggest down-sampling the first file and then combining them.
If you process the first run in LogCombiner (on its own) you can give
a resample frequency of 25000 (i.e., the same as the longer second
run). Then combine the result with the second run

See if that solves the problem with the burnin?

Andrew

___________________________________________________________________
Andrew Rambaut
Institute of Evolutionary Biology University of Edinburgh
Ashworth Laboratories Edinburgh EH9 3JT
EMAIL - a.ra...@ed.ac.uk TEL - +44 131 6508624

[phylo tree]

Hi Andrew,

I done this but it still doesnt resolve the burnin problem. Any other suggestions - manually by hand?

[phylo tree]

For people whom may have experienced the same problem as me. The solution is easy just press return when entering the burnin values

---------- Forwarded message ----------
From: phylo tree <phyl...@gmail.com>
Date: Jan 23, 2008 2:32 PM
Subject: Re: LogCombiner
To: Andrew Rambaut <a.ra...@ed.ac.uk >


Excellent that worked brill.

Many thanks Andrew for all your assistance!

On Jan 23, 2008 2:07 PM, Andrew Rambaut < a.ra...@ed.ac.uk> wrote:

You say that it happens irrespective of the order you add the files -
is it always the second file that doesn't have its burnin removed?
When you enter the burnin, are you sure that you are pressing return
so that the value gets entered into the table? If you don't then it
doesn't recognize the value you give - this is a problem with Java in
general.

Andrew

On 23 Jan 2008, at 14:04, phylo tree wrote:

> > never removes the burnin for the second file. This happens
> > irrespective of what order the files are added to logcombiner. I
> have
> > set the burnin to be the the number of states as in Tracer. So the

___________________________________________________________________

[Joseph W. Brown]

Hey,

I have encountered this problem once or twice.  When entering your burnin you need to hit the 'enter' button (i.e. do not simply enter the burnin value and hit 'Run').  It seems silly but it works.  An alternative would be to remove the burnin from each file individually, and then combine them in LogCombiner using a burnin of zero.

Joseph.

______________________________________________________________________________________________

Joseph W. Brown, PhD Candidate

Museum of Zoology, Bird Division

3015 Ruthven Museums Building

1109 Geddes Avenue

University of Michigan, Ann Arbor MI 48109-1079

Fax: (734) 763-4080

Email: jose...@umich.edu

Homepage: http://www-personal.umich.edu/~josephwb/

Biology 162 Laboratory: http://www-personal.umich.edu/~josephwb/Biol162/Lab.html

Queen's Conservation Genetics Group: http://biology.queensu.ca/~cgg

Tree of Life Web Project (Neornithes): http://tolweb.org/tree?group=Neornithes&contgroup=Aves

Family page: http://goghbirth.blogspot.com/