Likelihoods incorrectly calculated in BEAST 2.5.2 (using Cipres Portal)

144 views
Skip to first unread message

Pintoneta

unread,
Apr 27, 2019, 12:01:34 PM4/27/19
to beast-users
Hi all,

I'm running a very simple XML in BEAST 2.5.2 (produced with BEAUti 2.5.2), and I'm frequently seeing warnings saying "Likelihood incorrectly calculated". I'm using CIPRES portal.

I don't find where I can include the command line that prohibited the use of BEAGLE.
"
/usr/lib/jvm/java-8-oracle/bin/java -cp ::/home/nonada/.beast/2.5/MM/lib/MM.addon.jar:/home/nonada/.beast/2.5/SA/lib/SA.v2.0.0.jar:/home/nonada/.beast/2.5tarBEAST2/lib/guava-23.0.jar:/home/nonada/.beast/2.5/StarBEAST2/lib/starbeast2.addon.jar:/home/nonada/.beast/2.5/BEAST/lib/beast.jar:/home/nonada/.beast/2.5/BEAST/lib/beast.src.jar:/home/nonada/.beast/2.5/RBS/lib/RBS.addon.jar:/home/nonada/.beast/2.5/BEASTLabs/lib/BEASTlabs.addon.jar beast.app.beastapp.BeastMain -java /path/to/yourBeastFile.xml"

It is possible.? Any help?....

Attaching is the xml file and the output.

Cheers,
Nelsy
infile.xml.zip
STDERR

Remco Bouckaert

unread,
Apr 28, 2019, 4:34:17 PM4/28/19
to beast...@googlegroups.com
Hi Nelsy,

The error message shows that the posterior is a large positive number. What probably happened is that some of the parameters (maybe clock rates or birth rates) became very small or large resulting in numerical instability, which then gave large positive posteriors. You can check the trace log in Tracer to see which parameter escaped to an unrealistic value, and adjust the upper and lower bounds for these parameters to make sure they remain close to a value that is reasonable.

Hope this helps,

Remco

--
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 post to this group, send email to beast...@googlegroups.com.
Visit this group at https://groups.google.com/group/beast-users.
For more options, visit https://groups.google.com/d/optout.
<infile.xml.zip><STDERR>

Astrid Schuster

unread,
May 17, 2019, 6:07:35 AM5/17/19
to beast-users
Hi Remco, 
I am using BEAST v2.5.2 and my XML file also stopped running with the message: 
P(likelihood) = -397697.2506637145 (was -749803.9854690708)  **
P(treeLikelihood.fossil_recentASV3) = -397697.2506637145 (was -749803.9854690708)  **
At sample 0
Likelihood incorrectly calculated: -750076.7332999212 != -397969.99849456485(-352106.7348053563) Operator: SAExchange(SAWideFBD.t:fossil_recentASV3)
java.lang.UNIXProcess$ProcessPipeInputStream@506e1b77

Sample posterior likelihood prior
0 -750076.7332999212 -749803.9854690708 -272.74783085033977


No I am puzzled on what and where I should do the changes in my XML file? 
Any support would be super helpful, 
Many thanks in advance, 
Astrid 





Am Sonntag, 28. April 2019 22:34:17 UTC+2 schrieb Remco Bouckaert:
Hi Nelsy,

The error message shows that the posterior is a large positive number. What probably happened is that some of the parameters (maybe clock rates or birth rates) became very small or large resulting in numerical instability, which then gave large positive posteriors. You can check the trace log in Tracer to see which parameter escaped to an unrealistic value, and adjust the upper and lower bounds for these parameters to make sure they remain close to a value that is reasonable.

Hope this helps,

Remco
On 28/04/2019, at 4:01 AM, Pintoneta <nelsy...@gmail.com> wrote:

Hi all,

I'm running a very simple XML in BEAST 2.5.2 (produced with BEAUti 2.5.2), and I'm frequently seeing warnings saying "Likelihood incorrectly calculated". I'm using CIPRES portal.

I don't find where I can include the command line that prohibited the use of BEAGLE.
"
/usr/lib/jvm/java-8-oracle/bin/java -cp ::/home/nonada/.beast/2.5/MM/lib/MM.addon.jar:/home/nonada/.beast/2.5/SA/lib/SA.v2.0.0.jar:/home/nonada/.beast/2.5tarBEAST2/lib/guava-23.0.jar:/home/nonada/.beast/2.5/StarBEAST2/lib/starbeast2.addon.jar:/home/nonada/.beast/2.5/BEAST/lib/beast.jar:/home/nonada/.beast/2.5/BEAST/lib/beast.src.jar:/home/nonada/.beast/2.5/RBS/lib/RBS.addon.jar:/home/nonada/.beast/2.5/BEASTLabs/lib/BEASTlabs.addon.jar beast.app.beastapp.BeastMain -java /path/to/yourBeastFile.xml"

It is possible.? Any help?....

Attaching is the xml file and the output.

Cheers,
Nelsy

--
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...@googlegroups.com.

Remco Bouckaert

unread,
May 19, 2019, 4:35:50 PM5/19/19
to beast...@googlegroups.com
Hi Astrid,

Are you sampling tip dates? If so, are you using the SampledNodeDateRandomWalker operator instead the TipDatesRandomWalker operator? If not, can you send the XML for closer investigation?

Cheers,

Remco

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.
Visit this group at https://groups.google.com/group/beast-users.
Reply all
Reply to author
Forward
0 new messages