Posterior and Likelihood show "?", Extremely low Prior, and Turning on partial likelihood scaling to avoid precision loss

[GoogleCodeExporter]

What steps will reproduce the problem?
1. I have a 186 taxa dataset with 3 genes
2. UCLD molecular clock
3. Yule prior, starting tree randomly generated
4. 1 lognormal-distributed calibration prior

What is the expected output? What do you see instead?
The program runs normally. The first rounds of sampling of the chain are fine, 
but after the 3000 or 4000 generations the posterior and likelihood values go 
to "?" and the prior drops excessively (see below). I've left it running up to 
1000000 generations and the values don't change at all.

What version of the product are you using? On what operating system?
# BEAST v1.6.2, Build r4230
iMac 3.1 GHz Intel Core i5, 4 GB Ram, OS 10.6.8

Please provide any additional information below.
# BEAST v1.6.2, Build r4230
# Generated Wed Nov 16 12:33:41 EST 2011 [seed=1321464812158]
state   Posterior       Prior           Likelihood      rootHeight  
    caesareae_dating_EF1a.ucld.mean caesareae_dating_LSU.ucld.mean  caesareae_dating
_RPB2.ucld.mean
0   -164074.3910    -5624.0649      -158450.3260    449.344         1.00000         1.00000     
    1.00000         -
TreeLikelihood, caesareae_dating_LSU.treeLikelihood, turning on partial 
likelihood scaling to avoid precision loss
1000    ?               -123169229488117430000.0000 ?               7.80484E19      1.45468 
        1.06194         0.84590         -
TreeLikelihood, caesareae_dating_EF1a.treeLikelihood, turning on partial 
likelihood scaling to avoid precision loss
TreeLikelihood, caesareae_dating_RPB2.treeLikelihood, turning on partial 
likelihood scaling to avoid precision loss
2000    ?               -123169229488117430000.0000 ?               7.80484E19      1.45468 
        1.06194         0.84590         -
3000    ?               -123169229488117430000.0000 ?               7.80484E19      1.45468 
        1.06194         0.84590         -
.................

Original issue reported on code.google.com by santiago...@gmail.com on 16 Nov 2011 at 5:50

Attachments:

• caesareae_dating2.xml

[GoogleCodeExporter]

This looks at first glance to be an issue with the model specified rather than 
with the software. Have switched it from being a 'Defect' to 'Other'. The root 
height is moving off to some limit suggesting a lack of sufficient calibration 
information.

Original comment by ramb...@gmail.com on 18 Nov 2011 at 7:38

• Added labels: Type-Other
• Removed labels: Type-Defect

[GoogleCodeExporter]

Original comment by ramb...@gmail.com on 18 Nov 2011 at 7:38

• Removed labels: Priority-Medium

[GoogleCodeExporter]

Log strongly suggests lack of proper prior.

Original comment by msuch...@gmail.com on 7 Mar 2012 at 3:59

• Changed state: WontFix