Prior and clock model distributions in BEAUTi2

[Herman Mays]

What is the difference between the distribution settings in the clock model tab in BEAUTi2 for a relaxed lognormal clock and the ucldmean distribution settings in the priors tab?

Thanks in advance.

[Remco Bouckaert]

The distribution in the clock model tab is the distribution used for selecting the rates for the branches, while the distribution in the priors tab is the prior over the mean clock rate.

However, there should not be any distribution shown in the clock model panel. Can you let me know which packages are installed?

Cheers,

Remco

On 4/06/2015, at 6:28 am, Herman Mays <may...@gmail.com> wrote:

What is the difference between the distribution settings in the clock model tab in BEAUTi2 for a relaxed lognormal clock and the ucldmean distribution settings in the priors tab?

Thanks in advance.

--
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 http://groups.google.com/group/beast-users.
For more options, visit https://groups.google.com/d/optout.

[Francesco Cicconardi]

Sorry Remco if I say it... But I think somebody else may experience my same problem... Don't you think?

best,

F

[Herman Mays]

In BEAUTi2 I have had bModelTest, MGSM, MODEL_SELECTION and SubstBMA installed. But I've uninstalled them, or at least thought I did. The distribution sometimes appears as a dropdown called Distr when I select a clock model like lognormal relaxed in the clock tab. I say sometimes because I just went back into BEAUTi2 and now it doesn't appear? A bug perhaps?

Thanks
Herman

[Remco Bouckaert]

I am aware of this problem, but I cannot reproduce it on any of my installations — not v2.2.0, v2.2.1 or v2.3.0 — which makes it hard to diagnose.

So, I try to find out what is causing the problem, and it may be that there is a package that is interfering, or maybe something else is going wrong.

Remco

[Remco Bouckaert]

Yes this is a bug, and I have some trouble reproducing it. 

It would be helpful if you could tell some details on how you set up the analysis — number of partitions, linking of trees, clock models or site models, tree prior, site model details, whether you used the cloning function — so I can reproduce that and see whether I can get the same results.

Thanks, 

Remco

[Herman Mays]

I just tried to create a new xml file in BEAUTi2.2.1 on another machine and the "Distr" appears when I choose a relaxed lognormal model in the clock model panel. I have all the packages uninstalled on this machine but I likely had some installed in the past. I suspect there is something that is left behind from a package that does not get uninstalled?

Do you suspect this will interfere with my xml file? Should I ignore this distribution?

thanks
Herman

[Herman Mays]

StarBeast analysis in BEAST 2.2.1

8 partitions (3 mtDNA alignments and 5 nuDNA). The 3 mtDNA alignments are linked to the same tree model the site and clock models are unlinked.

Yule model. Linear with constant root .

Two of the partitions have a HKY+I site model and the rest are GTR some with I some with G and some both.

The three mtDNA and one nuDNA locus are set to relaxed lognormal clocks and I'm checking the estimate box for all the partitions exact for one for which I have a "known" rate from the literature. The known partition (cytb) I set to a lineage rate of 0.0105 and the rest I estimate with a clock rate of 0.01 for the mtDNA loci and a rate 1/10 of the mtDNA rate for the nuDNA loci (0.001)

I've set the priors to lognormal distributions in some runs and exponential in others. I'm trying to reproduce a result achieved in v1.7.5 and having enormous difficulty replicating that result in v2.2.1. The topology and the relative heights are consistent but the node ages are invariably off by at least two orders of magnitude.

I've attached my latest BEAST2 xml.

thanks
Herman

[Herman Mays]

I'm trying to replicate this xml file in BEAST2. It produces comparable results to a prior analysis on the same data from a colleague. I think he used prior clock.rate and ucldmean distributions that were normal with a lower bound of 0. I changed this to an exponential distribution and got very similar results in v1.7.5 but can not get the same result in BEAST2.

Thanks
Herman

[Graham]

When I load BEAST2_analysis36.xml into Beauti 2.2.0, it has 1.05 not 0.0105 for the clock rate for cytb in the Clock Model tab.