Odd BMT tracer file

[Amanda Santiago]

Hi guys, 

I've run *BEAST using BModelTest (transitionTransversionsplit) under a 3.5 million chain length. All ESS values looks ok but BMT model indicator looks odd. Does someone know if this is ok or not ? I can't understand this graph

thanks in advance, 

Amanda

[Alexei Drummond]

Looks fine to me. ESS is high. It is sampling discrete values rather than continuous so only about 30 values are possible. That is probably why is looks a little different from the traces you are accustomed to looking at.

Alexei

--
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.
<Captura de tela 2018-07-21 18.52.33.png>

[Remco Bouckaert]

HI Amanda,

It is probably more useful to look at the indicator under the ‘estimates’ tab, where it will show which models are sampled most often.

Alternatively, you can visualise the distributions using the BModelAnalyser (see https://github.com/BEAST2-Dev/bModelTest/wiki under post-processing).

Cheers,

Remco

[Amanda Santiago]

Hi Remco and Alexei, 

Thank you for your suggestions and help! 

Cheers, 

Amanda

[Alexei Drummond]

Hi Amanda,

One other thing: generally the site model is not a parameter of interest, but a nuisance parameter (i.e. you don’t care about the site model, except that you need to make sure you don’t pick the wrong site model because it might bias the inference of other parameters, like the tree, that you do care about).

The good news about bModelTest is that it is not so much a model test, as an implementation of model averaging. If you only actually care about the tree then (beyond making sure bModelTest parameters are mixing) you don’t need to look at the output of bModelTest at all. You can just look at the phylogenetic output safe in the knowledge that all possible site models have been averaged over.

Cheers

Alexei

On 22/07/2018, at 9:55 AM, Amanda Santiago <asbi...@gmail.com> wrote: