search

beast-dev / beast-mcmc

Bayesian Evolutionary Analysis Sampling Trees
http://beast.community
GNU Lesser General Public License v2.1
192 stars 73 forks source link

The MCMC sampling are still very poor after 200 million of MCMC samplings #1151

Closed biozhajie closed 1 year ago

biozhajie commented 1 year ago

Dear @msuchard @jtmccr1 @rambaut

I am trying to estimate the local clock rates among each lineages for a time tree, using the random local clock model and the method of ThorneyTreeLikelihood, implemented using BEAST v1.10.5 pre-release of ThorneyTreeLikelihood v0.1.2. But the MCMC sampling are still very poor after 200 million of MCMC samplings. I attached my input xml and the output log files, hoping someone can help fix this issue.

S_HCOV19.log S_HCOV19_SG-thorney.xml.txt

Thanks,

Jie

jtmccr1 commented 1 year ago

Hi Jie,

I believe the poor mixing is probably due to the random local clock model. You might be able to test your hypothesis with a simpler model - potentially a relaxed clock or a priori local clock model.

I'm going to close this since it is not an issue with the BEAST code base. But feel free to follow up over email or the beast user group if you have any further questions.

Best, JT

GuyBaele commented 1 year ago

Hi Jie,

If you are determined to use the random local clock model and want to try to improve its estimation, you may want to have a look at the following pre-print: https://arxiv.org/abs/2105.07119

Best regards, Guy

biozhajie commented 1 year ago

hoping someone can help fix this issue.

Hi Jie,

If you are determined to use the random local clock model and want to try to improve its estimation, you may want to have a look at the following pre-print: https://arxiv.org/abs/2105.07119

Best regards, Guy

Hi @GuyBaele ,

How can I use this model in BEAST v 1.10.5 pre-release of ThorneyTreeLikelihood v0.1.2 ?

Best, Jie