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

Continue of the issue 1153 for the unresolved promblem #1154

Closed biozhajie closed 1 year ago

biozhajie commented 1 year ago 
          > Hi Jie, This likelihood seems to be less numerically stable than the standard one. This line sets the initial mean of the relaxed clock model to 1. I think this what you’re looking for. `<parameter id="ucld.mean" value="1.0" lower="0.0"/>`

best, JT

Dear @jtmccr1

Thanks your reply very much! I have changed the initial value of ucld.mean to 0.01, but there was no effect to solve the error, whereas if I changed the value of clock rate to 0.01 under strict clock model, the run will be ok. Need your further advices.

The initial value of ucld.mean set to 1.0, the error promption is as follows:

The individual components of the posterior are as follows: The initial posterior is zero: CompoundLikelihood(compoundModel)=( Gamma(ucld.mean)=-4.6152, Gamma(ucld.stdev)=-4.6152, Gamma(skygrid.precision)=-4.6139, GMRFSkygridLikelihood(gmrfSkyrideLikelihood[skygrid])=-15463.4905, ContinuousBranchRates(continuousBranchRates[branchRates])=0.0 Total = -15477.334777110274 ), CompoundLikelihood(compoundModel)=( ThorneyTreeLikelihood(thorneyTreeLikelihood[treeLikelihood])=-Inf Total = -Infinity ) Total = -Infinity

The initial value of ucld.mean set to 0.01, the error promption is as follows:

The individual components of the posterior are as follows: The initial posterior is zero: CompoundLikelihood(compoundModel)=( Gamma(ucld.mean)=-4.6053, Gamma(ucld.stdev)=-4.6152, Gamma(skygrid.precision)=-4.6139, GMRFSkygridLikelihood(gmrfSkyrideLikelihood[skygrid])=-15463.4052, ContinuousBranchRates(continuousBranchRates[branchRates])=0.0 Total = -15477.239554256623 ), CompoundLikelihood(compoundModel)=( ThorneyTreeLikelihood(thorneyTreeLikelihood[treeLikelihood])=-Inf Total = -Infinity ) Total = -Infinity

Best,

Jie

Originally posted by @biozhajie in https://github.com/beast-dev/beast-mcmc/issues/1153#issuecomment-1610394840

biozhajie commented 1 year ago

Dear @jtmccr1

I have run the data successfully by using the maximum likelihood rooted tree produced under the unre UNREST substitution model by iqtree as the starting tree, while the previous starting tree is manually rooted tree using an outgroup from a unrooted tree which was the maximum likelihood tree produced under the general time reverse substitution model by MEGA X. Is this the key difference to the error?

Best,

Jie