-Inf in BD serial (BEAST v1.10.4)

48 views

Skip to first unread message

Artem B

unread,

Dec 6, 2023, 2:56:34 AM12/6/23

to beast-users

Hi there! I try to run analysis of a simulated dataset with the Birth-Death serial model but get -Inf and don’t know what BirthDeathSerialSamplingModel/bdss(treeModel)=-Inf means. What prior should I correct?

The individual components of the posterior are as follows: The initial posterior is zero: CompoundLikelihood(compoundModel)=( LogNormal(kappa)=-1.8654, CTMCScalePrior(ctmcScalePrior)=-101.9554, LogNormal(bdss.birthRate)=-1.5466, Uniform(bdss.relativeDeathRate)=0.0, LogNormal(bdss.psi)=-1.3046, LogNormal(bdss.origin)=-4.004, BirthDeathSerialSamplingModel/bdss(treeModel)=-Inf, StrictClockBranchRates(strictClockBranchRates[branchRates])=0.0 Total = -Infinity ), CompoundLikelihood(compoundModel)=( TreeDataLikelihood(TreeDataLikelihood[treeLikelihood])=-132649.8762 Total = -132649.8761731562 ) Total = -Infinity

Artem B

unread,

Dec 6, 2023, 8:57:44 PM12/6/23

to beast-users

setting the initial Origin value above 1.000.000 iterations solves the problem :)