BEAST v2.7.8 Documentation: beastlabs.inference.IndependentMCMC

Entry point for running a Beast task, for instance an MCMC or other probabilistic analysis, a simulation, etc.
MCMC chain. This is the main element that controls which posterior to calculate, how long to run the chain and all other properties, which operators to apply on the state space and where to log results.
Coupled MCMC. Two independent MCMC chains which will only stop after their clade posteriors have convergedNote that log file names should have $(seed) in their name so that the first chain uses the actual seed in the file name and all subsequent chains add one to it.Furthermore, the log and tree log should have the same sample frequency.

Reference:

Bouckaert, Remco, Timothy G. Vaughan, Joëlle Barido-Sottani, Sebastián Duchêne, Mathieu Fourment, Alexandra Gavryushkina, Joseph Heled, Graham Jones, Denise Kühnert, Nicola De Maio, Michael Matschiner, Fábio K. Mendes, Nicola F. Müller, Huw A. Ogilvie, Louis du Plessis, Alex Popinga, Andrew Rambaut, David Rasmussen, Igor Siveroni, Marc A. Suchard, Chieh-Hsi Wu, Dong Xie, Chi Zhang, Tanja Stadler, Alexei J. Drummond BEAST 2.5: An advanced software platform for Bayesian evolutionary analysis. PLoS computational biology 15, no. 4 (2019): e1006650.

doi:10.1371/journal.pcbi.1006650

Inputs:

checkEvery, cladeprob, convergedFor, tempDir, treeStorer, rhatLogger, rhat, ESS, chainLength, state, init, storeEvery, preBurnin, numInitializationAttempts, distribution, operator, logger, sampleFromPrior, operatorschedule

 

checkEvery
type: java.lang.Integer
number of samples in between checking for convergence
Optional input. Default: 100000

 

cladeprob
type: java.lang.Double
maximum difference in clade probability required to declare convergence
Optional input. Default: 0.15

 

convergedFor
type: java.lang.Integer
maximum difference in clade probability must be less than threshold this many checks in a row before convergence is declared and the chains are stopped
Optional input. Default: 3

 

tempDir
type: java.lang.String
directory where temporary files are written
Optional input. Default: /tmp/

 

treeStorer
type: beastlabs.inference.TreeStoreLogger***
list of tree loggers to check convergence with
Optional input

 

rhatLogger
type: beast.base.inference.Logger
The logger which will be used for computing Rhat (optional)
Optional input

 

rhat
type: java.lang.Double
The maximum acceptable value of Rhat (across all parameters in the log file) required to declare convergence (set to 0 to skip this step)
Optional input. Default: 1.05

 

ESS
type: java.lang.Double
The minimum ESS of all values in either chain required to declare convergence (set to 0 to skip this step)
Optional input. Default: 200.0

 

chainLength
type: java.lang.Long
Length of the MCMC chain i.e. number of samples taken in main loop
Forbidden: must not be specified

 

state
type: beast.base.inference.State
elements of the state space
Optional input

 

init
type: beast.base.inference.StateNodeInitialiser***
one or more state node initilisers used for determining the start state of the chain
Optional input

 

storeEvery
type: java.lang.Integer
store the state to disk every X number of samples so that we can resume computation later on if the process failed half-way.
Optional input. Default: -1

 

preBurnin
type: java.lang.Integer
Number of burn in samples taken before entering the main loop
Optional input. Default: 0

 

numInitializationAttempts
type: java.lang.Integer
Number of initialization attempts before failing (default=10)
Optional input. Default: 10

 

distribution
type: beast.base.inference.Distribution
probability distribution to sample over (e.g. a posterior)
Required input

 

operator
type: beast.base.inference.Operator***
operator for generating proposals in MCMC state space
Optional input

 

logger
type: beast.base.inference.Logger***
loggers for reporting progress of MCMC chain
Required input

 

sampleFromPrior
type: java.lang.Boolean
whether to ignore the likelihood when sampling (default false). The distribution with id 'likelihood' in the posterior input will be ignored when this flag is set.
Optional input. Default: false

 

operatorschedule
type: beast.base.inference.OperatorSchedule
specify operator selection and optimisation schedule
Required input