subset.twDEMCPops: subset twDEMCPops
In twDEMC: parallel DEMC

Description Usage Arguments Details Author(s) See Also Examples

Condenses MCMC results in an twDEMCPops object to the specified cases.

1
2
3

## S3 method for class 'twDEMCPops'
subset(x, boKeep, ..., iPops = seq_along(x$pops), 
    dropShortPops = FALSE)

`x`	twDEMCPops object
`boKeep`	either logical vector or numeric vector of indices of cases to keep
`...`
`iPops`	integer vector: only these populations are subset, others are kept
`dropShortPops`	if set to TRUE, pops in iPops with less samples than to what `boKeep` refers to are dropped

components thin,Y,nGenBurnin are kept, but may be meaningless after subsetting.

The twDEMCPops class

The class twDEMCPops encapsulates the result of a MCMC run of several populations each consisting of different chains. It is described as result value of twDEMCBlockInt.

The chains within each population have the same length. The chains of different populations, however, may have different length.

The chains within one population are not fully independent, because the proposals are based on a common past chain values. In order to combine each population to a single chain, use stackChainsPop.twDEMCPops. In order to make all the chains the same length and create results as arrays across all populations (as class twDEMC), use function concatPops.twDEMCPops.

There are several methods accessing properties of an object of the twDEMCPops class:

number of generations: getNGen.twDEMCPops
number of samples (only one sample each thinning inteval): getNSamples.twDEMCPops
number of chains: getNChains.twDEMCPops
number of populations: getNPops.twDEMCPops
number of chains per population: getNChainsPop.twDEMCPops
number of parameters: getNParms.twDEMCPops
thinning interval: res$thin
space replicate that the poplation belongs to: getSpacesPop.twDEMCPops
number of space replicates: getNSpaces.twDEMCPops
number of blocks: getNBlocks.twDEMCPops
parameter bounds: getParBoundsPop.twDEMCPops
current temperature of streams: getCurrentTemp.twDEMCPops
temperatue at aggregated level: getCurrentBaseTemp.twDEMCPops

There are several methods to transform or subset the results of the twDEMCPops class.

transforming to array representation across populations of type twDEMC: concatPops.twDEMCPops
select sub-populations: subPops.twDEMCPops
select subset of cases: subset.twDEMCPops (this function)
make populations the same length: squeeze.twDEMCPops
stack all the results of all chains to one big matrix: stackChains.twDEMCPops
thin all the chains: thin.twDEMCPops
combine several twDEMCPops results to a bigger set of populations: combineTwDEMCPops
combine populations for subspaces to bigger populations: stackPopsInSpace.twDEMCPops
combine MarkovChains of each population of a twDEMC to a single chain: stackChainsPop.twDEMCPops

There are several methods to utilize the functions of the coda package for the the twDEMCPops class.

convert an twDEMCPops to a coda mcmc.list as.mcmc.list.twDEMCPops
applying a function to all of the chains: mcmcListApply
stack all the results of all chains to one big matrix: stackChains.mcmc.list
transforming parameters transOrigPopt.mcmc.list

Thomas Wutzler

twDEMCBlockInt

if( FALSE){
	tmp <- subset(res,1:3)
	str(tmp)
	tmp <- subset(res,1:10)  # should produce an error
	#mtrace(subset.twDEMCPops)
	tmp <- subset(res,1:10,dropShortPops=TRUE)  
}