parallel.inits: Parallel RNGs for initial values

In datacloning/dclone: Data Cloning and MCMC Tools for Maximum Likelihood Methods

Parallel RNGs for initial values

Description

This function takes care of initial values with safe RNGs based on parallel.seeds of the rjags package.

Usage

parallel.inits(inits, n.chains)

Arguments

inits	Initial values (see Initialization at jags.model). If NULL, an empty list of length n.chains will be generated and seeded (RNG type and seed).
n.chains	Number of chains to generate.

Details

Initial values are handled similar to as it is done in jags.model.

RNGs are based on values returned by parallel.seeds.

If the "lecuyer" JAGS module is active, RNGs are based on the "lecuyer::RngStream" factory, otherwise those are based on the "base::BaseRNG" factory.

Value

Returns a list of initial values with RNGs.

Author(s)

Peter Solymos, [e-mail]. Based on Martyn Plummer's parallel.seeds function and code in jags.model for initial value handling in the rjags package.

See Also

parallel.seeds, jags.model

This seeding function is used in all of dclone's parallel functions that do initialization: parJagsModel, jags.parfit, dc.parfit

Examples

if (require(rjags)) {
## "base::BaseRNG" factory.
parallel.inits(NULL, 2)
## "lecuyer::RngStream" factory
load.module("lecuyer")
parallel.inits(NULL, 2)
unload.module("lecuyer")
## some non NULL inits specifications
parallel.inits(list(a=0), 2)
parallel.inits(list(list(a=0), list(a=0)), 2)
parallel.inits(function() list(a=0), 2)
parallel.inits(function(chain) list(a=chain), 2)
}

datacloning/dclone documentation built on Jan. 7, 2023, 2:38 p.m.