nimbleSCR (version 0.2.1)

sampler_categorical_general: nimble MCMC sampler function for general categorial distributions

Description

The categorical_general sampler operates within nimble's MCMC engine to perform Gibbs sampling for a single node, which must in essence follow a categorical distribution. However, the prior distribution need not be nimble's dcat distribution, but rather can be any (potentially user-defined) distribution which has the same support as a standard categorical (dcat) distribution. Specifically: the distribution must define a discrete random variable, which can only attain values from the set 1, 2, 3, ..., numCategories.

The categorical_general sampler requires one control list argument, named numCategories, which specifies the fixed upper-bound for the range of the random variable.

The categorical_general sampler is designed to be used in nimble's MCMC engine, and can be added to an MCMC configuration object using the addSampler method. See help(configureMCMC) for more information about MCMC configuration objects and adding custom samplers.

Usage

sampler_categorical_general(model, mvSaved, target, control)

Arguments

model

(uncompiled) model on which the MCMC is to be run.

mvSaved

modelValues object to be used to store MCMC samples.

target

node on which the sampler will operate.

control

named list containing an elemente named numCategories, which specifies the upper-bound for the range of the random variable.

Author

Daniel Turek

Examples

if (FALSE) {
## define custom dmy_categorical distribution as a nimbleFunction
dmy_categorical <- nimbleFunction(...)

## nimble model code, using custom-written dmy_categorical distribution
code <- nimbleCode({
  x ~ dmy_categorical(...)
})

## create NIMBLE model object
Rmodel <- nimbleModel(code)

## create MCMC configuration object with no samplers
conf <- configureMCMC(Rmodel, nodes = NULL)

## add categorical_general sampler to MCMC configuration
conf$addSampler(target = 'x', type = 'categorical_general', control = list(numCategories = 10))

## build MCMC algorithm
Rmcmc <- buildMCMC(conf)

## compile model and MCMC, run MCMC algorithm
}